Summary. Summary of Technical Analysis

Transcription

1 Summary Summary f Technical Analysis Alternatives t ptical character recgnitin technlgy. As a starting pint, OTA examined pssible alternatives t ptical character recgnitin fr pstal autmatin technlgy. Optical character recgnitin technlgy reads printed alphanumeric characters (letters and numbers) and recdes these characters int machine-readable frms such as a bar cde. OTA identified several electrnic, magnetic, and mechanical alternatives t ptical character recgnitin. Hwever, OTA cncluded that, at least fr the U.S. mail, it is as yet difficult t imprve n the infrmatin-carrying ability, readability, and cst effectiveness f printed characters n paper. As lng as this is the case, then ptical character recgnitin technlgy is the technlgy f chice. Electrnic mail is likely t be the strngest cmpetitr f pstal autmatin. But there mst likely will be a significant residual vlume f paper mail at least thrugh the year Thus there is a windw f pprtunity fr further USPS use f paper-based autmatin technlgy. Alternatives t a 9-digit ZIP cde. OTA fund that there are alternative cdes. Hwever, OTA cncluded that at this juncture there is n realistic alternative. The 5- digit ZIP is almst universally accepted and used (98 percent usage); the 9-digit ZIP directry is nw cmpleted; and ZIP+4 cdes are being distributed t large business mailers. If ZIP+4 becmes widely used, USPS culd cnsider adding a tenth digit (fr errr checking purpses) at sme future time. (Nte: The USPS bar cde already includes a crrectin character. ) Only if ZIP+4 des nt becme widely used culd alternative cdes realistically be cnsidered. Perfrmance f single-line ptical character readers (OCRs). OTA reviewed available data n perfrmance f the single-line OCRs nw being installed by USPS. Single-line OCRs read nly the last line f an address -- usually cntaining the city, State, and 5- r 9-digit ZIP cde. OTA cncluded that, despite initial start-up prblems, the already installed OCRs nw essentially meet USPS perfrmance specificatins. Perfrmance f multi-line OCRs. Over the last few years, multi-line OCR technlgy has emerged frm the labratry and prttype stage t peratinal units. Multi-line machines read up t fur lines f the address. OTA cncluded that, as f May 1984, the prepnderance f evidence indicates that multi-line OCR perfrmance is essentially equivalent t single-line fr reading 9-digit ZIP mail, and that multi-line perfrmance is substantially better fr reading 5-digit ZIP 6

2 .. mail t the 9-digit level. * OTA identified ne U.S. firm (Recgnitin Equipment, Inc.) and tw freign firms (Telefunken f Germany, ELSAG f Italy) that have prven multiline OCRs. OTA als cncluded that USPS has prbably underestimated the ultimate peratinal perfrmance level f multi-line OCRs by 5 t 15 percent. USPS estimated that multi-line OCRs wuld prcess 60 percent f 5-digit mail t the 9-digit level. OTA believes that 65 percent is mre likely and 75 percent pssible. Feasibility f lcal and natinal directries. In rder t read, cde, and srt S-digit ZIP mail t the 9-digit level, multi-line OCRs require a cmputerized address directry against which the address infrmatin can be cmpared t ascertain the crrect 9-digit ZIP cde. They then apply the crrespnding barcde, and finally srt the letter. Until recently, the absence f a lcal r natinal directry was a limiting factr fr use f multi-line OCRs. Hwever, in the perid, USPS cmpleted a natinal ZIP+4 directry and lcal ZIP+4 directries fr majr metrplitan areas. USPS and OTA agree that the cnversin f existing lcal ZIP+4 directries t a multi-line OCR frmat is technically feasible. Whereas lcal directries clearly wuld be necessary fr multi-line OCR peratin, OTA was nt able t determine whether natinal directries wuld ffer any significant advantage, particularly when cmpared t the technical difficulties and likely additinal cst. Feasibility f single-line t multi-line cnversin. OTA reviewed the technical feasibility and cst f cnverting single-line OCRs t multi-line. OTA cncluded that cnversin wuld be technically feasible and the USPS estimate f cnversin cst -- $200,000 per machine -- is as gd as can be develped frm available infrmatin. The actual cst f single- t multi-line OCR upgrade can nly be determined by detailed engineering analysis and cmpetitive prcurement prcess. It impssible that the cnversin culd be accmplished by an OCR manufacturer ther than the riginal surce, althugh this might require a high degree f technical cperatin between the tw vendrs. Technical pprtunities fr imprved perfrmance. OTA identified several areas where technical perfrmance f pstal autmatin might be imprved in the future. These include bar-cded reply envelpes, mailer printing f bar cdes, imprvements in perfrmance f ptical character readers, standards fr address frmat, and increased research and develpment n pstal autmatin. * The full address with 5-digit ZIP is read and cmpared against a cmputerized address directry that includes 9-digit ZIP cdes. If a match is made between the address n the envelpe and an address in the directry, the apprpriate 9-digit cde is applied. 7

