PAGE 1
Review of Postal Automation Strategy: A Technical and Decision Analysis June 1984 NTIS order #PB85-127843
PAGE 2
-- Recommended Citation: Review of Postal Automation Strategy: A Technical and Decision Analysis-A Technical Memorandum (Washington, DC.: U.S. Congress, Office of Technology Assessment, OTA-TM-CIT-22, June 1984). Library of Congress Catalog Card Number 84-601072
PAGE 3
OTA POSTAL AUTOMATION STAFF John Andelin, Assistant Director OTA Science, Information, and Natural Resources Division Frederick W. Weingarten, Communications and Information Technologies Program Manager Project Staff Fred B. Wood, Project Director Jean E. Smith, Analyst Administrative Staff Liz Emanuel Shirley Gayheart Jennifer Nelson Renee Lloyd Contractors Friendship Engineering Company William P. Gingras, Principal Investigator Decision Science Consortium, Inc. Jacob W. Ulvila, Principal Investigator
PAGE 4
* .* OTA POSTAL AUTOMATION WORKSHOP PARTICIPANTS March 5, 1984 Robert Buzard ElectroCom Automation Inc. Larry S. Davis University of Maryland Harry A. Dorn ELSAG Inc. (U. S.) William Ferguson Postal Rate Commission William Gingras Friendship Engineering Company Dana Grubb National Bureau of Standards Jacob Rabinow National Bureau of Standards Louis Rubinfield SDC-Burroughs Israel Sheinberg Recognition Equipment, Inc. Richard A. Stotler SDC-Burroughs Robert Tracy Pitney Bowes Jacob W. Ulvila Decision Science Consortium, Inc. Art Hamburgen u. s. IBM (Rochester) M. Wayne Holland Shell Oil Company u. s. Charles W. Hulburt SRI International Lloyd Krause Consultant Jerry W. Loftus Bell & Howell Jeryl Mumpower National Science Foundation Robert P. Otten The Reader's Digest James E. Pehta List Processing Company Matthias Prasser AEG Telefunken General Accounting Office Willis Elmore Robert E. Price Robert Hadley Postal Service James V. Jellison Lynn A. Kidd Harold Glass Robert Krause Ray Morgan Bill Chapp Nick Barranca
PAGE 5
. CONTENTS Page
PAGE 6
Table No. LIST OF FIGURES Page 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. Illustrative Optical Character Reader (OCR) Performance Data .**.**.*, 16a Productivity of Singleand Multi-Line Optical Character Readers .......0 2 la Summary of USPS Equipment Upgrades l ***O*********.****************.**. 3 la Decision Options ..*. ... ..** 0*** *.** **** ***. **** e*** **** ***0 ********0* 34a Uncertainties Included in Analysis of Decision Options l * 0 0 *.*.....* 41a Comparison of Code Usages ..*. .*. =**. *.*. *.. **** e * * *.***..** 42a Growth in ZIP+4 Usage Rate . . * . ***..*..* 43a Estimates of 1989 Clerk/Carrier Savings ... .*. .. ***********O.***********. 44a Usage Savings Factors . . . . . . . . . . . . 46a Estimated Percentage Rates of Return for Decision Options Under Various Conditions ..*.*.**.*..*.****.*.*******.****.***.**. 52a Estimated Net Present Values (in billions of dollars) for Decision Options Under Various Conditions ****...*..*..************ 53a Comparison of Net Present Values (in percentages and billions of dollars) of Options D, G, and H with Option A . . . .54a Eighty Percent Intervals on Net Present Value (Overall and Conditional) *0** **** **0* ***. .**. *.** ***. ..*. ..*. *.. ******0*****.* 54b Cumulative Probability Distributions of Net Present Value (Smoothed) 0*** ***. .*** **** **. *..0 .0. .**O*O* e*** *.. .e. *****.*0.* 56a Sensitivity of net Present Value to Multi-Line OCR Purchase Cost 0*** **** 0*. **** .*. ... .0. **** .*. .**. e ******..* 56b Comparative Net Cash Flows, Selected Options and Conditions, 1985-1998 (in billions of dollars) ..*.*.......**...******.*..**...** 57a Comparative Net Cash Flows, Selected Options, Conditions, and Years (in $ billions) .*. *.. *.*. .*. ..*. **** **. *e*. *e. **e* e* **e.** 57b A-1. Probabilistic Cash Flow Model l ***. m*** .*** 0 **0****..*.*......*********** 60 A-2. Simplified Representation of Continuous Probability Distribution ... ... .*. *.. ..o.O. **. **** . . . . . . . 61 A-3 through A-7. Simplified Schematic Model for Options A through E ... .**. .* .**....*....*.*.**..*.*.***.*.*****.*****.*****.* 62-66 B-1 through B-4. Cash Flow Results for Options A through D . . . ....68-71 LIST OF TABLES B-1 through B-19. Detailed Cash Flow Results ... ... ... ..o. ... ... ... ..abo e.. 72-97
PAGE 7
Foreword This OTA Technical Memorandum on postal automation responds to an October 5, 1983, letter of request from Congressman William D. Ford, Chairman of the House Committee on Post Office and Civil Service, Congressman Robert Garcia, Chairman of the Subcommittee on Postal Operations and Services, and Congressman Mickey Leland, Chairman of the Subcommittee on Postal Personnel and Modernization. The Committee asked OTA to review the United States Postal Service (USPS) decision to utilize single-line optical character readers (OCRs) instead of multi-line OCRs, and to conduct a comparative technical and economic analysis of the two technologies in the context of the overall postal automation program. OTA did not assess the impacts of postal automation in other areas such as labor force requirements, mail processing organization, and privacy and security of the mail. The Senate Committee on Governmental Affairs, Subcommittee on Civil Service, Post Office, and General Services, has also expressed an interest in this subject. [n preparing this memorandum, OTA has drawn on information and analyses from a variety of sources, including the October 1982 OTA report on Implications of Electronic Mail and Message Systems for the U.S. Postal Service. USPS has provided extensive materials on prior USPS automation studies as well as current operational data. The General Accounting Office has provided useful perspectives and analyses based on prior work and on a directly related study conducted in parallel with OTAs. On March 5, 1984, OTA held a postal automation workshop attended by mailers, manufacturers, and researchers who, along with USPS and GAO, participated in a vigorous discussion of key issues. Finally, OTA has benefited greatly from the excellent work of two OTA contractors -Friendship Engineering Company for technical analysis and Decision Science Consortium, Inc. for decision and economic analysis. OTA appreciates the participation of those who helped bring this study to fruition. The memorandum is, however, solely the responsibility of OTA, not of those who so ably advised and assisted in its preparation. Also, the purpose of this memorandum is to provide analysis and evaluation of options available to Congress and USPS, not to make recommendations as to which option(s) should be implemented.
PAGE 8
Overview The United States Postal Service (USPS) is in the midst of implementing a major postal automation program. This program includes acquisition of a large number of optical character readers (OCRs) and bar code sorters (BCSs) along with adoption Of the extended 9-digit ZIP code (known as ZIP + 4). This postal automation program is intended primarily to reduce the amount of labor required to process mail, and secondarily to improve the quality of mail service. Since labor accounts for about 85 percent of total postal costs, reductions in the labor component of mail processing offer the greatest potential to cut current postal costs and restrain future cost increases. USPS has already developed a national ZIP+4 directory, and since October 1, 1983, has been encouraging business mailers to use ZIP+4. Mailers receive a discount of 0.5 cent per piece of ZIP+4 presort first class mail when OCR-readable (can be read by optical character readers) and mailed in batches of 500 or more letters. For ZIP+4 non-presort first class mail, mailers receive a discount of 0.9 cent per piece, when OCR-readable and mailed in batches of at least 250 letters. Use of ZIP+4 is voluntary. At present, very few mailers (59 as of late May 1984) have converted to ZIP+4. Use of ZIP+4 allows USPS to sort letters down to the city block, building, or post office box, whereas the 5-digit zip code permits sorting only to the level of a smaller post office zone or a geographical area within a larger post office zone. The optical character readers are intended to read the ZIP+4 code, translate it into a bar code, and apply the bar code (with an ink jet printer) to the lower right-hand corner of the envelope. From then on, the letter can be sorted automatically by barcode sorters down to the level of carrier routes. All intermediate manual sorting is eliminated. To carry out the automated sorting, USPS has already bought 252 OCRs and 248 BCSs (Phase I of the automation procurement) at a combined cost (including ancillary equipment and installation expense) of $234 million. USPS expects this equipment to be fully installed and operational by the end of 1984. And USPS has received bids on procurement of an additional 403 OCRs and soon will be soliciting bids on an additional 452 BCSs (Phase II of the automation procurement). USPS has allocated $450.2 million for this procurement, of which $363 million is for capital expenditure. The central issue addressed by this OTA technical memorandum is whether the current USPS automation strategy is technically and economically sound, and whether USPS should proceed to actual procurement of this equipment as planned or revise its strategy in whole or in part. OTA concluded that the current postal automation strategy, while technically feasible, is not likely to achieve the greatest projected economic return to USPS when the uncertainty in ZIP+4 usage is taken into account. USPS has based their strategy on achieving 90 percent ZIP+4 usage (among large business mailers) within 5 years, and 27 percent after 1 year. Current estimates indicate that first year (1984) ZIP+4 usage will fall far short of original USPS projections. Based 2
PAGE 9
l on the preponderance of available evidence, OTA concluded that it is quite unlikely that ZIP+4 usage will grow as fast as assumed by USPS. Therefore, while the current USPS strategy of using single-line OCRs would provide an economic return considerably greater than not automating at all, other strategies offer a better return on investment, net present value, and net cash savings than the current strategy, especially if one assumes ZIP+4 usage at the lower range of alternate projections. These other strategies involve extensive use of a competitive technology -the multi-line optical character reader. Whereas the single-line OCR can read only the last line of an address (defined as city, State, and 5-or 9-digit ZIP code), the multi-line OCR can read up to four lines of the address and can process a large amount of 5-digit ZIP mail to the 9-digit level. In other words, the multi-line OCR is not as dependent on use of ZIP+4 to realize savings from automation. OTA concluded that, whereas the multi-line OCR may not have been a technically viable alternative 3 or 4 years ago when USPS made its initial decision to go with singleline OCRs, the multi-line OCR is now fully competitive. OTA found that the multi-line OCR performs as well as the single-line OCR in processing 9-digit ZIP mail, and significantly better than the single-line OCR in processing 5-digit ZIP mail to the 9-digit level. The purchase and/or conversion and maintenance costs of the multi-line OCR are expected to be only marginally higher than the single-line! and the difference is negligible when compared to the additional savings expected over the life of the investment. Based on the results of OTAs cash flow modeling, the strategy offering the greatest economic return to USPS would be for USPS to proceed with the Phase II singleline OCR procurement, but simultaneously initiate release-loan testing (and any necessary related research and development) on singleto multi-line conversion, and then convert all single-line OCRs to multi-line as soon as possible, regardless of the level of ZIP+4 use. OTA has designated this the automatic conversion strategy. Under conditions of high and median ZIP+4 usage, automatic conversion indicates a marginally greater ($40 to $180 million) net present value compared to the single-line OCR strategy. (Note: Net present value was calculated by discounting future cash flows at 15 percent per year.) But under low ZIP+4 usage, automatic conversion shows a substantially greater net present value of $250 to $820 million compared to single-line. As for total net cash flows (undiscounted) over the life of the investment (1985-98), at high ZIP+4 use, savings rate, and multi-line performance, automatic conversion shows a $560 million greater cash flow. All other things being equal, this increases to $790 million at median ZIP+4 usage and a dramatic $3.62 billion at low ZIP+4 usage, compared to single-line. [n the out years (1994-98), under these conditions, automatic conversion shows a greater annual net cash flow in the range of $440 to $580 million. In essence, the substantially greater performance and savings of the multi-line OCRs with non-ZIP+4 mail far more than offset the slightly higher conversion and maintenance costs, such that multi-line OCRs offer a clear economic (as well as technical) advantage over single-line OCRs. Put more simply, if USPS were starting from scratch today, multi-line OCRs would appear to be the logical choice. 3
PAGE 10
The automatic conversion strategy assumes that conversion of single-line OCRs to multi-line capability is technically feasible and legally viable and could be accomplished with no degradation in performance. Questions have been raised as to whether the single-line OCR vendors would have the ability to do the conversions and/or whether other vendors -perhaps more experienced with multi-line OCRs -would be able to do the conversions without having access to proprietary information. A possible solution would be to reissue the Phase II request for proposals (RFP) with additional criteria on singleto multi-line convertibility and/or with a procurement split between singleand multi-line OCRs. A split procurement would be intended to provide a stronger push to further improve multi-line OCR performance and perhaps provide a greater incentive for competition in the development of both multi-line OCRs and singleto multi-line conversion kits. OTA found that, overall, a 90-10 split procurement shows the second highest projected economic return, only marginally less than automatic conversion but higher than hedge conversion. Under the 90-10 split procurement option, USPS would cancel the current Phase 11 procurement, immediately reissue an RFP for 90 percent of the single-line OCRs (363 instead of 403), and simultaneously initiate release-loan testing of the multi-line OCR. A new RFP for procurement of the other 10 percent of Phase 11 OCRs, but using multiline technology (40 multi-line OCRs), would be issued as soon as possible, probably in about 2 years. The single-line OCRs (252 from Phase I and 363 from Phase II) would be converted to multi-line as soon as a conversion kit has been successfully developed and tested. The 90-10 option would result in a delay of about 2 to 3 months in procurement of the 90 percent Phase 11 single-line OCRs (the time required to reissue the RFP and receive and evaluate new bids). Procurement of the other 10 percent would be delayed about 2 to 3 years (the time required to complete release-loan testing of, issue an RFP on, and receive and evaluate bids for multi-line OCRs). OTA found that the cost of this delay for 10 percent of the Phase II procurement is very small, and would be negligible if the split procurement resulted in significantly higher multi-line OCR performance than would otherwise be the case. In sum, the 90-10 split procurement option is intended to reduce the uncertainty associated with automatic conversion by providing a greater incentive to companies to further improve multi-line OCR performance and to develop the best possible conversion kits. OTA is not recommending one option over another, but simply pointing out the trade-offs involved. The automatic conversion shows the highest projected economic return, followed in order by the 90-10 split procurement, hedge conversion, and 50-50 split procurement. All of these options depend on conversion kits that provide high multi-line performance. The principal question is how to stimulate development of the best possible conversion kit. OTA believes that some outside competition would help achieve this objective, and that it will be necessary to provide incentives to attract the best companies. One incentive is to keep open the decision on which company will do the
PAGE 11
conversions pending the results of several competitive development and testing efforts. This means that the best performing company would have a good chance for the conversion contract (estimated at about $130 million). A second incentive is the opportunity for participating companies to use the R&D results on the world market, even if USPS does not buy any multi-line OCRs. A third incentive would be the prospect of competing for the 10 percent of the Phase II procurement reserved for multi-line OCRs (estimated at $34 million) under a split procurement option. Thus, a 90-10 split procurement option could involve several elements: reissuing the Phase II RFP for 363 (rather than 403) single-line OCRs; initiating competitive release-loan testing on multi-line OCRs; and awarding several development contracts for conversion kits, either all on a competitive basis or at least one on an open competitive basis even if the others are awarded sole source to Phase I and Phase II single-line OCR companies. In addition to revising current automation strategy to give greater emphasis to multi-line OCRs, USPS may also wish to strengthen its commitment to research and development, which is still well below industry averages, and aggressively pursue further opportunities for improved performance of postal automation.
