Click for next page ( 51


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 50
50 APPENDIX A Chicago Transit Authority Case Study The CTA is the nation's second largest transit property, as well as developing applications for data analysis. The sys- serving 1.5 million weekday passengers with a fleet of 2,220 tems have been producing very large data volumes (3.2 mil- buses and 1,190 rail cars. Preliminary analysis pointed to the lion data records per day from the AVL and APC systems following reasons for selecting the CTA as a case study for this alone). A reorganization has shifted several staff to the IT sec- Guidebook: tion within the Management and Performance division to take responsibility for database management, while several The CTA maintains a comprehensive market research analysts remained within Data Services. program that is increasingly drawing on ITS data to monitor service delivery and to leverage traditional customer research Experience With Smart Cards practices. and Magnetic Stripe Cards The CTA is an industry leader in analyzing customer data from smart cards and magnetic stripe cards in support of Smart cards were introduced in 2002 and are offered in two market research and planning. versions. The Chicago Card (CC) is a stored value device with Although the CTA's experience with AVL and APC tech- optional registration. Card balances can be recharged at ticket nology has been fairly brief (less than 3 years), it has vending machines. The Chicago Card Plus (CCP) is managed developed an innovative array of intranet applications for through a Web-based account that is recharged by credit monitoring service delivery and improving market research card. All CCPs are registered to users. Registration informa- and planning. tion includes a billing address, an optional email address, and an optional field for customers to opt in for market research The research team met with CTA staff in March 2007. studies. The number of customers responding positively to the market research option has been limited. No demo- graphic information is collected in the registration process. Organizational Structure The demographics of smart and magnetic stripe card users CTA's market research function resides within the Planning are recovered from customer satisfaction surveys. The most section of the Transit Operations division (see Figure A-1). recent count of CC and CCP cards in circulation is 530,000, The Planning section includes Strategic Planning (which of which 250,000 are registered with email addresses. houses market research), Service Planning, Scheduling, Facil- Magnetic stripe cards were introduced in 1997. Following ities and Traffic Engineering, and Data Services. The creation their introduction all flash passes and tokens were eliminated. of the Data Services unit within the Planning section was a There are a number of alternative purchase options. Single consequence of the deployment of EFP, AVL and APC tech- ride cards are sold in packs of 10 and 20. Unlimited ride nologies, reflecting the interest of Planning section units in passes are offered for 1, 7, and 30 day's duration, as are accessing and analyzing data for their purposes. Previously, unlimited ride "visitor" passes covering 1 to 5 days. The un- for example, EFP data served the Finance section, and Plan- limited ride cards are activated on first use. Lastly, cards are ning units had to cross divisions and multiple sections to gain sold in stored value amounts ranging from $2.25 to $20.00 access to data. (rechargeable through ticket vending machines). Transac- Until recently, Data Services staff have been responsible for tions on the CTA rail system are limited to cards, while cash post-processing and merging data from the various systems, remains a payment option on the bus system.

OCR for page 50
51 Chicago Transit Board General Counsel Office of Inspector General Treasurer President Security & System Safety Management Construction Engineering Transit Operations & Performance & Facilities Maintenance Communications & Marketing Bus Operations Construction Employee Relations Control Center Engineering Finance Facilities Maintenance Planning Government and Community Relations & Affirmative Action Rail Operations Human Resources Service & Property & Real Estate Reliability Operations Asset Management Training & Purchasing/Warehousing Customer Service Technology Management Figure A-1. CTA organizational structure. About 50% of the trips on the CTA system involve a trans- ridership impacts of alternative pricing schemes. The CTA's fer. In January 2006, free transfers were eliminated for cash fare change model, first developed in the 1970s, has been up- fare riders, and the share of cash transactions subsequently dated to accommodate the evolution of fare payment options fell from about 30% to under 10%. In mid-2006, stored value (Multisystems and NuStats International 2000). The fare MS cards accounted for the largest share of transactions change model is structured around passengers' fare media (about 25%), followed by 7-day MS cards (about 22%). CC choices through stated preference exercises. Derived demand and CCP transactions accounted for about 18%. Over time, elasticities, in turn, are applied in estimating revenue and rid- transactions have become much more differentiated among ership. Calibration of the model has been periodically fare media. In January 1999, before the introduction of smart updated through stated preference surveys, and fare change cards, cash and stored value MS cards accounted for nearly model estimates can now be evaluated against the actual 80% of transactions. By July 2006, they accounted for just behavior of customers represented in fare media usage data. over 30%. The increasing differentiation in fare media has The fare change model was put to the test in assessing a recent been reinforced by focus group and survey analysis showing substantial fare increase. It predicted a large growth in very strong customer preferences for the fare payment op- revenue with limited impacts on ridership, which was subse- tions they select. quently borne out. Increasing diversity of fare payment options, coupled with Fare card transactions are recorded at the entry station on the pricing flexibility inherently available with card technol- the rail system. CTA analysts have been able to infer exit lo- ogy, has led to greater interest in exploring the revenue and cations from the sequence of card transactions that occur