3 The USPS recrd n pstal autmatin R&D is mixed. USPS cntinues t underspend n R&D cmpared t the U.S. industry average, and pstal R&D rganizatin and management appear t have lacked stability, clear directin and, at times, tp level cmmitment. Despite 20 years f USPS investment in ptical character recgnitin R&D, when USPS slicited manufacturers in 1980 fr single-line OCRs, all U.S. manufacturers previusly receiving USPS supprt fr single-line OCR R&D had withdrawn frm the market, in part due t several years f USPS indecisin n an autmatin strategy. On the ther hand, USPS has established a gd track recrd in narrwly fcused R&D n imprvements t upgrade existing equipment, such as the multi-psitin letter srting machine. Als, despite sme variability in funding and cmmitment, USPS has prvided enugh supprt ver the last 14 years t Recgnitin Equipment, Inc. (REI) such that REI has develped ne f the leading multi-line OCRs n the wrld market. Summary f Decisin Analysis USPS faces a decisin pint as t whether t cntinue its cmmitment t ZIP+4 and single-line OCR technlgy r t mdify that cmmitment in sme way. OTA emplyed decisin analysis techniques t: identify the range f ptins available t USPS; develp a prbabilistic cash flw mdel f each ptin; assign prbability distributins fr key variables such as ZIP+4 usage and multi-line OCR perfrmance; and calculate the rate f return (ROI), net present value (NPV), ttal net cash flw, and annual net cash flw fr each ptin; and cnduct sensitivity tests f the results t changes in key variables. Descriptin f decisin ptins. Optin A: Phase II single-line OCR is the current USPS strategy t prceed t prcurement f the 403 additinal single-line OCRs advertised fr Phase 11 f the pstal autmatin prgram, and n which bids have already been received. Under ptin A, there wuld be n further USPS expenditure n multi-line OCR research, develpment, and testing. Optin B: Multi-line OCR with ZIP+4 is a decisin t cancel the current Phase 11 single-line OCR prcurement, initiate release-lan testing f multi-line OCRs, and as sn as pssible reissue the Phase [1 request fr prpsals but fr multi-line rather than single-line OCRs, meanwhile retaining the ZIP+4 cde. Single-line OCRs already purchased wuld be cnverted t multi-line capability. * OTA did nt analyze the ptin f prcuring 403 additinal Phase I single-line OCRs instead f Phase II OCRs. This ptin was judged t be nt significantly different frm ptin A. ** The release-lan testing manufacturers actually test prttype equipment n USPS premises with real mail. 8

4 Optin C: Multi-line OCR withut ZIP+4 is the same as ptin B except that the ZIP+4 cde wuld be terminated. The 5-digit ZIP cde wuld be retained. Optin D: Autmatic cnversin is t prceed with the Phase 11 single-line OCR prcurement, but simultaneusly initiate release-lan testing (and any necessary related R&D) n single-line t multi-line cnversin and then cnvert all single-line OCRs as sn as pssible, regardless f the level f ZIP+4 use. Optin E: Hedge cnversin is similar t ptin D except that the single- t multi-line cnversin wuld take place nly if ZIP+4 use is lw at a specified future time (defined here as year-end 1987). Bth ptins D and E include the same initial decisin t purchase Phase II single-line OCRs, and t initiate release-lan testing f and any necessary research n cnversin. The difference is that under ptin D, the cnversin wuld be made regardless f the level f ZIP+4 use, while under ptin E, cnversin wuld take place nly if use is lw. Optin F: Cancel Phase II and ZIP+4 is t cancel the Phase II single-line OCR prcurement, terminate ZIP+4, and use the single-line OCRs already purchased t prcess 5-digit ZIP mail. Optin G: Split prcurement is a hybrid ptin that wuld cancel the Phase II prcurement, immediately reissue an RFP fr ne-half the number f single-line OCRs (202 instead f 403), and simultaneusly initiate release-lan testing f the multi-line OCR and single- t multi-line cnversin. A new RFP fr prcurement f the ther half f the OCRs but using multi-line technlgy (201 multi-line OCRs) wuld be issued as sn as pssible, prbably in abut 2 years, at which time the then existing single-line OCRs (252 frm Phase I and 202 frm Phase II) wuld be cnverted t multi-line. Optin H: Split Prcurement is similar t ptin G except that the Phase 11 RFP wuld be reissued fr 90 percent f the single-line OCRs (363), rather than 50 percent, and release-lan testing wuld be initiated n multiline OCRs leading t a new RFP fr prcurement f the ther 10 percent f the OCRs (40) using multi-line technlgy. Key assumptins. Where pssible and justifiable, OTA used the same assumptins as did USPS. Fr example, OTA and USPS used the same time hrizn (14 years, ), labr cst escalatin rate (7.42 percent annually), baseline cst and savings prjectins (fr single-line OCRs, as presented in January 1984 t the Bard f Gvernrs), discunt rate (15 percent per year), and single-line OCR perfrmance and cst. OTA assumptins abut multi-line OCR cst ($850,000 per machine), single- t multi-line cnversin cst ($200! 000 per machine), and the time required t release-lan test and prcure multi-line OCRs and cnversin kits (3 years) were generally cnsistent with USPS and GAO estimates.