PAGE 12
Summary Summary of Technical Analysis Alternatives to optical character recognition technology. As a starting point, OTA examined possible alternatives to optical character recognition for postal automation technology. Optical character recognition technology reads printed alphanumeric characters (letters and numbers) and recodes these characters into machine-readable forms such as a bar code. OTA identified several electronic, magnetic, and mechanical alternatives to optical character recognition. However, OTA concluded that, at least for the U.S. mail, it is as yet difficult to improve on the information-carrying ability, readability, and cost effectiveness of printed characters on paper. As long as this is the case, then optical character recognition technology is the technology of choice. Electronic mail is likely to be the strongest competitor of postal automation. But there most likely will be a significant residual volume of paper mail at least through the year 2000. Thus there is a window of opportunity for further USPS use of paper-based automation technology. Alternatives to a 9-digit ZIP code. OTA found that there are alternative codes. However, OTA concluded that at this juncture there is no realistic alternative. The 5digit ZIP is almost universally accepted and used (98 percent usage); the 9-digit ZIP directory is now completed; and ZIP+4 codes are being distributed to large business mailers. If ZIP+4 becomes widely used, USPS could consider adding a tenth digit (for error checking purposes) at some future time. (Note: The USPS bar code already includes a correction character. ) Only if ZIP+4 does not become widely used could alternative codes realistically be considered. Performance of single-line optical character readers (OCRs). OTA reviewed available data on performance of the single-line OCRs now being installed by USPS. Single-line OCRs read only the last line of an address -usually containing the city, State, and 5or 9-digit ZIP code. OTA concluded that, despite initial start-up problems, the already installed OCRs now essentially meet USPS performance specifications. Performance of multi-line OCRs. Over the last few years, multi-line OCR technology has emerged from the laboratory and prototype stage to operational units. Multi-line machines read up to four lines of the address. OTA concluded that, as of May 1984, the preponderance of evidence indicates that multi-line OCR performance is essentially equivalent to single-line for reading 9-digit ZIP mail, and that multi-line performance is substantially better for reading 5-digit ZIP 6
PAGE 13
. mail to the 9-digit level. OTA identified one U.S. firm (Recognition Equipment, Inc.) and two foreign firms (Telefunken of Germany, ELSAG of Italy) that have proven multiline OCRs. OTA also concluded that USPS has probably underestimated the ultimate operational performance level of multi-line OCRs by 5 to 15 percent. USPS estimated that multi-line OCRs would process 60 percent of 5-digit mail to the 9-digit level. OTA believes that 65 percent is more likely and 75 percent possible. Feasibility of local and national directories. In order to read, code, and sort S-digit ZIP mail to the 9-digit level, multi-line OCRs require a computerized address directory against which the address information can be compared to ascertain the correct 9-digit ZIP code. They then apply the corresponding barcode, and finally sort the letter. Until recently, the absence of a local or national directory was a limiting factor for use of multi-line OCRs. However, in the 1981-83 period, USPS completed a national ZIP+4 directory and local ZIP+4 directories for major metropolitan areas. USPS and OTA agree that the conversion of existing local ZIP+4 directories to a multi-line OCR format is technically feasible. Whereas local directories clearly would be necessary for multi-line OCR operation, OTA was not able to determine whether national directories would offer any significant advantage, particularly when compared to the technical difficulties and likely additional cost. Feasibility of single-line to multi-line conversion. OTA reviewed the technical feasibility and cost of converting single-line OCRs to multi-line. OTA concluded that conversion would be technically feasible and the USPS estimate of conversion cost -$200,000 per machine -is as good as can be developed from available information. The actual cost of singleto multi-line OCR upgrade can only be determined by detailed engineering analysis and competitive procurement process. It impossible that the conversion could be accomplished by an OCR manufacturer other than the original source, although this might require a high degree of technical cooperation between the two vendors. Technical opportunities for improved performance. OTA identified several areas where technical performance of postal automation might be improved in the future. These include bar-coded reply envelopes, mailer printing of bar codes, improvements in performance of optical character readers, standards for address format, and increased research and development on postal automation. The full address with 5-digit ZIP is read and compared against a computerized address directory that includes 9-digit ZIP codes. If a match is made between the address on the envelope and an address in the directory, the appropriate 9-digit code is applied. 7
PAGE 14
l The USPS record on postal automation R&D is mixed. USPS continues to underspend on R&D compared to the U.S. industry average, and postal R&D organization and management appear to have lacked stability, clear direction and, at times, top level commitment. Despite 20 years of USPS investment in optical character recognition R&D, when USPS solicited manufacturers in 1980 for single-line OCRs, all U.S. manufacturers previously receiving USPS support for single-line OCR R&D had withdrawn from the market, in part due to several years of USPS indecision on an automation strategy. On the other hand, USPS has established a good track record in narrowly focused R&D on improvements to upgrade existing equipment, such as the multi-position letter sorting machine. Also, despite some variability in funding and commitment, USPS has provided enough support over the last 14 years to Recognition Equipment, Inc. (REI) such that REI has developed one of the leading multi-line OCRs on the world market. Summary of Decision Analysis USPS faces a decision point as to whether to continue its commitment to ZIP+4 and single-line OCR technology or to modify that commitment in some way. OTA employed decision analysis techniques to: identify the range of options available to USPS; develop a probabilistic cash flow model of each option; assign probability distributions for key variables such as ZIP+4 usage and multi-line OCR performance; and calculate the rate of return (ROI), net present value (NPV), total net cash flow, and annual net cash flow for each option; and conduct sensitivity tests of the results to changes in key variables. Description of decision options. o Option A: Phase II single-line OCR is the current USPS strategy to proceed to procurement of the 403 additional single-line OCRs advertised for Phase 11 of the postal automation program, and on which bids have already been received. Under option A, there would be no further USPS expenditure on multi-line OCR research, development, and testing. o Option B: Multi-line OCR with ZIP+4 is a decision to cancel the current Phase 11 single-line OCR procurement, initiate release-loan testing of multi-line OCRs, and as soon as possible reissue the Phase [1 request for proposals but for multi-line rather than single-line OCRs, meanwhile retaining the ZIP+4 code. Single-line OCRs already purchased would be converted to multi-line capability. OTA did not analyze the option of procuring 403 additional Phase I single-line OCRs instead of Phase II OCRs. This option was judged to be not significantly different from option A. ** The release-loan testing manufacturers actually test prototype equipment on USPS premises with real mail. 8
PAGE 15
o Option C: Multi-line OCR without ZIP+4 is the same as option B except that the ZIP+4 code would be terminated. The 5-digit ZIP code would be retained. o Option D: Automatic conversion is to proceed with the Phase 11 single-line OCR procurement, but simultaneously initiate release-loan testing (and any necessary related R&D) on single-line to multi-line conversion and then convert all single-line OCRs as soon as possible, regardless of the level of ZIP+4 use. o Option E: Hedge conversion is similar to option D except that the singleto multi-line conversion would take place only if ZIP+4 use is low at a specified future time (defined here as year-end 1987). Both options D and E include the same initial decision to purchase Phase II single-line OCRs, and to initiate release-loan testing of and any necessary research on conversion. The difference is that under option D, the conversion would be made regardless of the level of ZIP+4 use, while under option E, conversion would take place only if use is low. o Option F: Cancel Phase II and ZIP+4 is to cancel the Phase II single-line OCR procurement, terminate ZIP+4, and use the single-line OCRs already purchased to process 5-digit ZIP mail. o Option G: 50-50 Split procurement is a hybrid option that would cancel the Phase II procurement, immediately reissue an RFP for one-half the number of single-line OCRs (202 instead of 403), and simultaneously initiate release-loan testing of the multi-line OCR and singleto multi-line conversion. A new RFP for procurement of the other half of the OCRs but using multi-line technology (201 multi-line OCRs) would be issued as soon as possible, probably in about 2 years, at which time the then existing single-line OCRs (252 from Phase I and 202 from Phase II) would be converted to multi-line. o Option H: 90-10 Split Procurement is similar to option G except that the Phase 11 RFP would be reissued for 90 percent of the single-line OCRs (363), rather than 50 percent, and release-loan testing would be initiated on multiline OCRs leading to a new RFP for procurement of the other 10 percent of the OCRs (40) using multi-line technology. Key assumptions. Where possible and justifiable, OTA used the same assumptions as did USPS. For example, OTA and USPS used the same time horizon (14 years, 198598), labor cost escalation rate (7.42 percent annually), baseline cost and savings projections (for single-line OCRs, as presented in January 1984 to the Board of Governors), discount rate (15 percent per year), and single-line OCR performance and cost. OTA assumptions about multi-line OCR cost ($850,000 per machine), singleto multi-line conversion cost ($200! 000 per machine), and the time required to release-loan test and procure multi-line OCRs and conversion kits (3 years) were generally consistent with USPS and GAO estimates.
PAGE 16
l The major differences between USPS and OTA were assumptions about the incentive rates for ZIP+4 usage (OTA and GAO treated these as a cost, since incentives appear to be required to get large mailers to use ZIP+4), multi-line OCR performance rates (OTA and GAO concluded that performance would likely be somewhat better than USPS estimated), ZIP+4 usage, and clerk/carrier savings rate. For the latter three variables, OTA developed low, median, and high estimates. For low, median, and high estimates of multi-line OCR performance. OTA concluded that production model multi-line OCRs would read 60, 65. and 75 percent of 5-digit mail to the 9-digit level, respectively. For ZIP+4 usage, USPS projects that usage would reach 90 percent of the machinable metered first class mail base within 5 years. This is considerably more optimistic than actual experience with either the U.S. 5-digit ZIP code or the Canadian 6-digit postal code. The 5-digit ZIP took 12 years to reach 90 percent usage; after 5 years, the 5-digit usage level was about 51 percent. Thus, the USPS projection shows ZIP+4 reaching 90 percent about two and one-half times as fast as the 5-digit ZIP (in 5 years rather than 12). At present, the projected actual volume of ZIP+4 mail for 1984 is about 20 percent of the original USPS projection. After reviewing all available evidence, OTA concluded that the USPS ZIP+4 projection should be considered optimistic (high), that an appropriate median estimate would be the 5-digit ZIP growth pattern, and that an appropriate pessimistic (low) estimate would be a growth pattern peaking out at about 40 percent ZIP+4 usage. At the present time, the first year ZIP+4 usage could turn out to be even more pessimistic. The estimated 2.73 billion pieces of ZIP+4 first class mail at year end 1984 represents about 5.4 percent of the target mail base as compared to about 7 percent under the low scenario, 13 percent under the median, and 28 percent under the optimistic scenario. (GAO made no estimate of ZIP+4 usage but did find that businesses are still concerned about the cost of converting and whether the USPS presort discount will be modified because of ZIP+4.) For clerk/carrier labor savings, OTA concluded that the USPS baseline estimate was probably somewhat optimistic, since the quality and mix of the OCR processed mail may be less than anticipated, and labor costs (e.g., for maintenance) may be more than expected. OTA assumed high, median, and low labor savings rates of 100 percent, 90 percent, and 80 percent of the USPS estimate. Results of decision analysis. With respect to internal rates of return (ROIs), every option except option F (cancel), under all conditions modeled, shows an ROI above the 15 percent threshold established by USPS. OTA assumed, therefore, that under any scenario, the Phase I single-line OCRs already purchased would be kept in service. All ROIs, net present values, and cash flows were calculated net of cash flows associated with the Phase I single-line OCRs. Use of ROIs for decision making has a serious limitation. When more than one option clears the hurdle rate (that is, has more than the minimum required ROI, in this case 15 percent), the ROI itself gives no indication of the cash flow differences of the various options as a basis for comparing the options. An alternative to ROI frequently used in capital investment decision making is net present value (NPV). NPV discounts the cash flows of each option at the hurdle or threshold rate, in this study 15 percent. 10
PAGE 17
Under conditions of high savings and high multi-line performance, option D (automatic conversion) has about a 5 percent and 11 percent higher NPV with high and median ZIP+ 4 usage, respectively, than option A (single-line OCR). At low ZIP+4 usage, all other things being equal, the option D advantage increases to a substantial 134 percent or about $820 million in NPV. At a low savings rate (along with low ZIP+4 usage and high multi-line performance), the relative advantage of option D over A increases further to about 310 percent although the absolute advantage decreases to about $650 million in NPV. Even at low multi-line performance, option D has 53 to 119 percent relative advantage in NPV and a $320 to $250 million absolute advantage in NPV, at a high and low savings rate, respectively. Option E (hedge conversion) has the same NPV as option A at high or median ZIP+4 usage and the same NPV as option D at low ZIP+4 usage. Option H (90-10 split procurement) also has a higher NPV than option A under almost all conditions. Option G (50-50 split procurement) has a significant although somewhat smaller advantage over option A at low ZIP+4 usage. Option G has a 34 to 271 percent relative advantage in NPV and a $170 to $710 million absolute advantage in NPV at low ZIP+4 usage, depending on the multi-line OCR performance rate and savings rate. The ranking of the options by NPV is summarized below: Overall NPV Rank Low ZIP+4 Use NPV Rank Option D 1 highest Option D 1 highest H 2 E 2 (tie) E 3 H 3 G 4 c 4 A 5 G 5 B 6 B 6 c 7 lowest A 7 lowest OTA found that the NPV results are not very sensitive to the purchase price of the multi-line OCR or the number of multi-line OCR units. An increase in the purchase price from $850,000 to $970,000 or an increase in the number of units from 403 to 444 (as estimated by GAO to be required if the entire Phase [1 procurement was switched from singleto multi-line OCRs) would reduce NPV by about $20 to $30 million. Net present value appears to be the best basis for comparative quantitative evaluation of the decision options. However, the actual undiscounted net cash flows over the 13 year payback period (1985-1998) can provide another dimension to the evaluation. Option A (single-line) is estimated to show positive cash flows of $8.8, $8.24, and $3.57 billion at high, medium, and low ZIP+4 usage. At high ZIP+4 usage, option B (multi-line with ZIP+4) is somewhat lower at $8.14 billion, options D (automatic conversion) and H (90-10 split procurement) somewhat higher at $9.36 billion and $9.24 billion respectively, and option G (50-50 split procurement) about the same at $8.7 5 billion. The comparisons between options change relatively little at median ZIP+4 usage. However, at low ZIP+4 usage there is a substantial difference in net cash flows. Option A(single-line) shows a net cash flow of $3.57 billion. But, depending on the multi-line OCR performance rate, options D (automatic conversion) and H (90-10 split 11
PAGE 18
. ., procurement) show a net cash flow of $5 to $7.2 billion, or about $1.4 to $3.6 billion greater than option A. Option G (50-50 split procurement) shows about $1.1 to $3.3 billion greater cash flow than option A, and option B (multi-line with ZIP+4) shows about $0.8 to $3.0 billion greater cash flow than option A. A comparison of yearly cash flows gives similar results. By 1994, all equipment will presumably have been installed (or converted) and up and running at optimal performance. Options B, D, G. and H will by that time look exactly the same -ail multi-line OCRs. The single-line OCRs procured under options D, G, and H will have been converted to multi-line capability. Option A will continue to be solely single-line OCRs. With high ZIP+4 usage, option A shows an annual net cash flow of about $870 million to $1.2 billion from 1994 to 1998. Options B, D, G, and H show almost identical annual cash flows, only slightly higher by about $70 to $100 million per year. However, at low ZIP+4 usage, the differences again become substantial. With high multi-line performance, options B, D, G, and H show between $440 and $580 million per yea r additional net cash flow compared to option A, from 1984 to 1998. With median multiline performance, the advantage of options B, D, G, and H over A ranges from $370 to $490 million per year. And even at low multi-line performance, the advantage over option A, while reduced, is still significant at $180 to $240 million per year. 12
PAGE 19
., Technical Analysis Alternatives to Optical Character Recognition As a starting point, OTA examined possible alternatives to optical character recognition technology for postal automation. OTA concluded that the strongest competition to postal automation is likely to come from electronic mail. If a significant portion of the paper-based mainstream were to divert to either Generation II (electronic input-hardcopy output) or Generation [11 (electronic input and output) recognition technology would electronic mail and message assumptions about growth of electronic mail, then the need for optical character be reduced. However, in a previous (1982) study of systems, OTA found that, even under very optimistic electronic mail, there most likely will be a significant residual volume of paper mail at least through the year 2000. Therefore, while electronic mail is a strong competitor of postal automation, the major effects of electronic competition are likely to be delayed for at least 10 to 15 years. From this vantage point, there is a window of paper-based automation technology. Some leaders in the optical character recognition the current information revolution promises to replace opportunity for further use of industry already recognize that the traditional media on which information has been carried (paper) with electronic media.... Thus, OCR provides a bridge between the Paper Age and. the [formation Age. It is a transitional system which aids users who have one foot in each era. But as the (electronic) Information Age matures, the role of OCR promises to diminish.... The irony is that OCR will be faced with increased opportunities before the electronic axe falls (Schantz, 1983, p.?). 13
PAGE 20
Optical character recognition technology reads printed alphanumeric characters (letters and numbers) and recodes these characters into machine-readable forms such as a barcode. There are other recognition technologies like magnetic or mechanical, but these seem clearly impractical for conventional mail. For example, rather than showing addresses in black and white alphanumeric printing on envelopes, the address information could be stored in bits of information in magnetic stripes on the envelope, similar to credit cards and farecards. However, magnetic stripes are not readable by the human eye and would be difficult for senders to encode and receivers to decode. Another form of magnetic code is MICR (magnetic ink character recognition) used on bank checks. characters and requires special equipment. used. Thus, address information could be attached to the envelope. punches. In sum, while there These would be are electronic, This code is readable but lacks alphabetic Alternatively, a mechanical code could be stored as punched holes in cards or tape difficult to read and would require special magnetic, and mechanical alternatives to optical character recognition, none are both readable and readily and cheaply available as a substitute at the present time. In the final analysis, OTA found that, at least for the U.S. mail, it is as yet difficult to improve on the information carrying ability, readability, and cost effectiveness of printed characters on paper. As long as this is the case, then optical character recognition technology is the technology of choice.
PAGE 21
. Alternatives to a 9-Digit ZIP Code OTA also examined whether there are viable alternatives to the 9-digit ZIP or ZIP+4 code. OTA concluded that there are alternative codes. and, indeed, some are used today by other countries. For example, both Canada and Britain use alphanumeric zip codes, that is, a combination of letters and numbers. Other code schemes have been suggested, for example, using individual telephone numbers as zip codes. Telephone numbers would permit sorting down to the level of each individual street address. In 1976, USPS considered a wide range of alternative ZIP schemes, including scrapping the 5-digit ZIP, using an alphanumeric code, and adding a check digit (e.g., a tenth digit to the 9-digit code). USPS ruled out any change in the basic 5-digit ZIP, since almost all mail (about 94 percent as of 1976) used a ZIP code. A change in the 5-digit code was judged by USPS to be unfair and excessively burdensome to mailers. This left the alternative of adding 3, 4 or 5 digits to the existing 5-digit codes. USPS elected to add 4 digits. Three was ruled out since this would have required an alpha or alphanumeric add-on. Five was likewise ruled out, since the additional digit, while helping to detect code errors and preventing letters from sorting to the wrong destination, would have increased mailing list information and maintenance cost. (The USPS barcode does contain a correction character.) At the March 5 OTA workshop, several participants expressed the view that the current 9-digit ZIP was not the best code, but that it was too late to make any major changes. The 5-digit ZIP is almost universally accepted and used (98 percent usage) and the 9-digit ZIP directory is now completed. ZIP+4 codes are being distributed to and beginning to be used by large business mailers. 15
PAGE 22
-. At this juncture, OTA concluded that there is no realistic alternative. If the ZIP+4 becomes widely used, USPS could consider adding a tenth digit (for error checking purposes) at some future time. If ZIP+4 does not become widely used, alternative codes could be considered. Performance of Single-line Optical Character Readers (OCRs) OTA reviewed available data on performance of the single-line OCRs now being installed by USPS. Single-line OCRs read only the last line of an address -usually containing the city, State, and 5-or 9-digit ZIP code. [n the 1976-80 period, when the basic USPS automation program was developed, the single-line optical character reader was, in the opinion of USPS, the only proven equipment. Even so, in 1980 when USPS issued specifications for single-line OCRs, no U.S. manufacturer had OCRs meeting USPS specifications. As a result, and to meet USPS domestic content requirements, four foreign companies teamed with U.S. manufacturers who were licensed to produce single-line OCRs. Two U.S. OCR manufacturers (under foreign license) were selected -Burroughs (under license to NEC) and Pitney Bowes (under license to code sorter (BCS) manufacturer, Bell and Howell. Both OCR manufacturers experienced start-up performance specifications. However, based on review of ELSAG) -along with one bar problems in meeting USPS current performance data and on-site inspection, OTA concluded USPS performance specifications. figure 1. that the already installed OCRs now [illustrative OCR performance data essentially meet are presented in 16
PAGE 23
Figure 1 Illustrative Optical Character Reader (OCR) Performance Data Burroughs/NEC Actual Test Results Meter Belt Mail Managed Mail Performance Specs Meter Belt Mail Managed Mail Actual Test Results Meter Belt Mail Managed Mail Performance Specs Meter Belt Mail Managed Mail Accept Throughput Error Rate Rate Rate 80.0% 28,500 pieces/hour 1.50% 60.0% 31,300 pieces/hour 2.40% 62.1% 28,908 pieces/hour 53.4% 30,083 pieces/hour Pitney Bowes/ELSAG Accept Throughput Rate Rate 72.0% 28,500 pieces/hour 62.0% 30,500 pieces/hour 67.1% 26,224 pieces/hour 57.8% 26,730 pieces/hour 2.37% 2.00% Error Rate 1.10% 1.10% 1.50% 1.50% Definitions: Accept Rate Percent of letters read by the OCR as a portion of the total fed into the OCR. Throughput Total mail pieces fed through the OCR per hour. Error Rate Percent of letters sent to the wrong pocket. Meter Belt Mail High quality, generally OCR readable mailfrom large mailers. Managed Mail Mail from other Post Offices containing widely variable levels of OCR readable mail. Source: United States Postal Service: August 1983 2-week tests. 16a
PAGE 24
- -.. The 252 OCRs already purchased by USPS averaged $645,000 per unit in capital cost, although the Burroughs price was apparently significantly lower (approximately $300,000 per unit) than the Pitney Bowes price. These recognition acceptance USPS), and significant prices include all OCR equipment (e.g., mail transport, optical character unit, computer directory, ink jet printer, sorting stackers) plus installation, (equipment must perform to USPS specifications prior to acceptance by 2 years worth of spare parts. There is no obvious explanation for th e price difference between Burroughs and Pitney Bowes, since both manufacturers bid on the same number of units meeting the same technical performance specifications. USPS apparently split the procurement between two vendors in order to encourage future competition for subsequent procurements, but at an additional cost of about $37.8 million (126 units Pitney Bowes and Burroughs). In the so-called Phase II OCRs. Competitive bids have times the estimated $300,000 price differential between procurement, USPS intends to purchase an additional 403 been solicited from four qualified U.S. manufacturers (all under license to foreign companies): U.S. Manufacturers Foreign Licenser Burroughs NEC (Japan) Pitney Bow-es ELSAG (Italy) Recognition Equipment Inc. Toshiba (Japan) ElectroCom Telefunken (Germany) USPS has budgeted for a capital cost of $660,000 per unit for been received by USPS, but a selection decision and contract before late June 1984. the 403 OCRs. Bids have award(s) will not be made
PAGE 25
Performance of Bar Code Sorters (BCSs) Performance of the BCSs has not been in dispute. USPS has already procured 248 units from Bell and Howell at a capital cost of approximately $129,000 each. The BCSs sort 24,000 to 28,000 letters per hour depending on the type of sort, with an accept rate of 96 percent. USPS plans to procure an additional 452 BCSs as part of the Phase II procurement, and has budgeted approximately agreements with five BCS manufacturers to summer of 1984. The manufacturers are: $154,000 per unit. USPS has signed conduct release-loan tests during the Bell and Howell (U. S.) Hotchkiss-Brandt Sogeme (France) USPS will require that at manufacture. Therefore, any Leigh Instruments (Canada) National Presort (U. S.) Telefunken (Germany) [east 75 percent of the machine cost be of domestic foreign manufacturer whose equipment tests satisfactorily will have to license a U.S. company to produce all or most of the machines in order to qualify for the competitive procurement. Performance of Multi-Line OCRs Over the last few years, multi-line OCR technology has emerged from the laboratory and prototype stage to operational units. OTA examined available research and data on multi-line OCR performance. The major difference between singleand multi-line OCRs is that the multi-line machines read up to four lines of the address while the single-line machines read only the bottom line (with city, state, and ZIP). 18
PAGE 26
At present, USPS has no definite plans for use has largely funded the development of a multi-line (REI), a U.S. company based in Dallas, Texas. REI of multi-line OCRs. However, USPS OCR by Recognition Equipment Inc. has a proven track record in optical character recognition technology and is a leading U.S. corporation in high performance OCRs. There are a total of five prototype REI multi-line OCRs operating installations (two in Chicago; one each in New York, Philadelphia, and Dallas) OCRs are known as RCS/OCR for Read Code Sort/Optical Character Readers. at posta l The REI The computer software of the RCS/OCR is programed so that the address search is bottom up. That is, the bottom line containing city, State, and ZIP code (5or 9-digit) is read first, followed by followed by the third (and, building, etc. The address the second line containing the street number and name, if necessary, fourth) line containing company name, office information on the envelope is compared with information maintained in a computerized ZIP+4 address directory. Once a match between the address information on the envelope and in the directory is obtained, a bar code is applied to the envelope, which from then on is sorted automatically down to the carrier level. The multi-line provides additional redundancy since, for example, the street number and name as well as city and State can be cross-checked against the ZIP code. A direct comparison between singleand multi-line machine performance is difficult, since the USPS has not subjected both machines to equivalent acceptance testing on a comparable mail base. REI had proposed comparative testing, but this suggestion was declined by USPS on the grounds that it would be unfair to other potential multi-line OCR manufacturers and would violate the ongoing competitive procurement process for single-line OCRs. USPS also asserts that the REI multi-line OCR did not 19
PAGE 27
. meet USPS performance specifications in 1980, when USPS initially decided to use single-line OCRs, and that test results on the prototype multi-line OCRs were not available until April 1983, after USPS had decided to purchase single-line OCRs. OTA did not itself investigate and has reached no conclusions on the OCR procurement history. However, OTA did conclude that, as of May 1984, the preponderance of evidence indicates that multi-line OCR performance is essentially equivalent to that of single-line OCR performance for processing 9-digit ZIP mail, and that multi-line performance is substantially better for processing 5-digit ZIP mail to the 9-digit level. For 9-digit ZIP (ZIP+4) mail, USPS performance data indicate that the Burroughs and Pitney Bowes single-line OCRs and the REI multi-line OCRs all correctly read, code, and sort 98 to 99 percent of OCR-readable ZIP+4 analysis, USPS assumes 100 percent correct reading of mail. For purposes of mail flow ZIP+4 mail. For 5-digit ZIP mail, both single-line and multi-line OCRs correctly read, code, and sort 98 to 99 percent correctly to 5 digits. However, only the multi-line OCR can read, code, and sort 5-digit ZIP mail to 9 digits. USPS has estimated that the multi-line OCR can read, code, and sort 60 percent of 5-digit ZIP mail to 9 digits. USPS believes that this 60 percent estimate may be high, since USPS assumed that the total local metropolitan area would be included in the OCR computer directory. If the local directory has less than total coverage, the read-code-sort rate would be reduced. Also, USPS notes that adequate test data are not available on how the multi-line OCR performs two-stage encoding (e.g., placing a 5-digit bar code on a non-local letter at an originating post office and subsequently placing the 20
PAGE 28
4-digit add-on bar code -or the entire 9-digit bar code -on the letter at the destination post office). In contrast, OTA has concluded that the 60 percent USPS estimate may be low for the following reasons. First, full coverage computerized local address directories appear to be technically and economically feasible. USPS already has partial local directories in several metropolitan areas. Second, there is no evidence that two-stage encoding will pose a significant problem for multi-line OCRs. Reasserts that two-stage encoding can be accomplished with no significant degradation in performance. Third, it is reasonable to expect that production model multi-line OCRs would have improved performance compared to the prototype RSC/OCRs. USPS has found that single-line OCR performance improved 5 to 10 percent between prototype and production. With respect to overall multi-line OCRs are roughly shown in figure 2. productivity, USPS has concluded that the single-line and equal. Average data from USPS performance reports are In addition to REI, OTA has identified two other companies that manufacture multi-line OCRs: Telefunken (Germany) and ELSAG (Italy). Japanese firms may have the capability and interest, judging from their activity in the single-line OCR market. Other than REI, no U.S. companies are known to currently have multi-line OCR capability. At one time Control Data Corporation, IBM Corporation, and Ford-Aerospace (Philco-Ford) all had single-line OCR products, and might have been able to develop multi-line OCRs, but left the business in the mid-1970s. Burroughs, Pitney Bowes, and ElectroCom have acquired single-line OCR capability under licenses to foreign manufacturers (NEC, ELSAG, and Telefunken, respectively).