5 The majr differences between USPS and OTA were assumptins abut the incentive rates fr ZIP+4 usage (OTA and GAO treated these as a cst, since incentives appear t be required t get large mailers t use ZIP+4), multi-line OCR perfrmance rates (OTA and GAO cncluded that perfrmance wuld likely be smewhat better than USPS estimated), ZIP+4 usage, and clerk/carrier savings rate. Fr the latter three variables, OTA develped lw, median, and high estimates. Fr lw, median, and high estimates f multi-line OCR perfrmance. OTA cncluded that prductin mdel multi-line OCRs wuld read 60, 65. and 75 percent f 5-digit mail t the 9-digit level, respectively. Fr ZIP+4 usage, USPS prjects that usage wuld reach 90 percent f the machinable metered first class mail base within 5 years. This is cnsiderably mre ptimistic than actual experience with either the U.S. 5-digit ZIP cde r the Canadian 6-digit pstal cde. The 5-digit ZIP tk 12 years t reach 90 percent usage; after 5 years, the 5-digit usage level was abut 51 percent. Thus, the USPS prjectin shws ZIP+4 reaching 90 percent abut tw and ne-half times as fast as the 5-digit ZIP (in 5 years rather than 12). At present, the prjected actual vlume f ZIP+4 mail fr 1984 is abut 20 percent f the riginal USPS prjectin. After reviewing all available evidence, OTA cncluded that the USPS ZIP+4 prjectin shuld be cnsidered ptimistic (high), that an apprpriate median estimate wuld be the 5-digit ZIP grwth pattern, and that an apprpriate pessimistic (lw) estimate wuld be a grwth pattern peaking ut at abut 40 percent ZIP+4 usage. At the present time, the first year ZIP+4 usage culd turn ut t be even mre pessimistic. The estimated 2.73 billin pieces f ZIP+4 first class mail at year end 1984 represents abut 5.4 percent f the target mail base as cmpared t abut 7 percent under the lw scenari, 13 percent under the median, and 28 percent under the ptimistic scenari. (GAO made n estimate f ZIP+4 usage but did find that businesses are still cncerned abut the cst f cnverting and whether the USPS presrt discunt will be mdified because f ZIP+4.) Fr clerk/carrier labr savings, OTA cncluded that the USPS baseline estimate was prbably smewhat ptimistic, since the quality and mix f the OCR prcessed mail may be less than anticipated, and labr csts (e.g., fr maintenance) may be mre than expected. OTA assumed high, median, and lw labr savings rates f 100 percent, 90 percent, and 80 percent f the USPS estimate. Results f decisin analysis. With respect t internal rates f return (ROIs), every ptin except ptin F (cancel), under all cnditins mdeled, shws an ROI abve the 15 percent threshld established by USPS. OTA assumed, therefre, that under any scenari, the Phase I single-line OCRs already purchased wuld be kept in service. All ROIs, net present values, and cash flws were calculated net f cash flws assciated with the Phase I single-line OCRs. Use f ROIs fr decisin making has a serius limitatin. When mre than ne ptin clears the hurdle rate (that is, has mre than the minimum required ROI, in this case 15 percent), the ROI itself gives n indicatin f the cash flw differences f the varius ptins as a basis fr cmparing the ptins. An alternative t ROI frequently used in capital investment decisin making is net present value (NPV). NPV discunts the cash flws f each ptin at the hurdle r threshld rate, in this study 15 percent. 10