PAGE 29
Figure 2 Productivity of Singleand Multi-Line Optical Character Readers Burroughs Single-line OCR 52.8 22,324 8,527 Pitney Bowes Single-line OCR 51.8 19,305 9,127 REI Multi-line OCR 51.3 22.095 10.397 Definitions: Gross Accept Rate Pieces of mail accepted by the machine per hour divided by pieces of mail fed to the machine per hour. Throughput Pieces of mail fed through the machine per hour. Productivity Throughput divided by workhours to arrive at pieces of mail processed per work hour. Source: United States Postal Service 21a
PAGE 30
. While it is difficult to extrapolate from foreign experience to U.S. postal needs, due in part to major differences in the mail make-up and use of postal codes, multi-line OCRs appear to be performing well in other countries. ELSAG has 20 two-line OCRs operating in French post offices with an apparently very low error rate (0.1 to 0.5 percent). Telefunken has twoor three-line OCRs operating in Norway, the Netherlands, and Britain. The British Post Office reports that its one Telefunken three-line OCR is undergoing a field trial to be completed by December 1984, and is handling United Kingdom mail at rates between 28,000 and 30,000 letters per hour. Productivity and error rates are not known. Feasibility of Local and National Directories [n order to read, code, and sort 5-digit ZIP mail to the 9-digit level, multi-line OCRs require a computerized address directory against which the address information can be compared to ascertain the correct 9-digit ZIP code, then apply the corresponding bar code, and finally sort the letter. OTA has reviewed the current state-of-the-art in computerized directories to determine if such directories for postal purposes are technically and economically feasible. Accordingly to USPS, about 40 percent of a multi-line OCR is to process local mail to mail is local and 60 percent non-local. [f the 9-digit level at the originating post office, a local directory is needed. non-local) to the 9-digit level at needed. If a multi-line OCR is to process all mail (local and the originating post office, a national directory is 22
PAGE 31
Until recently, the absence of a local or national directory was a limiting factor for use of multi-line OCRs. However, in the 1981-83 period, USPS completed a national ZIP+4 directory and local ZIP+4 directories for major metropolitan areas. The national directory is stored on a computer in San Francisco and is essentially the sum of all local directories. In the four cities where multi-line OCRs are already operating (Chicago, New York, Philadelphia, Dallas), the local directories have been partially converted to a format usable by the multi-line OCRs. For example, the Philadelphia multi-line OCR uses a converted local ZIP+4 directory containing about 185,000 local ZIP+4 codes covering more than one-half of the addresses in the Philadelphia metropolitan area. USPS and OTA agree that the conversion of existing local ZIP+4 directories to a multi-line OCR format is technically feasible. This would be simply an extension of the partial conversions already accomplished in the four cities noted above. Whereas local directories clearly would be necessary for multi-line OCR operation, OTA was not able to determine whether national directories would offer any significant advantage. With local directories only, non-local mail would have to be processed by multi-line OCRs twice, once at the second time at the destination post would, in theory, eliminate the need OCR processing, the ZIP+4 bar coded originating post office to the 5-digit level and a office to the 9-digit level. A national directory for two-stage OCR processing. After the initial mail could bypass subsequent OCR processing and be handled entirely by the less expensive bar code sorters. USPS argues that this would not result in a reduction in the number of OCRs, since these machines would still be needed for processing outgoing mail at the destination post 23
PAGE 32
off ice. In addition, USPS points out that when taken together, local directories in all major metropolitan areas would amount to a de facto national directory. Local mail (about 40 percent of the total) would be processed using a local directory in the originating post office, and non-local mail (the other 60 percent) would be processed using local directories in the various destination post offices. Nonetheless, OTA estimated the technical and cost requirements of a national directory, even though the need for such a directory has not been firmly established. The memory size of a national directory has been grossly estimated at 20 gigabits or 20 billion bits. A directory of this size would contain all 20 million ZIP+4 codes plus address information including street number and address, city, State, and, where necessary, building floor and suite numbers. The size could be reduced to include only the most frequently used ZIP+4 codes and related address information. For example, if 15 percent of ZIP+4 codes account for 75 percent of ZIP+4 code use, then a memory size of 3 billion bits might suffice. Currently available magnetic disc memory technology can provide a 3 billion bit capacity at a cost of about $30,000. But the average access time appears to be too long. Optical disc and magnetic bubble memories have similar limitations. Large random access memories (RAMs) have fast access times (microseconds as compared to milliseconds) and may be the best approach. A 3 billion bit directory using 256K RAMs might cost on the order of $300,000 in 3 years (1987 dollars). If five OCRs shared each directory, then the cost per OCR would be about $60,000. [n sum, a national directory would be technically feasible with memory technology now coming on the market. A full national directory of 20 billion bits of information
PAGE 33
. would be very expensive (about $4 million each in 1987), even if shared among five OCRs ($800,000 per OCR). If the memory size is reduced to 3 billion bits, the cost would be about $300,000 per memory, or $60,000 per OCR if used on a shared basis (as above). In contrast, a local directory would require much smaller memory size (e.g., about 72 million bits for the Philadelphia metropolitan area). OTA estimated that the cost of a typical local directory would be about $20,000 (roughly $300,000 times 72 million/3 billion times a multiplier of 3x). At this low cost, sharing a directory among several OCRs may not be necessary. But if shared among five OCRs, the cost per OCR would be further reduced to about $4,000. In all likelihood, the capital cost of local directories would be small compared to either the total directory data conversion and maintenance. cost of multi-line OCRs or to the cost of Feasibility of Single-line to Multi-line Conversion OTA reviewed the technical feasibility and cost of converting single-line OCRs to multi-line. OTA concluded that conversion would be technically feasible and that the USPS estimate of conversion cost -$200,000 per machine -is as good as can be developed from available information. Conversion is relatively simple because a large part of the single-line OCR could be retained almost as is, as discussed below. Letter sorter -no change. Letter transport -no change. The current mechanical transport is designed to move faced and bottom justified letters at a constant speed past an OCR window. This function is common to singleand multi-line OCRs and would not change. 25
PAGE 34
. . Prelook window -probably no change. The prelook is used to find the address location, and probably could be used as Lenses and scanners -probably are adequate. Otherwise, the OCR prelook") would need to be replaced. is. no change if the scan height and pixel resolution hardware from the lens system back (including OCR electronics -probably some change needed to upgrade the image registers, recognition logic, etc. Since the multi-line OCR must process much more address information than the single-line OCR, the basic scanning and clocking rates may have to be increased. If the single-line OCR already at sufficient speed and resolution, then little needed -only a change in computer software captures three or four lines of the address or no change in OCR electronics would be and directory. Computer software -change needed to upgrade the software so that multiple address lines could be formatted and queued processed in a bottom-up fashion, and the resulting data properly into the directory. Computerized directory -change needed to expand the directory from city, State, and ZIP+4 to include street number and name and, as appropriate, building floor and suite. Ink jet printer and verifier -probably no change needed with a local directory. [f a national directory is used, the access and processing time may lengthen to the point where the ink jet printer and verifier (sprays the bar code on the bottom right-hand edge of each envelope) would need to be moved farther downstream in the transport path. 26
PAGE 35
The actual cost of singleto multi-line OCR upgrade can only be determined by detailed engineering analysis and a competitive procurement process. It is possible that the conversion could be accomplished by an OCR manufacturer other than the original source, although this would require a high degree of technical cooperation between the two OCR vendors. Technical Opportunities for Improved Performance OTA identified several areas where technical performance of postal automation might be improved in the future. These include bar-coded reply envelopes, mailer printing of bar codes, character recognition upgrades, address format standards, and increased postal research and development. Bar-coded reply envelopes. Mailers-and especially large business mailers -already frequently provide reply envelopes to customers, presumably to facilitate payment of bills such as those mailed out monthly by utility, telephone, and gas companies. Preprinting of the bar code along with the return address on the reply envelope would appear to be cost-effective, and could permit processing of business reply mail by the less expensive BCSs rather than OCRs. Some business mailers are already preprinting bar codes. But in order for this to be successful, the bar codes must be readable by the Bell and Howell barcode sorters. That is, the color spectrum of the ink and location of the bar code on the envelope must match the capability and location of the photo-detector in the BCS. Also, in order to achieve savings by bypassing some or all OCR processing, the bar-coded reply envelopes would need a unique facing indicia mark (FIM) so that these envelopes could be detected (perhaps by the facer/canceler) early in the mainstream and diverted to BCSs. 27
PAGE 36
- l Mailer printing of outgoing bar codes. Most major business mailers already use computers to store, update, and print out addresses on outgoing mail. Addresses are printed directly on the envelopes, on mailing labels, or on letters that show through window envelopes. Almost all business mailers have already entered the 5-digit ZIP codes into their address data bases. USPS would, of course, like them addresses from 5-digit to 9-digit ZIP codes. At present, few mailers have actually converted their address files. of concern over the cost of conversion. Some participants at the March 5 to convert the partly because OTA workshop pointed out that, if and when mailers convert, consideration should be given to building in a capability to print outgoing bar codes in addition to or as a substitute for the ZIP+4 numeric codes. In principal, outgoing bar-coded mail could, with proper FIMs on the envelopes, bypass In practice, speed non-impact the OCRs completely and go directly to the technical and economic feasibility printers (such as laser printers) could BCSs. is unclear. For example, high be programed to print the bar code immediately under the last address line on an envelope or label. However, if the barcode location is too far up from the bottom of the envelope, then the Bell and Howell bar code sorters used by USPS would not be able to read the bar code unless a second scanning channel was added. could also be used, but this processing changes on the part Ink jet printers or special photo offset printing devices would involve significant equipment procurement and of mailers. And proper location of the bar code could be difficult when labels or window envelopes are or barcode labels might be needed. Bell and and/or inserting machines could be easily used used. Special bar code window envelopes Howell indicated to OTA that addressing to print bar codes on outside envelopes of outgoing mail, and that 30 to 35 billion pieces of mail annually could be processed way. A clear understanding of the potential and pitfalls must await further study. 28 in this
PAGE 37
Character recognition upgrades. The percentage of OCR-readable mail actually read by the single-line OCRs used by USPS, although meeting USPS performance specifications, is still far less than 100 percent. USPS has assumed an average read rate of 70 percent. One way to improve performance is to upgrade the character recognition technology. Current character recognition technology uses mask matching, whereby character patterns are stored in electronic memory and matched against the actual characters in the address. But because there are so many different sizes, shapes, and forms of alphanumeric characters in U.S. mail addresses, not all characters are stored in electronic memory. If the computer cannot make a match within 100 milliseconds, the address is not read. The overall read rate could be improved by, first, studying the rejected addresses and determining what types of characters are not being read. Then, OCR manufacturers could some be solicited to develop improved character recognition technology that would read or all of the characters rejected most frequently. Address format standards. Some OCR rejects are due to problems with the location and format of addresses on envelopes. To receive a ZIP+4 discount, mailers must meet several mandatory requirements. These include: a barcode clear zone (at the bottom right-hand portion of the envelope) in which no printing or markings whatsoever are permitted; an OCR read area in which the city, State, and ZIP+4 code line must be visible and unobstructed (no extraneous printing or markings); machineprinted address with uniform character single and line spacing; and a reasonable degree of color contrast between the address and mail piece. 2 9
PAGE 38
. . However, there are several voluntary guidelines with respect to margins; State abbreviations; character fonts; character pitch, height, and height to width ratio; character and line skew; upper case characters; and line and character spacing. Also, while black ink on white paper is preferred, other color combinations are permitted (except for brilliant colors and reverse printing and any others that do not meet minimum reflectance standards). The overall read rate could determining the cause(s). The most be improved by studying the rejected address and common causal factors could be mitigated by: 1. better enforcement of mandatory format requirements; 2. improved compliance with voluntary format requirements, possibly through incentives, and if necessary, by making some voluntary requirements mandatory; and/or 3* technical upgrades as discussed previously, so that variety of address characters and formats. Postal research and development. The USPS record on mixed. On the negative side, USPS continues to underspend the OCRs can read a wider postal automation R&D is on R&D, despite repeated recommendations from congressional oversight committees and the Commission on Postal Service to raise postal R&D closer to private sector levels. Postal R&D was about $24 million in 1983, or about compares to a U.S. industry organization and management times, top-level commitment. one-tenth of one percent (0.1 percent) of revenue. This average of perhaps 3 percent. Second, postal R&D appear to have lacked stability, clear direction, and, at Third, despite 20 years of USPS investment in optical character recognition R&D, when USPS solicited manufacturers in 1980 for single-line OCRs, no U.S. manufacturer was judged to be qualified. All U.S. manufacturers previously receiving USPS support for single-line OCR R&D had withdrawn from the market by 1980, in part due to USPS indecision on an automation strategy. 30
PAGE 39
For example, Philco Corporation USPS awarded development contracts for first in 1960-65, and for second generation OCRs to generation OCRs to IBM and Philco-Ford in 1968 and 1969, respectively. Both companies developed successful OCR designs and were subsequently awarded prototype contracts. The Philco-Ford second generation OCRs were installed in Boston in November 1971 and successfully tested in early 1972; and IBM OCRs were installed in New York in June 1972 and tested in late 1972. By the time USPS had settled on a postal automation strategy and solicited manufacturers to provide OCRs on a line OCR business Philco-Ford second the IBM OCR units release-loan basis in 1979-80, the only companies left in the singlewere foreign manufacturers. This was despite the fact that the generation OCR units in Boston remained operational until 1982, and in New York are still in service. Thus, it is at least arguable that USPS could have reasonably opted for wide deployment of single-line OCRs in the early 1970s, perhaps using a 5-digit bar code (5digit ZIP code use had reached 84 percent by 1972). Had USPS opted for this strategy, some U.S. manufacturers of single-line OCRs might well have stayed in the business. And it would be reasonable to expect that OCR technology would be further advanced than it is today. On the positive side, USPS has established a good track record in narrowly focused R&D on improvements to upgrade existing equipment. For example, the multi-position letter-sorting machine (MPLSM), in wide use since the late 1960s, has been upgraded several times, most recently by a not yet fully implemented electronic ZIP retrofit (known as EZR) that allows four-digit keying of ZIP+4 codes. Facer cancellers, single position letter sorting machines, and flat sorting machines also have been, or will be, upgraded. USPS equipment upgrades are highlighted in figure 3. 31
PAGE 40
Figure 3 Summary of USPS Equipment Multiposition Letter Sorting Machine ( MPLSM ) 1969 1973 1974 1976 1978 1981 1982 ZIP Mail Translator (ZMT) Converted keying (cordal) to sequential keying. Upgrades MPLSMs from simultaneous ZIP Codes keyed by operators then could be translated by the ZMT into BIN assignments. Engineering Data Isolation Technique (EDIT) An electronic modification to the ZMT which allowed keyed data to be monitored for accuracy. Automatic Density Analysis Profile Technique (ADPT) -An Upgrade that enabled automatic tabulation of MPLSM sweepside BIN densities. Electronic Sort Processor (ESP) A modification to MPLSMs which replaced mechanical code setting with electronic code setting to provide more accuracy, reduce maintenance costs, and reduce noise. Micro-Key An upgrade to the ZMT which allowed the first digit of a carrier route number to be locked in each time an operator keyed an incoming secondary distribution. ZIP Data Logger (ZDL) Electronic ZIP Retrofit (EZR)-A modification to MPLSMs to allow four digit keying of ZIP+4 Codes. Facer Cancelers 1982 Micro Mark A modification to Mark II Facer Cancelers to upgrade the electronics to Solid State circuitry. Single Position Letter Sorting Machine (SPLSM) 1972 Automated Business Mail Processing System (ABMPS) A modification to the Universal Business Machine (UBM) SPLSM to allow automated distribution of destinating Bar Coded Business Mail. Flat Sorting Machine (FSM 775) 1984 Software Modification An upgrade to provide Micro key capability on the FSM 775. Source: United States Postal Service 31a
PAGE 41
Also on the positive side, despite some variability in funding and commitment, USPS has provided enough support over the last 14 years to Recognition Equipment,Inc. (of Dallas, Texas) such that REI has developed one of the leading multi-line OCRs on the world market. Ironically, at various times from the late 1960s to the late 1970s, USPS appeared to actually favor the multi-line over the single-line OCR. [n the late 1970s, USPS procured one multi-line OCR from REI, and as insurance solicited every known OCR manufacturer in the world to provide a single-line OCR on a release-loan basis. As it turned out, the REI multi-line OCR did not satisfy USPS performance requirements, but the single-line OCRs of five foreign manufacturers did (ELSAG, NEC, Telefunken, Toshiba, and ITT Belgium). USPS decided to deploy single-line OCRs and awarded production contracts to Pitney Bowes (under license to ELSAG) and Burroughs (under license to NEC) in early 1981. According to USPS, for insurance purposes an additional contract was awarded to REI for five multi-line OCRs. These were installed and tested between June 1982 and April 1983. As discussed earlier, USPS test data indicate that the multi-line OCR performance is now fully competitive with single-line OCR performance. Finally, USPS may wish to consider: (1) new approaches to R&D and procurement (including the release-loan testing process) with a view towards speeding up the time delay from R&D to installation of new equipment; and (2) new ways to organize mail processing in order to achieve faster and more reliable delivery. 32
PAGE 42
Decision Analysis Introduction USPS is approaching a critical decision point on postal automation strategy. At present, the USPS commitment is to the 9-digit ZIP and single-line OCR. USPS has developed 9-digit ZIP codes (ZIP+4) for the entire United States and prepared a national ZIP+4 directory, although as yet very few mailers are using ZIP+4. Also, USPS has procured and is installing Phase I automation equipment (including 252 single-line OCRs and 248 BSCs) and has received bids on an additional 403 single-line OCRs as part of a Phase II procurement. A selection decision is pending. However, multi-line OCR technology has advanced to the point where it is fully competitive with single-line OCRs ZIP+4 mail. In advantage over addition, multi-line single-line OCRs in Thus USPS faces a decision with respect to technical performance in processing technology offers a significant technical performance processing 5-digit ZIP mail to the 9-digit level. point as to whether to continue its commitment to ZIP+4 and single-line OCR technology or to modify that commitment in some way. [n order to analyze the USPS decision, OTA has employed decision analysis techniques to: identify the range of options available to USPS; develop a probabilistic cash flow model of each option; assign probability distributions for key variables such as ZIP+4 usage and multi-line OCR performance; calculate the rate of return (ROI), net present value (NPV), total net cash flow, and annual net cash flow for each option; and conduct sensitivity tests of the results to changes in key variables. 33
PAGE 43
The options and assumptions used in the decision analysis along with the major results are discussed below. Further details on the modeling methodology and the complete cash flows for each option are included in the appendices. Decryption of Decision Options OTA identified eight major decision options. These are listed in figure 4 and described below. Option A: Phase II Single-Line OCR. Option A is the current USPS strategy to proceed to procurement of the 403 the postal automation program, on A, there would be no furthe r development, and testing. additional single-line OCRs advertised for Phase 11 of which bids have already been received. Under option USPS expenditure on multi-line OCR research, Option B: Multi-Line OCR with ZIP+4. Opt on B is a decision to cancel the current Phase 11 single-line OCR procurement, initiate release-loan testing (where manufacturers actually test prototype equipment on USPS premises using real mail) of multi-line OCRs, and as soon as possible reissue the Phase 11 request for proposals but for multi-line rather than single-line OCRs, meanwhile retaining the ZIP+4 code. Single-line OCRs already purchased would be converted to multi-line capability. that line Option C: Multi-Line OCR without ZIP+4. Option C is the same as option B except the ZIP+4 code would be terminated. The 5 digit ZIP code would be retained. Option D: Automatic Conversion. Option D is to proceed with the Phase II singleOCR procurement, but simultaneously initiate release-loan testing (and any necessary related R&D) on single-line to multi-line conversion, and then convert all single-line OCRs as soon as possible, regardless of the level of ZIP+4 use. 34
PAGE 44
. Figure 4 Decision Options A: Single-Line OCR, Phase II B: Multi-Line OCR, with ZIP+4 c: Multi-Line OCR, No ZIP+4 D: Automatic Conversion: Purchase Single-Line, Cont Research and Testing on Mu Line, and Convert Regardle of ZIP+4 Usage E: Hedge Conversion: Purchase Single-Line, Continue Research and Testing on Multi-Line, Convert if ZIP+4 Usage is Low. inue ltiss F: Cancel Phase II, No ZIP+4 \ 50-50 Split Procurement: Purchase Single-Line now, Multi-Line Later. 90-10 Split Procurement: Purchase 90% Single-Line now, 10% Multi-Line Later. Source: Office of Technology Assessment 34a
PAGE 45
. Option E: Hedge Conversion. Option E is similar to option D except that the singleto multi-line conversion would take place only if ZIP+4 use is low at a specified future time (defined here as year-end 1987). Both options D and E include the same initial decision to purchase Phase II single-line OCRs, and to initiate release-loan testing of and any necessary research on conversion. The difference is that under option D, the conversion would be made regardless of the level of ZIP+4 use, while under option E, conversion would take place only if use is low. Option F: Cancel Phase II and ZIP+4. Option F is to cancel the Phase 11 single-line OCR procurement, terminate ZIP+4, and use the single-line OCRs already purchased to process 5 digit ZIP mail. Option G: 50-50 Split Procurement. Option G is a hybrid option that would cancel the Phase 11 procurement, immediately reissue a Request for Proposals (RFP) for onehalf the number of single-line OCRs (202 instead of 403), and simultaneously initiate release-loan testing of the multi-line OCR and singleto multi-line conversion. A new RFP for procurement of the other half of the OCRs but using multi-line technology (201 multi-line OCRs) would be issued as soon as possible, probably in about 2 years, at which time the then existing single-line OCRs (252 from Phase I and 202 from Phase 11) would be converted to multi-line. Option H: 90-10 Split Procurement. Option H is similar to option G except that the Phase II RFP would be reissued for 90 percent of the single-line OCRs (363), rather than 50 percent, and release-loan testing would be initiated on multi-line OCRs leading to a new RFP for procurement of the other 10 percent of the OCRs (40) using multi-line technology. 35
PAGE 46
. Discussion of Key Assumptions In developing and modeling the decision options, OTA made a variety of assumptions. The starting point for the OTA analysis was the January 1984 cost, savings, and cash flow projections for single-line OCR procurement (and related equipment including bar code readers and extended ZIP retrofit kits) provided by USPS to the Postal Board of Governors. Using the USPS data as a base, key assumptions were adopted as is or modified as necessary to fit the decision options analyzed by OTA. These key assumptions are discussed below by option or groups of options. AllOptions o Time horizon. OTA assumed a 14-year time horizon, the same as was used by USPS. Thus, cash flows and ROI/NPV precalculated for the 1985-1998 time period. o Labor cost escalation rate. OTA assumed a 7.42 percent annual escalation in clerk/carrier labor costs, as was used by USPS. This escalation rate is based on a 10-year historical average. o Baseline cost and savings projections. OTA used the USPS cost and savings projections for single-line OCRs and related equipment. These projections were adjusted for the various options depending on extent and timing of single-line OCR deployment and uncertainties (where applicable) in ZIP+4 use and savings rate. o Discount rate. OTA assumed a 15 percent discount rate (or required ROI), as was used by USPS. Compared to the U.S. Governments cost of capital (estimated to be 12.0 to 12.4 percent based on yields on U.S. Treasury bonds maturing in 1998-2001), the USPS discount rate appears to be reasonable. o Phase 11A procurement. USPS has identified a possible future procurement of automation equipment as Phase IIA. OTA has excluded this from all options and limited analysis to Phases I and II. 36
PAGE 47
. . All Options Using ZIP+4 (Options A, B, D, E, G, H) o Incentive rates for ZIP+4 use. OTA assumed that the incentive rates offered to volume mailers who use ZIP+4 (0.5 cent for presorted first class and 0.9 cent for nonpresorted first class) is a cost. USPS argues that this is a return to mailers and thus a benefit of ZIP+4, not a cost. However, OTA concluded that the incentive rates are required to get large mailers to convert to ZIP+4, and are therefore appropriately considered a cost. OTA assumed a fixed incentive rate, and that escalating rates would not be necessary to maintain a given level of ZIP+4 usage. Based on these incentives and a detailed mail flow analysis, the General Accounting Office (GAO) estimated an annual cost for incentive rates of $140 million at 90 percent ZIP+4 usage (GAO, 1983a, p.152). OTA assumed the GAO estimate, with the cost reduce d proportionately at lower ZIP+4 usage levels. o Savings as a function of ZIP+4 use. OTA assumed for these options that some portions of clerk/carrier savings resulting from automation is a function of the level of ZIP+4 use. For single-line OCRs, OTA used the curve of savings versus ZIP+4 use developed by USPS. For multi-line OCRs, OTA assumed that the USPS curve was pessimistic and developed additional curves (median and optimistic). These curves are presented and discussed in a later section. All options using multi-line OCRs (Options B, C, D, E,G,H). o OTA concluded that USPS assumptions about multi-line OCR performance were pessimistic with respect to the multi-line OCRs ability to read, code, and sort 5-digit ZIP mail to the 9-digit level. OTA developed additional curves, as noted above and discussed later. All options using singleor multi-line OCRs (options A, B, C, D,E,G, H). o OTA concluded that the USPS baseline estimates of clerk/carrier savings were likely to be optimistic, for a variety of reasons discussed later. Therefore, OTA analyzed savings at 100 percent, 90 percent, and 80 percent of the USPS estimates. Additional assumptions that apply to individual options are presented below. Option A: Single-line OCR. OTA used the USPS cash flow estimate as the base line, and treated ZIP+4 usage as an uncertainty. During 1985-1988, the deployment period for single-line OCRs, OTA reduced USPS savings estimates by an amount proportional to reduced ZIP+4 usage. For example, if projected clerk/carrier savings in 1986 were based on an USPS-assumed ZIP+4 usage of 57 percent, but estimated by OTA 37
PAGE 48
. .to be 28 percent, the savings at the lower ZIP+4 level would be calculated at 28/57 = 49 percent of the USPS estimate. Beyond 1989, when the single-line OCRs would be fully installed and operational, OTA estimated savings by using the USPS curve of clerk/carrier savings versus ZIP+4 use. Option B: Multi-line OCR with ZIP+4. Here, OTA assumed that for the next three years, 1985-1987, only Phase I single-line OCRs would be in operation, and the Phase I cost and savings estimates apply. The Phase I single-line OCRs would be converted to multi-line at an estimated conversion cost of $200,000 each, with the cost spread equally over three years, 1988-90, based on best available engineering judgment. OTA estimates that the cost of Phase II multi-line OCRs would be $850,000 each (capital and expense), again based on engineering judgment, and compares to the USPS-estimated unit cost (capital and expense) of $750,120 for Phase 11 single-line OCRs. OTA assumed that the cost of multi-line OCRs would be spread overthrew years, 1988-90, which reflects a 3-year delay period (1985-87) for release-loan tests, competitive bidding, and contract award, and that the total number of OCRs would be the same, whether singleor multi-line. OTA assumed an additional multi-line OCR cost of $5 million per year for 3 years, 1985-87, to cover any further research and development and the release-loan testing of multi-line OCRs prior to procurement. The $5 million equates to about one-fifth of the 1983 USPS R&D budget. Otherwise, OTA assumed that Phase II equipment costs (bar code sorters, electronic ZIP retrofits, site preparation, address directory information update, and contingency) would be the same as for single-line (Option A). OTA assumed that savings from the multi-line OCRs would phase in over 1988-90 period and that full savings would begin in 1991, the year following 38 the full
PAGE 49
.. installation. Multi-line savings at a given level of ZIP+4 use are based on the curve of savings versus ZIP+4 use presented later. The curve is treated as an uncertainty, with pessimistic, median, and optimistic performance assumptions built in. Option C: Multi-line OCR without ZIP+4. Option C is based on the same assumptions as Option B, except that the cost of the ZIP+4 rate incentive is dropped (since ZIP+4 would be terminated) and savings are estimated based on zero use of ZIP+4. For 1985-87, OTA assumed that savings would be the same as for the Phase I single-line OCRs with zero ZIP+4 use. For 1988-90, OTA assumed that one-third of the full savings benefit of Phase I single-line OCR conversion to multi-line would be realized in 1989, and two-thirds of the savings benefit in 1990. OTA assumed that Phase II multi-line OCR savings would phase in over 1988-90, and that full savings of converted Phase I OCRs and Phase II multi-line OCRs would begin in 1991. Option D: Automatic Conversion. Here, OTA assumed that single-line OCRs would be purchased and installed on the sam e schedule as in Option A. Over the 1985-8 7 period, R&D and release-loan testing on singleto multi-line conversion would be conducted at $5 million per year. All single-line OCRs (Phase I and Phase II) would be converted to multi-line at a total cost of $31 million ($200,000 per conversion times 655 units) spread over 3 years, 1988-90. Clerk/carrier savings are assumed to be the same as option A savings through 1990, and the same as option B from 1991 on. Option E: Hedge Conversion. In option E, OTA assumed conversion of single-line OCRs to multi-line only if ZIP+4 use is low at the end of 1987. If ZIP+4 use is at the high or median level, conversion would not take place and option E would be the same as option A except for a $5 million per year R&D and release-loan cost for 3 years, 198587. If ZIP+4 use is low, then conversion would take place and option E would be the same as option D. 39
PAGE 50
. Option F: Cancel. Here, OTA assumed that the Phase 11 OCR procurement would be cancelled, as would the ZIP+4 program and related rate incentives. The 252 singleline OCRs purchased in Phase I would be used to process 5-digit ZIP mail. OTA assumed that maintenance and spare parts costs would be the same, but that there would be no address directory information update cost. Clerk carrier savings for option F were assumed to be the same as with zero percent ZIP+4 use; that is, about 21 percent of the savings achievable at 90 percent ZIP+4 use, per USPS estimates. OTA used option F as the baseline against which incremental cash flows of other options can be measured. Option G: 50-50 Split Procurement. For option G, OTA assumed that the 252 Phase I single-line OCRs would be converted to multi-line in 1988-90 (as in options B, C, D); the Phase 11 procurement would be split, with 202 additional single-line OCRs purchased now, installed in 1985-87, and converted to multi-line in 1988-90 (same as option D except at one-half the number of single-line OCR units); and 201 multi-line OCRs would be purchased option B except at Thus, option one-half G is an after release-loan testing and installed in 1988-90 (same as the number of multi-line OCR units). intermediate option between options B and D and would be expected to roughly split the difference between the two. Option H: 90-10 Split Procurement. Option H is based on the same assumptions as option G except for the Phase 11 split: 363 (instead of 202) additional single-line OCRs would be purchased now, installed in 1985-87, and converted to multi-line in 1988-90; and 40 (instead of 201) multi-line OCRs would be purchased after release-loan testing and installed in 1988-90. 40
PAGE 51
Discussion of Key Uncertainties OTA gave explicit consideration to several uncertainties in the decision analysis. These included uncertainty about ZIP+4 usage, multi-line OCR performance, singleand multi-line savings rate, multi-line OCR cost, singleto multi-line conversion cost, multiline ZIP+4 use, national directory feasibility and cost, single/multi-line OCR obsolescence, and multi-line OCR procurement delay. The ZIP+4 usage, multi-line performance, and single/multi-line savings rate have proven to be the most controversial uncertainties. These are discussed first, followed by the less controversial uncertainties. As shown in figure 5, the three most controversial uncertainties were included in many of the decision options. ZIP+4 usage was treated as an uncertainty in analysis of options A, B, D, E, and, by extension, G and H. Multi-line OCR performance was treated as an uncertainty in analysis of options B, C, D, E, G, and H. The single/multi-line savings rate was treated as an uncertainty in options A, B, C, D, E, G, and H. ZIP+4 Usage. USPS has based its analysis on the assumption that 90 percent ZIP+4 usage will be achieved by the end of 1988. The 90 percent is calculated as a of total machinable, metered first class letter mail of about 51 billion pieces. achieve 90 percent by 1988, USPS assumed the following interim usage rates: percentage In order to 41
PAGE 52
. Decision Opt ion Figure 5 Uncertainties Included in Analysis of Decision options A. B. c. D. E. F. G. H. Single-Line Multi-Line with ZIP+4 Multi-Line Without ZIP+4 Automatic Conversion Hedge Conversion Cancel 50-50 split Procurement 90-10 split Procurement Multi-Line Single-/MultiZIP+4 OCR Line Usage Performance Savings Rate Yes No Yes Yes No Yes Yes No Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes No Yes Yes Source: Office of Technology Assessment 41a
PAGE 53
.. ... ZIP+4 Pieces Year Ending Usage Rate Of Mail 1984 27 percent 13.8 Billion 1985 48 24.5 1986 66 33.7 1987 85 43.4 1988 90 45.9 As of late May 1984, 59 mailers had actually converted their mailing address files to ZIP+4. Of these, 42 had qualified for ZIP+4 discounts and are expected to generate 401 million pieces on an annual basis. Sixteen of the other 17 are expected to collectively generate 25 million pieces of non-qualifying ZIP+4 first class mail, and the seventeenth, 200 million pieces of third-class mail. Another 258 mailers had given USPS definite commitments to convert to ZIP+4 by December 1984. When fully converted, these 258 mailers are expected to generate a total of 2.1 billion pieces of first class mail annually. Thus, at present, the projected actual volume of ZIP+4 mail (as a 1984 year-end volume and percentage of total machinable, metered first class mail) is about 2.73 billion pieces or about 19.8 percent of the original USPS projection. It impossible, of course, that additional mailers will decide to convert before the end of 1984. It is also possible that some of these already committed to convert will not actually do so. As shown in figure 6, the original USPS projection of 90 percent ZIP+4 usage within 5 years is considerably more optimistic than actual digit ZIP code or the Canadian 6-digit postal code. reach 90 percent usage; after 5 years, the 5-digit experience with either the U.S. 5The 5-digit ZIP took 12 years to usage level was about 51 percent. Thus, the USPS projection shows ZIP+4 reaching 90 percent about two and one-half times as fast (in 5 years rather than 12). 42
PAGE 54
Figure 6 100 90 80 40 30 20 Comparison of Code Usages U.S. 5 Digit, Zip+4, and Canadian SOURCE : United States Postal Service 42a
PAGE 55
USPS justifies its projection on the grounds that the technology used by mailers today is dramatically more sophisticated than that of the 1960s, when the 5-digit ZIP code was introduced. The widespread use of the 5-digit ZIP code, the proliferation of office automation and automated mailings, and the ZIP+4 incentive all point to successful and accelerated acceptance of ZIP+4." (USPS, Jellison, 1984c). In 1983, GAO reviewed the one ZIP+4 market study cited by USPS (a 1982 survey conducted for USPS by R.H. Bruskin Associates). GAO expressed reservations about the study methodology (response rate too low, study universe not representative) and was unable to endorse the study results (GAO, 1983b, pp. iv-v, 31-36). After reviewing all available evidence, OTA concluded that the USPS ZIP+4 projection should be considered optimistic, that inappropriate median estimate would be the 5-digit ZIP growth pattern, and that an appropriate pessimistic estimate would be a growth pattern similar to that of the USPS Electronic Computer Originated Mail Service (E-COM), where actual usage was about one-third of USPS projections. At the present time, the first year ZIP+4 usage could turn out to be even more pessimistic. The estimated 2.73 billion pieces of ZIP+4 first class mail at year end 1984 represents about 5.4 percent of the target mail base as compared to about 7 percent under the pessimistic scenario, 13 percent under the median, and 28 percent under the optimistic scenario. OTAs ZIP+4 growth curves are shown in figure 7. OTA assumed that there is a 5 percent chance that actual ZIP+4 usage will equal or exceed the USPS projection (the high growth curve), a 50-50 chance that actual usage will be above or below the median growth curve (that is, it is equally likely that ZIP+4 usage will be above or below the 5digit ZIP growth curve), and a 5 percent chance that ZIP+4 usage will be equal to or less than the low growth curve. At the present time, ZIP+4 growth is tracking a growth curve lower than the low curve in figure 7. 43
PAGE 56
z z u 4 u k c 43a . L Li3
PAGE 57
As a final note, GAO recently surveyed six business associations whose members mail large volumes of first class mail. Although not a representative sample, GAO found that while there is some interest in ZIP+4, businesses are still concerned about the cost of converting their address files (even with the current rate incentives) and whether the USPS presort program (and discounts) will end as a result of ZIP+4. In order to promote conversion, USPS is allowing "comingling" or mixing of up to 15 percent non-ZIP+4 mail in a ZIP+4 presort first class mailing until February 1, 1985, and up to 10 percent until October 1, 1985. As yet, however, there is little evidence that business mailers are giving ZIP+4 conversion a high priority. Multi-line OCR performance. A major advantage of multi-line OCRs is the ability to read, code, and sort 5-digit ZIP mail to the 9-digit level. That is, unlike the singleline OCR, the multi-line OCR can process a significant percentage of 5-digit ZIP mail as if the ZIP+4 were being used but without actually requiring the ZIP+4 code to be on each letter. The multi-line OCR does this by matching the multi-line address information on the envelope with address and ZIP+4 information stored in a computerized address directory. Even though there is no ZIP+4 code on the envelope, when a match is made, the multi-line OCR prints the 9-digit barcode on the envelope. At issue is not whether but how well the multi-line OCR can process 5-digit mail to the 9-digit level. USPS has estimated that the multi-line OCR can process 60 percent of 5-digit mail accepted by the OCR to the 9-digit level extra based on acceptance tests of the REI equipment. However, USPS notes that the 60 percent is "a projection that was not fully tested. Based on the 60 percent multi-line performance estimate (5-digit to 9digit level) and more complete data available on singleand multi-line OCR processing of 9-digit (ZIP+4) mail, USPS developed a set of curves shown in figure 8 as alternatives A, 44