6 Under cnditins f high savings and high multi-line perfrmance, ptin D (autmatic cnversin) has abut a 5 percent and 11 percent higher NPV with high and median ZIP+ 4 usage, respectively, than ptin A (single-line OCR). At lw ZIP+4 usage, all ther things being equal, the ptin D advantage increases t a substantial 134 percent r abut $820 millin in NPV. At a lw savings rate (alng with lw ZIP+4 usage and high multi-line perfrmance), the relative advantage f ptin D ver A increases further t abut 310 percent althugh the abslute advantage decreases t abut $650 millin in NPV. Even at lw multi-line perfrmance, ptin D has 53 t 119 percent relative advantage in NPV and a $320 t $250 millin abslute advantage in NPV, at a high and lw savings rate, respectively. Optin E (hedge cnversin) has the same NPV as ptin A at high r median ZIP+4 usage and the same NPV as ptin D at lw ZIP+4 usage. Optin H (90-10 split prcurement) als has a higher NPV than ptin A under almst all cnditins. Optin G (50-50 split prcurement) has a significant althugh smewhat smaller advantage ver ptin A at lw ZIP+4 usage. Optin G has a 34 t 271 percent relative advantage in NPV and a $170 t $710 millin abslute advantage in NPV at lw ZIP+4 usage, depending n the multi-line OCR perfrmance rate and savings rate. The ranking f the ptins by NPV is summarized belw: Overall NPV Rank Lw ZIP+4 Use NPV Rank Optin D 1 highest Optin 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 lwest A 7 lwest OTA fund that the NPV results are nt very sensitive t the purchase price f the multi-line OCR r the number f multi-line OCR units. An increase in the purchase price frm $850,000 t $970,000 r an increase in the number f units frm 403 t 444 (as estimated by GAO t be required if the entire Phase [1 prcurement was switched frm single- t multi-line OCRs) wuld reduce NPV by abut $20 t $30 millin. Net present value appears t be the best basis fr cmparative quantitative evaluatin f the decisin ptins. Hwever, the actual undiscunted net cash flws ver the 13 year payback perid ( ) can prvide anther dimensin t the evaluatin. Optin A (single-line) is estimated t shw psitive cash flws f $8.8, $8.24, and $3.57 billin at high, medium, and lw ZIP+4 usage. At high ZIP+4 usage, ptin B (multi-line with ZIP+4) is smewhat lwer at $8.14 billin, ptins D (autmatic cnversin) and H (90-10 split prcurement) smewhat higher at $9.36 billin and $9.24 billin respectively, and ptin G (50-50 split prcurement) abut the same at $8.75 billin. The cmparisns between ptins change relatively little at median ZIP+4 usage. Hwever, at lw ZIP+4 usage there is a substantial difference in net cash flws. Optin A(single-line) shws a net cash flw f $3.57 billin. But, depending n the multi-line OCR perfrmance rate, ptins D (autmatic cnversin) and H (90-10 split 11

7 .., prcurement) shw a net cash flw f $5 t $7.2 billin, r abut $1.4 t $3.6 billin greater than ptin A. Optin G (50-50 split prcurement) shws abut $1.1 t $3.3 billin greater cash flw than ptin A, and ptin B (multi-line with ZIP+4) shws abut $0.8 t $3.0 billin greater cash flw than ptin A. A cmparisn f yearly cash flws gives similar results. By 1994, all equipment will presumably have been installed (r cnverted) and up and running at ptimal perfrmance. Optins B, D, G. and H will by that time lk exactly the same -- ail multi-line OCRs. The single-line OCRs prcured under ptins D, G, and H will have been cnverted t multi-line capability. Optin A will cntinue t be slely single-line OCRs. With high ZIP+4 usage, ptin A shws an annual net cash flw f abut $870 millin t $1.2 billin frm 1994 t Optins B, D, G, and H shw almst identical annual cash flws, nly slightly higher by abut $70 t $100 millin per year. Hwever, at lw ZIP+4 usage, the differences again becme substantial. With high multi-line perfrmance, ptins B, D, G, and H shw between $440 and $580 millin per year additinal net cash flw cmpared t ptin A, frm 1984 t With median multiline perfrmance, the advantage f ptins B, D, G, and H ver A ranges frm $370 t $490 millin per year. And even at lw multi-line perfrmance, the advantage ver ptin A, while reduced, is still significant at $180 t $240 millin per year. 12