PAGE 58
. II v 0 0 C4 0 o 7 dIZ 44a
PAGE 59
. B [low], and C. These curves show the estimated annual dollars) as a function of ZIP+4 usage for single-line OCRs and without ZIP+4. clerk/carrier savings (in 1989 and for multi-line OCRs with As estimate discussed previously in the technical analysis, OTA concluded that the USPS of 60 percent was pessimistic. OTA assumed that there was only a 5 percent chance that actual multi-line performance would be equal to or less than this low estimate. OTA concluded that a reasonable median estimate of performance was 65 percent, with a 50-50 chance that actual performance will be above or below, and that a reasonable high estimate was 75 percent, with a 5 percent chance that actual performance would equal or exceed this level. The multi-line OCR savings curves associated with low (60 percent), median (65 percent), and high (75 percent) performance are shown in figure 8. OTA used the USPS multi-line OCR savings curve as the low performance curve (marked as Alt. B (low) in figure 8), and developed new savings curves for median and high multi-line OCR performance (marked as Alt. B [median] and Alt B. [high], respectively, in figure 8). The x-intercepts of the three multi-line curves (savings at O percent ZIP+4 usage) correspond to about 67, 73, and 83 percent of the single-line OCR savings at 90 percent ZIP+4 usage. Thus, based on this set of curves, at zero percent ZIP+4 usage the annual savings estimates are approximately $580 million, $635 million, and $720 million for the three multi-line OCR alternatives, as compared to about $230 million for the single-line OCR alternative. calculated by $870 million, (Note: The savings percentage at O percent ZIP+4 for Alt. B [low] was dividing $580 million, the multi-line savings at O percent ZIP+4 usage, by the single-line savings at 90 percent ZIP+4 usage equals approximately 67 percent. Then, to estimate the multi-line savings percentage at 65 percent and 75 45
PAGE 60
l .* percent performance levels corresponding to Alt. B [median] and Alt. B [high], respectively, simple ratios were calculated: 65 percent/60 percent X 0.67 equals approximately 73 percent for Alt. B [median], and 75 percent/60 percent X 0.67 equals approximately 83 percent for Alt. B [high].) As is evident from figure 8, the USPS multi-line with ZIP+4 curve (Alt. B. [low]) has an elbow in it, with no increase in the savings level occurring until ZIP+4 usage exceeds about 20 to 25 percent. USPS defends this elbow on the grounds that up to about 20 percent ZIP+4 usage, the read redundancy in the address and the ZIP+4 code negates any advantage from the multi-line OCR. [n other words, USPS believes that the higher quality mail will be the first to use ZIP+4, and thus there will be no immediate benefit from multi-line processing. OTA was unable to establish a satisfactory engineering justification for this redundancy effect, and USPS was unable to provide a detailed explanation. Therefore, while OTA included the elbow in the USPS-estimated curve used as multi-line alternative B [low], OTA excluded the elbow for alternatives B [median] and B [high]. For these multi-line alternatives, OTA assumed a linear relationship between ZIP+4 usage and savings. For modeling purposes, OTA converted the figure 8 curves into a set of normalized linear equations using ZIP+4 usage as the independent variable and usage savings factor as the dependent variable. A usage savings factor of 1.0 equates to 100 percent of the savings projected for the single-line OCR alternative at 90 percent ZIP+4 usage. The set of curves corresponding to the linear equations is shown in figure 9. The slope of the single-line OCR curve was adjusted slightly to be consistent with the ZIP+4 sensitivity analyses included in the 1984 USPS proposal to the Postal Board of Governors (savings factors of 1.0, 0.866, and 0.72 at ZIP+4 usage rates of 0.9, 0.76, and 0.57 [corresponds to 90 percent, 76 percent, and 57 percent ZIP+4 usage]). 46
PAGE 61
. -. Figure 9 Usage Savings Factors 0 I 1 I I I 1 I I i .1 .2 3 .4 .5 ZIP+4 USAGE (U) .6 .7 .8 .9 1.0 1.0875 Alto A = Single-Line OCR Alt. B= Multi-Line OCR with (low) performance (median) performance (high) performance SOURCE: Office of Technology Assessment 46a
PAGE 62
* .* Single/multi-line savings rate. In addition to uncertainty over ZIP+4 usage and multi-line OCR performance, OTA included an uncertainty over the baseline level of clerk/carrier savings estimated by USPS. USPS asserts that their estimates are conservative, since the estimates do not include additional savings from a reduction in operator scheme training, increased employee flexibility, error reduction, or the potential for encoding mail manually. USPS also points out that the savings estimates are based on machinable first-class mail only, and do not include savings from nonmachinable first class mail or from semi-automated processing of ZIP+4 coded flats and parcels (for example, by using a wand reader or laser scanner). On the other hand, the USPS savings estimates are vulnerable to lower than anticipated volume and/or higher than expected costs. Some large mailers may develop techniques for totally bypassing the OCRs, for example, by applying bar codes to outgoing letters at the same time addresses are printed. This would be an extension of presorting letters to the carrier route level, which is apparently already competing with postal automation. [f a significant amount of machinable, easy to read (trayed and clean) first class mail is presorted and bypasses the OCRs, the OCRs would be left with a higher relative volume of lower quality mail with a lower OCR performance level to be expected. The net result could be a considerable decrease in OCR productivity and savings. This downside potential is dramatized by comparing single-line OCR accept rates (percentage of mail fed to an OCR that is accepted by the machine) for different types of mail. USPS assumed, for example, an accept rate of 90 percent for presort first class mail but only 50 percent for collection box mail and 75 percent for bulk business mail. To the extent the OCR mail mix changes such that presort decreases relative to other
PAGE 63
mail types, then the overall all other things being equal. As to the possibility OCR accept, productivity, and savings rates would decline, of higher than expected costs, the greatest vulnerability appears to be in the labor area, not in equipment. Although it is too early to have hard figures, it is possible that OCR maintenance labor costs will be higher than anticipated. All factors considered, OTA concluded that the USPS baseline estimate of clerk/carrier savings was probably somewhat optimistic. OTA assumed that there is a 5 percent chance that the actual savings rate will equal or exceed 100 percent of the USPS baseline savings estimate, a 50-50 chance that the savings rate will be above or below 90 percent of the USPS baseline estimate, and a 5 percent chance that the actual rate will be equal to or less than 80 percent of the USPS estimate. Multi-line OCR Cost. OTA noted some uncertainty about the purchase and maintenance costs of the multi-line OCR. Firm estimates are not possible in the absence of a competitive bidding on and operational experience with a large number of multi-line OCRs. USPS has estimated that multi-line OCRs would cost $200,000 more per unit to buy than single-line OCRs, or about $950,000 compared to the $750,000 (per single-line OCR unit (capital and expense) used in the USPS Governors. However, based on best available engineering $100,000 purchase cost difference is more realistic. proposal to the Postal Board of judgment, OTA concluded that a Also, analysis on multi-line OCR purchase prices of $750,000, found that the impact on ROI/NPV was negligible, as will be OTA conducted a sensitivity $850,000, and $950,000 and discussed in a later section. 48
PAGE 64
* As for multi-line OCR maintenance cost, OTA concluded that the cost of updating the multi-line OCR address directory would probably be greater than updating the singleline OCR directory. USPS has very roughly estimated this additional maintenance cost at one work year per local directory per year (equivalent months). Since 209 local directory updates would be projected to have OCRs), the additional yearly directory to three persons working for 4 needed (the number of cities maintenance cost is estimated to be about $9.32 million (209 local directory updates times $44.6K per average work year). This amount is negligible (on the average one percent or less) compared to the total yearly multi-line savings, and therefore was excluded from further analyses. Singleto multi-line conversion cost. assumed a conversion cost of $200,000 per judgment. Multi-line ZIP+4 usage. USPS believes In the absence of firm estimates, OTA unit, based on best available engineering that use of multi-line OCRs would likely have a detrimental effect on mailer acceptance and use of ZIP+4. In other words, all other things being equal, USPS believes that ZIP+4 usage would be less for a multi-line OCR system than for a single-line OCR system. While a few mailers have indicated that a USPS switch to multi-line OCRs would reduce the likelihood of converting to ZIP+4, the cost of conversion, level and stability of USPS ZIP+4 rate incentives, and relationship to USPS presort rate incentives appear to be much more important to mailers than the type of OCR equipment used by USPS. [n sum, the available evidence suggests that mailers base their decisions about ZIP+4 use primarily on economic and financial factors and not on technological factors. 49
PAGE 65
. In addition, even the concerns stated by USPS would appear to be moot for options including an initial purchase of some or all of the Phase II single-line OCRs and subsequent conversion to multi-line capability. Since actual conversion would probably not begin for about 3 years, by that time it should be clear to what extent USP S projections of ZIP+4 usage with a single-line OCR system are being met. National directory feasibility and cost. OTA originally included the cost of charge against multi-line OCR options, on the be necessary for effective use of multi-line OTA concluded that national directories, while computerized national directories as a theory that national directories would OCRs. However, upon further analysis, technically feasible, were not necessary, since and destination cities amounts to a de facto national directory has not yet been developed, As a result of these factors, the cost of national directories was removed from further consideration. (The cost of local directories was, of course, reflected in the estimates of multi-line OCR purchase price and singleto multi-line conversion price.) the use of local directories in the origin national directory. In addition, since a estimating costs would be very difficult. Single/multi-line OCR obsolescence. OTA originally included as an uncertainty the obsolescence date for Phase II single-line OCR equipment (i. e., a 5 percent chance of becoming obsolete in 1994 or earlier). However, OTA concluded that the single-line OCRs could be upgraded, if necessary, to use new technologies and/or perform new or expanded functions. Therefore, the equipment obsolescence date was excluded as an uncertainty. All equipment was assumed to remain operational through 1998. Multi-line OCR procurement delay. OTA originally assumed a 2-year delay in OCR procurement if USPS switched from single-to multi-line OCRs. Upon further analysis. OTA concluded that a 3-year delay was more reasonable, based on best available 50
PAGE 66
engineering judgment and USPS experience with single-line OCR release-loan testing, assuming no change in USPS procurement-practices. USPS has likewise estimated a 3year delay as more realistic, given the need to issue a new solicitation for release-loan testing of multi-line OCRs. Results of Decision Analysis Probabilistic cash flow models were developed for each option using the assumptions and uncertainties discussed previously. The models were then used to project estimated yearly cash flows, rate of return, and net present value for each decision option. (See appendices A and B for a detailed presentation of the models and cash flows for all options except options G and H, which were estimated by interpolation. ) The purpose of the modeling is to provide both absolute and comparative financial projections for each decision option. However, while the models generate numbers that appear to be very specific, the user of the results must understand that all projections are subject to some imprecision, especially in view of the large number of assumptions and uncertainties. But as long as these assumptions and uncertainties are treated consistently, the results should provide a valid basis for comparison among options. Results of the decision analysis are presented below in the following order: rates of return, net present values, supplemental sensitivity analysis, and overall cash flows. Rates of return. The cash flow models were used to estimate internal rates of return (ROIs) for each option under each condition of uncertainty. The ROIs were estimated on an incremental basis over option F (cancel), since OTA assumed that under 51
PAGE 67
. any scenario, the Phase I single-line OCRs already purchased would be kept in service. Thus option F is in effect the baseline option. In essence, all ROIs are net of cash flows associated with Phase I single-line OCRs. The estimated ROIs for all options under various conditions are shown in figure 10. Every option under all conditions modeled shows an ROI above the 15 percent threshold established by USPS. The lowest ROI is 20.6 percent, for option A (single-line OCR) under low ZIP+4 usage and a low savings rate. Only if ZIP+4 was even lower (e.g., peaked out at 20 percent usage instead of the 40 percent assumed in the low usage scenario) and/or the savings rate was even lower (e.g., 70 percent of USPS estimates instead of the 80 percent assumed in the low savings rate scenario) would the option A ROI drop below 15 percent. The highest ROI is 84.6 percent, for option D (automatic conversion) under high ZIP+4 usage, multi-line performance rate, and savings rate. However, under these conditions options A (single-line OCR), E (hedge conversion), and H (90-10 split procurement) have only a slightly lower estimated ROI. The greatest difference between option D (automatic conversion) and option A (single-line OCR) occurs with low ZIP+4 usage and high multi-line OCR performance. Under these conditions, the option D ROI is anywhere from 33 to 50 percent higher than the option A ROI. The relative ROI ranking of the various options as a function of ZIP+4 usage is as follows (excluding option F, cancel, which has a negative ROI): 52
PAGE 68
Figure 10 Estimated Percentage Rates of Return for Decision Options Under Various Conditions Multi-line OCR Performance Rate Low Median High I I I ROIs for Option E are the same as Option A at high and median ZIP+4 usage and the same as Option D at low usage. ROIs for Options G and H were calculated by interpolation from Options B and D (i.e. Option G was assumed to have an ROI that split the difference between Options B and D). NOTE: All ROIs are incremental over Option F (cancel) Source: Office of Technology Assessment 52a
PAGE 69
Option D Option H Options A and E Option G Option B Highest ROI Lowest ROI Highest ROI Low ZIP+4 Usage Option C Options D and E Option H Option G Option B Option A Lowest ROI Thus, at high or median ZIP+4 usage, options D and H have the highest ROIs. And at low ZIP+4 usage, options C, D, E, and H have the highest ROIs. more ROI, Net present values. Use of ROIs for decisionmaking has a serious limitation. When than one option clears the hurdle rate (that is, has more than the minimum required in this case 15 percent), the ROI itself gives no indication of the cash flow differences of the various options as a basis for comparing the options. An alternative to ROI frequently used in capital investment decisionmaking is net present value (NPV). NPV discounts the cash flows of each option at the hurdle or threshold rate, in this study 15 percent. Estimated NPVs for all options under all conditions (except option F, cancel, which has a negligible NPV of $232,199) are shown in figure 11. The relative ranking of the options based on NPV is the same as the ranking based on ROI, except for option C. However, there is a significant difference in the absolute rankings when using NPV. 53
PAGE 70
Figure 11 Estimated Net Present Values (in billions of dollars) for Decision Options Under Various Conditions Multi-line OCR Performance Rate Low Median High I L I M I H I L I M I H I L I M H I L M I H I I I High ZIP+4 Usage Option A 1.50 1.91 2.33 Option B 1.07 1.42 1.77 1.12 1.48 1.84 1.14 1050 1.86 Option D 1.51 1.93 2.35 1.57 1.99 2.42 1.59 2.01 2.44 Option E* 1.50 1.91 2.33 Option G** 1.29 1.68 2.06 1.35 1.74 2.13 1.36 1.76 2.15 Option H** 1.47 1.88 2.29 1.53 1.94 2.36 1.55 1.96 2.38 Median ZIP+4 Usage Option A 1.05 1.38 1.71 Option B 0.82 1.12 1.41 0.90 1.21 1.51 0.93 1.24 1.55 Option D 1.09 1.43 1.77 1.17 1.52 1.87 1.19 1.55 1.90 Option E* 1.05 1.38 1.71 Option G** 0.96 1.28 1.59 1.04 1.37 1.69 1006 1.40 1.73 Option H** 1.06 1.40 1.73 1.14 1.49 1.83 1.16 1.52 1.87 Low ZIP+4 Usage Option A 0.21 0.41 0.61 Option B 0.30 0.50 0.70 0.59 0.83 1.07 0.70 0.95 1.20 1.03 1.30 0.86 1.15 1.43 Zero ZIP+4 Usage Option C 0.58 i 0.78 0.99 0.69 0.91 1.13 0.87 1.11 1.36 NPVs for Option E are the same as Option A at high and median ZIP+4 use and the same as Option D at low usage. NPVs for options G and H were calculated by interpolation from Options B and D (i.e. Option G was assumed to have a NPV that split the difference between Options B and D). NOTE: All NPVs are incremental over Option F (cancel). Source: Office of Technology Assessment 53a
PAGE 71
Under conditions of high and medium ZIP+4 usage, high savings, and high multi-line performance, option D (automatic conversion) has about a 5 percent and 11 percent higher NPV, respectively, than option A (single-line OCR). At low ZIP+4 usage, another things being equal, the option D advantage increases to a substantial 134 percent or about $820 million in NPV. At a low savings rate (along with low ZIP+4 usage and high multi-line performance), the relative advantage of option D over A increases further to about 310 percent although the absolute advantage decreases to about $650 million in NPV. Even at low multi-line performance, with low ZIP+4 usage option D has a 53 to 119 percent relative advantage in NPV and a $320 to 250 million absolute advantage in NPV, at a high and low savings rate, respectively. Option H (90-10 split procurement) also has a higher NPV than option A under almost all conditions. Option G (50-50 split procurement) has a significant although somewhat smaller advantage over option A at low ZIP+4 usage. The net NPV advantage or disadvantage of options D, G, and H compared to option A is shown in figure 12 for various conditions. Relative as well as absolute comparisons are included for low ZIP+4 usage. The results clearly show that options D and H have a marginally higher NPV at high and medium ZIP+4 usage under all conditions and a substantially higher NPV at low ZIP+4 usage. On the other hand, option G has a marginally lower NPV than option A at high ZIP+4 usage and median ZIP+4 usage with low or median multi-line performance, a marginally higher NPV at median ZIP+4 usage and high ZIP+4 performance, and a substantially higher NPV under all low ZIP+4 conditions. Another way to present these results is shown in figure 13. Here, the 80 percent credible values are shown for each option along with the overall net present value for 54
PAGE 72
Figure 12 Comparison of Net Present Values (in percentages and billions of dollars) of Options D, G, and H with Option A Multi-line OCR Performance Rate Low Median High Savings Rate Savings Rate Savings Rate Low Med. High Low Med. High Low Med. High High ZIP+4 Usage Option D ($) 0.01 0.02 0.02 0.07 0.08 0.09 0.09 0.10 0.11 Option G ($) -0.21 -0.23 -0.27 -0.15 -0.17 -0.20 -0.13 -0.15 -0.18 Option H ($) -0.03 -0.03 -0.04 0.03 0.03 0.03 0.05 0.05 0.05 Median ZIP+4 Usage Option D ($) 0.04 0.05 0.06 0.12 0.14 0.16 0.14 0.17 0.19 Option G ($) -0.09 -0.10 -0.12 -0.01 -0.01 -0.01 0.01 0.02 0.03 Option H ($) 0.01 0.02 0.02 0.09 0.11 0.12 0.11 0.14 0.16 Low ZIP+4 Usage Option D (%) 119 71 53 257 151 113 310 181 134 ($) 0.25 0.29 0.32 0.54 0.62 0.69 0.65 0.74 0.82 Option G (%) 81 46 34 209 127 92 271 156 116 ($) 0.17 0.19 0.21 0.46 0.52 0.56 0.57 0.64 0.71 Option H (%) 110 66 49 248 146 105 300 176 131 ($) 0.23 0.27 0.30 0.52 0.60 0.67 0.63 0.72 0.80 Notes: Option D= Automatic conversion. Option G = 50-50 Split procurement. Option H = 9O-10 Split procurement. All dollar figures in billions and are net of Option A NPV from Figure 11. All NPVs are incremental over Option F (cancel). Source: Office of Technology Assessment 54a
PAGE 73
Options A. Single B. Multi with ZIP+4 C. Multi no ZIP+4 D. Convert E. Hedge G. 50-50 Split Procurement H. 90-10 Split Procurement o Conditional H. 90-10 Split Procurement 54b Figure 13 Credible Intervals on NPV (Overall & Conditional) .5 1.0 1.5 2.0 NPV ($B E c.
PAGE 74
.each option. The NPV being above smallest value. ZIP+4 usage. Figure 13 and the range 80 percent credible interval means that there is a 10 percent chance of the largest value and a 10 percent chance of NPV being below the The NPVs and credible intervals are shown overall and conditional on shows, in effect, the ranking of the options by expected net present value of uncertainty in NPV associated with each option, both overall and conditional on ZIP+4 use. The ranking of the options by expected NPV is summarized below: Overall NPV Rank Option D H E G A B c Low ZIP+4 Usage Option D E H c G B A 1 highest 2 3 4 5 6 7 lowest 1 highest 2 tie 3 4 5 6 7 lowest Thus, option D (automatic conversion) ranks first in NPV both overall and with low ZIP+4 usage. Option H (90-10 split procurement) ranks second in NPV overall and third with low ZIP+4 usage. Option E for first with low ZIP+4 usage. OCR) rank almost identically in relatively low (6th) in NPV, both (hedge conversion) ranks third in NPV overall and ties Options G (50-50 split procurement) and A (single-line overall NPV. Option B (multi-line with ZIP+4) ranks overall and with low ZIP+4 usage. Option C (multi-line --
PAGE 75
without ZIP+4) ranks the lowest in overall NPV, but somewhat higher (4th) in NPV with low ZIP+4 usage. Option A (single-line) ranks the lowest in NPV with low ZIP+4 usage. The dominance of option D (automatic conversion) can also be illustrated by plotting the cumulative probability distributions of NPV for each option. As shown in figure 14, for any value of NPV (incremental over option F (cancel]), the probability is greatest for option D. For example, for an NPV of $1 billion, the probability is about 0.9 or 90 percent that option D will exceed that NPV (cumulative probability of about 0.1), about 0.75 that option A will exceed, about 0.7 that option B will exceed, and only 0.5 (or 50 percent) that option C will exceed $1 billion. Because option D dominants all other options under all conditions of uncertainty, option D is stochastically dominant. Supplemental sensitivity analysis. The basic models built in three uncertainties (ZIP+4 usage, savings rate, multi-line performance rate). A supplemental analysis was conducted to check the sensitivity of NPV results to changes in the purchase price of the multi-line OCR or the number of multi-line OCR units. The results, summarized in figure 15, showed that an increase in multi-line OCR purchase price from $850,000 to $970,000 would have very little effect on NPV. The effect would be to reduce NPV by about $0.02 and 0.03 billion. Likewise, an increase in the number of multi-line OCR units from 403 to 439 (as estimated by GAO to be required if the entire Phase 11 procurement was switched from singleto multi-line OCRs) would cost about an additional $34.9 million (41 units x 850,000/unit), which is less than the $40.3 million cost of a $120,000 price increase for 403 machines. Therefore, the effect on NPV again would be very little. 56
PAGE 76
. Figure 14 Cumulative Probability Distributions of Net Present Value (Smoothed) 1.00 .50 .40 .30 .20 10 0 Option G 50-50 Split Option H Procurement Option A 90-10 0 7 .4 .6 .8 1.0 .0 2.5 .2 Net Present Value (in $ billions) n Source: Office of Technology Assessment 56a
PAGE 77
Figure 15 Sensitivity of Net Present Value to Multi-line OCR Purchase Cost NPV* with Different Multi-line OCR Unit Costs Multi-line Savings I ($ in billions) ZIP+4 Use Performance Rate Rate @ $750K @ $850K @ $970K I HIGH High High 1.88 1.86 1.84 Median 1.52 1.50 I 1.48 Low 1.16 1.14 Median High 1.86 1.84 Median 1.50 1.48 1.12 1.81 1.46 Low 1.14 1.12 1.10 Low High 1.79 1.77 1.75 Median 1.44 1.42 1.40 Low 1.09 1.07 1.04 MEDIAN High High 1.57 Median 1.26 Low .95 Median High 1.53 Median 1.23 Low .92 Low High 1.43 Median 1.14 Low .84 LOW High High 1.22 Median l 97 Low .72 Median High 1009 Median I .85 Low .61 Low High .72 Median .52 Low .32 1.55 1.24 .93 1.51 1.21 .90 1.41 1.12 .82 1.20 .95 .70 1.52 1.21 .90 1.49 1.18 .88 1.39 1.09 .80 1.18 .93 l 68 1.07 1.04 .83 .80 .30 .27 NPV for option B (multi-line OCR with ZIP+4) adjusted to reflect different purchase costs. Source: Office of Technology Assessment 56b
PAGE 78
In sum, neither a modest the total number of units would increase in price per unit for the multi-line OCRs or in significantly change the NPVs. Overall and selected year cash flows. Net present value comparative quantitative evaluation of the decision options. is the best basis for However, the actual undiscounted net another dimension cash flows over the 13 year payback period (1985-98) can provide to the evaluation. Comparative net cash flows for selected options and conditions are shown in figure 16. Option A (single-line) is estimated to show positive cash flows of $8.8. $8.24, and $3.57 billion at high, medium, and low ZIP+4 usage. At high ZIP+4 usage, option B (multi-line with ZIP+4) conversion) and H (90-10 billion respectively, and is somewhat lower at $8.14 billion, options D (automatic split procurement) somewhat higher at $9.36 billion and 9.24 option G (50-50 split procurement) about the same at $8.75 billion. The comparisons between options change relatively little at median ZIP+4 usage. However, at low ZIP+4 usage there is a substantial difference in net cash flows. Option A (single-line) shows a net cash flow of $3.57 billion. But, depending on the multi-line OCR performance rate, options D (automatic conversion) and H (90-10 split procurement) show a net cash flow of $5 to 7.2 billion, or about $1.4 to 3.6 billion greater than option A. Option G (50-50 split procurement) shows about $1.1 to 3.3 billion greater cash flow than option A, and option B (multi-line with ZIP+4) shows about $0.8 to 3.0 billion greater cash flow than option A. A comparison of yearly cash flows gives similar results. Yearly cash flows for selected options and conditions are shown in figure 17, for the years 1994-98. By this time, all equipment will presumably have been installed (or converted) and up and running 57
PAGE 79
Figure 16 Comparative Net Cash Flows, Selected Options and Conditions, 1985-1998 (in $ billions) Compared to Option A -0.65 Low ZIP+4 Usage High Savings Rate High Multi-line Performance Net Cash Flow +$3.57 +$6.57 Compared to Option A +3.01 Low ZIP+4 Usage High Savings Rate Median Multi-line Performance Net Cash Flow +$3.57 +$5.98 Compared to Option A +2.41 Low ZIP+4 Usage High Savings Rate Low Multi-line Performance Net Cash Flow +$3.57 +$4.38 Compared to Option A I +0.82 Option D Option G* Option H* (Automatic (50-50 split (90-10 split Conversion) Procurement) Procurement) $9.36 +$8.75 +$9.24 +0.56 -0.05 +0.44 +$9.03 I +$8.31 I +$8.89 +0.79 +0.07 +0.65 +$7.19 +$6.88 +$7.13 +3.62 +3.31 +3.56 +$6.59 +$6.29 +$6.54 +3.02 +2.72 +2.97 +$5.00 I +$4.69 I +$4.94 +1.43 +1.12 +1.37 Options G and H calculated by interpolating between Options B and D. NOTE: All net cash flow figures in undiscounted dollars. SOURCE: Office of Technology Assessment 57a
PAGE 80
Figure 17 Comparative Net Cash Flows, Selected Options, Conditions, and Years (in $ billions) High ZIP+4 Usage High Savings Rate High Multi-line Performance Net Cash Flow 1994 1995 1996 1997 1998 Low ZIP+4 Usage High Savings Rate High Multi-line Performance Net Cash Flow 1994 1995 1996 1997 1998 Low ZIP+4 Usage High Savings Rate Median Multi-line Performance Net Cash Flow 1994 1995 1996 1997 1998 Low ZIP+4 Usage High Savings Rate Low Multi-line Performance Net Cash Flow 1994 1995 1996 1997 1998 Option A (Singleline) $0.87 0.94 1.02 1.11 1.20 0.43 0.46 0.50 0.54 0.59 0.43 0.46 0.50 0.54 0.59 0.43 0.46 0.50 0.54 0.59 Options B (Multi-line) D (Automatic Conversion) G (50-50 Split Procurement) H (90-10 Split Procurement) $0.94 1.02 1.11 1.20 1.30 0.87 0.93 1.01 1.09 1.17 0.80 0.86 0.93 1.00 1.08 0.61 0.66 0.71 0.77 0.83 Options B, G, D, H I Compared to Option A I $0.07 0.08 0.09 0.09 0.10 0.44 0.47 0.51 0.55 0.58 I 0.37 0.40 0.43 0.47 0.49 0.18 0.20 0.21 0.23 0.24 Source: Office of Technology Assessment 57b
PAGE 81
at optimal performance. Options B, D, G, and H will by that time look exactly the same -all multi-line OCRs. The single-line OCRs procured under options D, G, and H will have been converted to multi-line capability. Option A will continue to be solely singleline OCRs. With high ZIP+4 usage, option A shows an annual net cash flow of about $870 million to $1.2 billion from 1994 to 1998. Options B, D, G, and H show almost identical annual cash flows, only slightly higher by about $70 to $100 million per year. However, at low ZIP+4 usage, the differences again become substantial. With high multi-line performance, options -B, D, G, and H show between $440 and 580 million per year additional net cash flow compared to option A, from 1994 to 1998. With median multiline performance, the advantage of options B, D, G, and H ranges from $370 to 490 million per year. And even at low multi-line performance, the advantage over option A, while reduced, is still significant at $180 to 240 million per year. 58
PAGE 82
The basic methodology employed was to develop probabilistic cash flow models for each of the decision options (except for the split procurement options, which were analyzed by interpolating from results of other options). A decision tree was developed for each option. The tree included the uncertainties discussed previously, as applicable to each specific option, and a cash flow valuation measured by rate of return (ROI) and net present value (NPV) discounted at 15 percent, as illustrated conceptually in figure A-1. The uncertainties (i.e., ZIP+4 usage, savings rate, multi-line OCR performance rate) were treated as continuous random variables. The continuous random distributions were approximated by the Pearson-Tukey approximation which uses values of the variable at three discrete points: the 5, 50, and 95 percentiles. Pearson-Tukey assigns probabilities of 0.185, 0.63, and 0.185 to these three percentiles, as shown in figure A-2. Simplified schematic models for options A, B, C, D, and E are shown in figures A-3 through A-7. The full models are shown in appendix B. The models were run on an IBM Personal Computer using Lotus 1-2-3 and proprietary software. 59
PAGE 83
.. Figure A-1 Probabilistic Cash Flow Model (Illustrated for Option A: Single-Line OCR) UNCERTAINTIES ZIP+4 USAGE CASH FLOW VALUATION SAVINGS RATE The uncertainties are continuous random variables. A simplified representation of these distributions is used in the analysis (as explained in figure A-2. Source: Office of Technology Assessment
PAGE 84
Continuous Figure A-2 Simplified Representation of Continuous Probability Distribution Distribution 5% F .05 F .95 VALUE .63 Pearson-Tukey Approximation of Continuous Distribution F .05 F .50 F .95 VALUE Model Representation of Approximation .9 5 .185 .185 This representation is a Pearson-Tukey approximation. It was chosen because it provides an excellent approximation to a wide range of continuous probability distributions, as explained in Keefer and Bodily (1983). Source: Office of Technology Assessment 61
PAGE 85
Figure A-3 Simplified Schematic Model for Option A: Single-Line OCR ZIP+4 USAGE High .185 Median .185 \ Low (See figures 7, 8, 9) SAVINGS RATE High (100%) .185 / Median (90%) .185 \ L OW (80%) SOURCE: Office of Technology Assessment 62
PAGE 86
. I I 63
PAGE 87
Figure A-5 Simplified Schematic Model for Option C: Multi-Line OCR Without ZIP+4 MULTI-LINE PERFORMANCE RATE \ Low (See figures 8 and 9 at O% ZIP+4 usage) SAVINGS RATE High (100%) / .185 / Same as for all other options SOURCE: Office of Technology Assessment. 64 r
PAGE 88
. Figure A Simplified Schematic Model for Option D: Automatic Conversion ZIP+4 USAGE MULTI-LINE PERFORMANCE RATE SAVINGS RATE SOURCE : Office of Technology Assessment
PAGE 89
; I w ,4 3 ii 7 c .. L Un m I m I h an 0 ./ \ Lr .0 I
PAGE 90
Appendix B: Detailed Cash Flow Models and Results The detailed cash flow model and results for options A, B, C, D, and F are presented in this appendix. For each option, a complete probabilistic event tree is shown, along with the incremental rate of return (ROI) and net present value (NPV) over option F (cancel) for each condition of all options modeled. Also for each option, detailed year-by-year cash flows are shown for each condition. The detailed cash flows show two total cash flows of the option modeled. less cash flows for Option F (cancel). All dollar figures are in thousands, ROIs and NPVs. The first ROI and NPV are for The second ROI and NPV are for total cash flows and are carried to the nearest thousand dollars. 67
PAGE 91
... .... Figure B-1 Cash Flow Model and Results for Option A: Single-Line OCR INCREMENTAL LABEL ROI (%) NPV (SB) ZIP+4 USAGE SAVINGS RATE 111 112 113 121 83.5 2.33 l 1 69.2 1.91 56.4 1.50 49.7 1.71 HIGH 185 .185 43.6 1.38 1.05 .185 LOW 131 .61 132 25.3 .41 133 .21 SOURCE : Off ice of Technology Assessment 68
PAGE 92
.Figure B-3 Cash Flow Results for Option C: Multi-Line OCR Without ZIP+4 MULTI-LINE PERFORMANCE RATE AT O% ZIP +4 SAVINGS RATE LABEL ROI (%) NPV ($B) 111 48.5 1.36 100% 185 HIGH (.83) 90% 185 80% 43.1 1.11 112 113 37.6 .87 45.3 1.13 100% .91 39.9 .18 80% 123 .69 .99 34.4 100% 131 .185 LOW (.67) 90% 132 .185 43.2 37.8 .78 80% 133 32.3 .58 SOURCE: Office of Technology Assessment 70
PAGE 93
Figure B-2 .185 Multi-Line With ZIP+4 .185 ZIP+4 USAGE Cash Flow Model and Results for Option B: Multi-Line OCR with ZIP+4 MULTI-LINE PERFORMANCE RATE LABEL 74.7 1.77 58.2 1.42 45.6 1.07 47.5 1.55 41.5 1.24 35 4 l 93 47.0 1.51 41.0 1.21 35.0 .90 45.7 1.41 39 7 1.12 33.7 .82 37. 8 1.20 33.8 .95 29.5 .70 36.1 1.07 32.1 .83 27.8 .59 30.8 26.7 22.3 .70 .50 .30
PAGE 94
Figure B-4 Cash Flow Results for option D: Automatic Conversion MULTI-LINE PERFORMANCE RATE Incrementa l ZIP+4 USAGE LABEL SAVINGS RATE 84.5 70.0 57.1 2.42 1.99 1.57 84. 2 69. 6 56. 6 2.35 1.93 1.51 .185 51.5 1.90 45.4 1.55 39.2 1.19 / .185/ 51. 2 45. 1 38. 8 1.8 7 1.5 2 1.17 i 1.77 50.3 44.1 37. 8 1.4 3 .185 1.09 39.4 1.4 3 35. 4 1.1 5 31.1 .86 I 38.1 34.0 29.7 1.30 1 03 .75 .93 34.0 29.8 25.3 .70 .46 SOURCE: Office of Technology Assessment 71
PAGE 95
t Table B-1 Detailed Cash Flows for Option A (Single-Line OCR) at High Savings Rate 111 I SOURCE: Office of Technology Assessment
PAGE 96
Table B-1 (continued) --------------,162,220 1,258,844 ,107,037 1,199,567 131 =x===sx=xc*=sz8======xm==xxg==*===:*================================== =========*================================a,=========a,=====.===*===========8==8======m===**=`==,=== SOURCE: Office of Technology Assessment I
PAGE 97
1 Table B-2 Detailed Cash Flows for Option A (Single-Line OCR) at Median Savings Rate 412,572 SOURCE: Office of Technology Assessment
PAGE 98
Table B-3 Detailed Cash Flows for Option A (Single-Line OCR) at Low Savings Rate 113 SEDUCED SAVINGS To 80% AVERAGE USEAGE 37.5% 57.0$ 75.5% 87.5$ 90.0$ 90.0; 90.0% 90.0% 90.0% 90.0% 90.0% 90.0% 90.0% 90.0% SOURCE: Office of Technology Assessment
PAGE 99
Table B-4 ; Detailed Cash Flows for Option B (Multi-Line with ZIP+4) at High Performance Rate and High Savings Rate 1111 (140,000) (140,000) (140,000) (140,000) SOURCE: Office of Technology Assessment I
PAGE 100
Table B-4 (continued) 1995 19% 1997 1998 1311 SOURCE: Office of Technology Assessment I
PAGE 101
1112 1212 1312 1113 1213 1313 Table B-5 Detailed Cash Flows for Option B (Multi-Line OCR with ZIP+4) at High Performance Rate and Median or Low Savings Rate 8=8==8s*888=c8==s=8s8888=88=x8===8=====*=8===*======z================= *================8=========88==88888888=8=B8ss8888=8==x8888x88*c8c=xs= ==sx=xx===s PRICE PER OCR DATE $850 NUMBER 1985 1986 1987 (131,525) (105,970) (66,809) (151,173) (129,829) (97,288) 33.8% RED NPV 655 i 1988 1989 1990 1991 1992 1993 1994 1995 19% 1997 1998 --------------------------------------------------(53,599) 14,436 144,947 --------------------631,176 688,397 749,864 815,892 886,819 1,501,261 963,009 1,044,853 1,132,769 (132,571) (76,647) 49,971 622,361 679,959 742,711 810,433 884,692 ,237,034 963,009 1,044,853 1,132,769 (173,044) (145,310) (101,990) 597,271 646,205 698,770 951,878 755,236 815,891 881,047 951,038 1,026,222 (151,201) (104,491) 937 524,022 573,524 627,519 685,925 749,954 817,742 888,806 965,143 927,103 (183,634) (161,290) (128,609) 508,115 550,434 595,893 644,725 697,181 753,528 814,057 879,077 699,882 l SOURCE: Office of Technology Assessment
PAGE 102
. Table B-6 Detailed Cash Flows for Option B (Multi-Line with ZIP+4) at Median Performance Rate and High Savings Rate 1121 ====x=====:==z=:=s=s=8===xx====:===z==x=xs========xxxxc===x=========== ========::=:=x:=xs:=x==s6xz::====az========x======x=a=8zsx==:z=ss=::== PRICE PER OCR NUMBER OF OCRs ESCALATION FACTOR BEGINNING OF YEAR AVERAGE USEAGE YEAR DATE USEAGE (140,000) (140,000) (140,000) (140,000) (140,000) (140,000) SOURCE: Office of Technology Assessment
PAGE 103
1 Table B-6 (continued)
PAGE 104
Table B-7 Detailed Cash Flows for Option B (Multi-Line with ZIP+4) at Median Performance and Median or Low Savings Rate ===========================*====:g--.-.z==============:s====::=====:== =====================:================================================ ======================:======= ---.PRICE PER OCR $850 NUMBER 655 l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... l ......0. . . ....*.*.** . . . . . . . l .......*. . . ....***... . . . . . l . . . . . . . SOURCE : Office of Technology Assessment
PAGE 105
Table B-8 I I I 1 I I I I I Detailed Cash Flows for Option B (Multi-Line with ZIP+4) at Low Performance Rate and High Savings Rate =z==ss===s===s==8==2xs8=8=s=88x=====s=8s=====c88=88=a=8=8====:======x= =====================================================*====8s===:===zs= ======88==ZX:====ZZS:=:=SS=:Z: SOURCE: Office of Technology Assessment I I I
PAGE 106
Table B-8 (continued) 1231 =xx*s:========s===*=88=========8====================8================= =:=: =:=z:=:ss==::===s=...-.-===z:=====:===x=88z======x==xs=s==x=::::z= -----:==a=scxz===s:xsz=====x==:==== NET CASH FLW CASH FL(ll PHASE 1331 =x8=8cmxsccx=zsc8*=88=2cacm8s*c=c==s8==========8m*===============a---xc====xx=x=xs===c====x=========*============s============s============ =ss==ssss2==8s========1======= SOURCE: Office of Technology Assessment
PAGE 107
I Table B-9 Detailed Cash Flows for Alternative B (Multi-Line with ZIP+4) at Low Performance Rate and Median or Low Savings Rate 88xxz=x8*zs8===c===a8s8888*===88=z==================================== ============================================================x88=xz==s*===28=====8z===a=z*8x8::===xa~ . . . . . . . . . . . . . . . . . . . . l ,.. . . . . . l . . . . . l =*******.. l ******.*.* l **...*.... .....00...: l *.eO**o.. ..0.:.0.... . . . . . . . . . . . . . . . . . . . . . . SOURCE: Office of Technology Assessment
PAGE 108
Table B-10 Detailed Cash Flows for Option C (Multi-Line OCR Without ZIP+4) at High Performance Rate 111 PRICE PER OCR NUMBER OF OCRs
PAGE 109
Table B-n 1 122 1 123 Detailed Cash Flows for Option C (Multi-Line OCR Withou t ZIP+4) at Median Performance Rate 121 . . . . . . . . . . .*.*..*. ...**..., . . l . . *.,.*.... . . ..,....** l .0...... . . .*..:*... ..*.*.,.. . . l .**.*** .000..0.. ...**.*,. . .....*,. . . . . . . . ...,*... . . . . . . . . . . . a8s=s--.----===z=s==g==:====:====:=:8===-----.==========8========8====x====================--------========================================= =:====================---------.-----..----------.---SOURCE: Office of Technology Assessment
PAGE 110
e I Table Detailed Cash Flows for Option C at Low Performance 131 RATE REDUCTION o 0 0 0 B-12 (Multi-Line OCR Without ZIP+4) Rate 5 6 7 8 9 10 11 12 l a 0 0 0 0 0 a 0 0 0 0 :: ::::.*...........*. . . . . . . . . . . . . . . . ......*..**...,..*.. ..*......*,...*.*... ,....,O.OOO,.O.OOOO .**...*.* . . .*,...,... . . . . . . . . . .*...*.**. l *****.*** l ***.*.*** . . . . . . . . . SOURCE: Office of Technology Assessment
PAGE 111
Table B-13 Detailed Cash Flows for Option D (Automatic Conversion) at High Performance Rate and High Savings Rate 1111 1.0742 o 1 2 3 1985 1986 1987 1988 7 1992 SOURCE: Office of Technology Assessment
PAGE 112
Table B-13 (continued) 1211 1311 SOURCE: Office of Technology Assessment
PAGE 113
I Table d-14 I I I I I I I Detailed Cash Flows for-Option D (Automatic Conversion) at High Performance Rate and Median or Low Savings Rate =*x==E=s===xss=sxx8sx8*s*88s888s8xzxssz=88xsz=c=szs=::==z:==z===xxxcs= =:=x:======sz=ss:=:s=2:=:==z.-.-...==s=axs=zs8sz====8===*===~~======== ==**===============~=*~=*===:= ------SOURCE: Office of Technology Assessment I
PAGE 114
Table B-15 Detailed Cash Flows for Option D (Automatic Conversion) at Median Performance and High Savings Rate 1121 SOURCE: Office of Technology Assessment
PAGE 115
Table B-15 (continued) 1221 I 1321 X=X8 ==s== =8== == s ::= ::= :=== ====== ===== == == ===== = == === ==== =* = == ==== ===== ======== = === = == = === === = == = == = = == == == == == = ===== = === = === ============= =6 = = = l = =s ===== ==6= == == = === = 8 ===s= SOURCE : Office of Technology Assessment
PAGE 116
Table B-16 Detailed Cash Flows for Option D (Automatic Conversion) at Median Performance and Median or Low Savings Rate x:==::=:=====:::=::::::=:s=:=2s::=xs=xzz*x==:===z=:x=xc=s-------------------------------------------.-----.--.--------.-------=:===:==========:=:===:======================== ==============:===-.----==:=== -----DATE 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 ------------------------------------------1122 ------------------------------------------------CF-PH I, SAVINGS @ 90%, HI USE (193,295) (111,967) 81,615 365,228 454,038 ------464,631 622,585 679,169 739,951 805,243 875,380 950,722 1,031,653 1,118,590 REDUCED ROI 70.0% RED NPV 1,993,421 1222 CF-PH I, SAVINGS @ 90%, MED USE (239,525) (190,008) (66,252) 134,708 292,87 2 355,784 602,601 661,50 2 725,85 9 794,99 2 871,537 950,722 1,031,653 1,118,590 REDUCED ROI 45.1% RED NPV 1,517,346 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... . . l .,0.,.... l .*....... . . . . . l . . . . . . . .,....,0.. ......0.,0 . . .**,.,.*., . . . . . SOURCE: Office of Technology Assessment 8 )
PAGE 117
. Table B-17 1 I I I I I I I I I I Detailed Cash Flows for Option D (Automatic Conversion) at Low Performance and High Savings Rate 1131 8=98s8*8*=88=8=8=m8D8s88===8==8888=mm=c==8x=B=*s88888888=======8==xcxx ===============================8======*=====*==========~=~===~*======= l =================eB=n====c8a c 1.0742 SOURCE: Office of Technology Assessment I I
PAGE 118
Table B-17 (continued) 1231 1331 ========================:*=======:x==zsz====--------..-----.----.-z==:s-------.------...-.------..--======xz==:==zxs::::====x=:=:=========:: X======Z --------------------------------------------------.-----Assessment
PAGE 119
Table B-18 I 1132 I 1232 I 1233 I 1333 Detailed Cash Flows for Alternative D (Automatic Conversion) at Low Performance Rate and Median or Low Savings Rate SOURCE: Office of Technology Assessment
PAGE 120
. Table B-19 Cash Flows for Option F: Cancel DATE PHASE I ONLY CASH FLOW 1985 1986 1987 1988 --------- ------- ----------- --54,469 34,884 33,426 35,717 NPV 232,199 1989 1990 1991 1992 1993 1994 1995 -------- ---------------------- ------ --------------31,126 33,436 35,916 38,581 41,444 44,519 47,823 1996 1997 1998 ----.-.0 ---------------51,371 55,183 59,277 SOURCE : Office of Technology Assessment 97
PAGE 121
SELECTED REFERENCES Michael T.Y. Lai and Ching Y. Suen, Automatic Recognition of Characters by Fourier Descriptors and Boundary Line Encodings, Pattern Recognition, N OS 1-6, pp. 383-392, 1981. Recognition Equipment, Inc. Unsolicited Proposal for Build, [installation and Test of RCSOCR vol. 14, Systems" f submitted to United States Postal Service, undated (late 1982). Letter from REI President William G. Moore, Jr. to Senior Assistant Postmaster General James V. Jellison, dated July 11, 1983. Letter from William G. Moore, Jr., to Postmaster General William Bolger, dated August 24, 1983. Letter from REI President William G. Moore, Jr., to Postmaster General William Bolger, dated September 9, 1983. Herbert F. Schantz, The Potential and Economics of Automated Data-Entry (Recognology) A High Technology in The Information Age, Recognition Technologies Today, December 1983, pp. 5-7. Ching Y. Suen, et. al., f Automatic Recognition of Handprinted Characters The State o f the Art, Proceedings of th e IEEE, vol. 68, No. 4, April 1980, pp. 469-487. U.S. Congress, Committee on Government Operations, Subcommittee on Government Information, Justice, and Agriculture, Postal Service Electronic,Mail: The Price Still Isnt Right, House Report No. 98-552, 98th Congress, 1st Session, November 16, 1983. U.S. Congress, Office of Technology Assessment, Implications of Electronic Mail and Message Systems for the U.S. Postal Service, October 1982. U.S. General Accounting Office Conversion to Automated Mail Processing Should Continue; Nine-Digit Zip Code Should be Adopted If Conditions Are Met, Report No. GAO/GGD-83-24, General Accounting Office, January 6, 1983 (GAO, 1983a). Background Paper o n Multi-Line Read Optical Character Reader, September 1983. Letter from General Government Division Director William J. Anderson to The Honorable William D. Ford, Chairman, House Post Office and Civil Service Committee, dated September 19, 1983. 98
PAGE 122
. Conversion to Automated Mail Processing and Nine-Digit Zipcode: A Status -~ RePort 0 G~o/GGD-83-84 enerd Accounting Office September 28, 1983 (GAO, 1983 b). Comparative Review of Single-Line Read and Multi-Line Read Optical Character Readers Used in Mail Processing, May 1984 draft. U.S. Postal Service Letter from Senior Assistant Postmaster General James V. Jellison to REI President William G. Moore, Jr., dated July 26, 1983. Letter and enclosures from Postmaster General William F. Bolger to The Honorable William D. Ford, Chairman, House Post Office and Civil Service Committee, dated July 27, 1983. Letter from Postmaster General William F. Bolger to REI President William G. Moore, dated September 2, 1983. Proposal to Board of Governors: Automated System Phase II, January 1984. Letters and Attachments from Senior Assistant Postmaster General James V. Jellison to OTA Project Director Fred B. Wood, dated January 23, 1984, and February 9, 1984. Letter from Senior Assistant Postmaster General James V. Jellison to OT A Project Director Fred B. Wood, dated March 14, 1984 (1984c). Letter from Lynn A. Kidd to Fred B. Wood, dated April 2, 1984. Letters from Senior Government Relations Representative Kenneth F. Currier to OTA Project Director Fred B. Wood, dated April 2, 1984, and May 11, 1984. Why a Four-Line OCR is Not the Way to Go, undated. 99
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ERFEL0NYI_T9WK67 INGEST_TIME 2017-05-18T17:31:03Z PACKAGE AA00055041_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
|