OCR for page 50
52 Figure A-2. Derived CTA exit locations from entry location sequence data. over the course of a given day (see Figure A-2). This station, indicating that billing and residential addresses may approach assumes that passengers return to the destination differ or that other (non-walk) travel modes may be used to station of their previous trips and that at the end of the day access stations. they exit the same station they first entered. Using this The CTA has used CCP card data in preparing for major approach, Rahbee and Czerwinski (2002) were able to suc- reconstruction along the Brown Line north of the Loop, cessfully infer 70% of passengers' destinations from se- which is among the system's most heavily-traveled corridors. quential station entry data. Using scaling factors to account Information on station closures is being targeted to riders for station-specific variations in unsuccessful entry-exit whose transaction information shows them to be users of the matches, they estimated an origin-destination trip table for affected stations (see Figure A-4). GIS has also been used to the rail system. create a buffer along the Brown Line corridor to examine cen- For many rail origin-destination pairs the route path is de- sus data on languages spoken in the home, so that informa- terminant. Passengers boarding at stations in the Loop, how- tion on construction and travel options can be more effec- ever, have directional options. Given that most Loop stations tively communicated. serve either direction, path choice cannot be directly deter- Travel disruptions from reconstruction along the Brown mined from card transaction data. A few Loop stations are di- Line may be mitigated by the fact that many CTA customers rection-specific, allowing analysis of path choice. Selection of have the option of traveling on bus routes that also serve the the shortest path may minimize travel time, but transactions corridor. Analysis of CCP data revealed a pattern of rail and data from directional stations indicates that some passengers bus preferences in the corridor (see Figure A-5). Estimates of opt for a lengthier alternative path possibly because the expected changes in rail preference related to the Brown Line choice improves their likelihood of finding an available seat. reconstruction should prove helpful in planning for addi- Analysis of the travel time differences involved in such tional bus service, as well as in targeting information to cus- choices, along with passenger surveys, can yield estimates of tomers on travel options that will be available during the the implicit monetary value that customers place on im- project. provements in comfort and convenience associated their trip making (Zhao 2004). On the bus system, card transaction data include the time of transaction, the vehicle's farebox ID, and the route ID. The location of the transaction is not directly identified. However, by joining the card and AVL data streams, the stop associated with the transaction can be inferred from the location status of the bus at the time of the transaction (Cui 2006). CCP card holders' station or stop accessibility can be ana- lyzed by relating billing addresses to entry station transactions or inferred stop transactions using a GIS (see Figure A-3). Analysis of CCP station transaction data found that 40% of card holders "resided" more than two miles from their entry Figure A-3. Station accessibility.

OCR for page 50
53 Brown Line Station Preference, Chicago Card Plus Customers Figure A-4. Station usage of Chicago Card Plus holders along the CTA Brown Line. CTA's Experience With AVL post-processing. Data archiving is managed in an Oracle and APC Technology database and processed twice daily. Archived data can be ac- cessed from desktop workstations through an intranet web AVL and APC technologies are integrated in the CTA's au- server, where reporting and analysis tools reside. tomated voice annunciation system (AVAS), which was de- AVAS summary reports and analysis tools available on ployed on the bus fleet in 2002 following a court order related CTA's intranet web server are shown in Figure A-7. Summary to the Americans with Disabilities Act. All buses in the fleet reports of passenger movements from APC data are gener- are AVL-equipped, and about 45% (35% of standard buses ated by Ridecheck Plus reporting software. The structure of and 100% of articulated buses) of buses have APC units. APC the summary reports differentiates by route, time period and units are being specified in all new bus acquisitions, with a direction. Load and maximum load reports can also be gen- goal of 100% coverage by 2013. erated for selected routes and locations. AVL data are The general organization of ITS data flows in AVAS is il- processed to report vehicle running time information with lustrated in Figure A-6. Data that are uploaded to buses from respect to schedule adherence and on-time performance the BusTools workstation include data from the HASTUS from the stop to the system level for selected time periods. Scheduling System and timepoint/stop location coordinates. Reports on the operating status of AVAS equipment and the These data allow assignment of APC passenger movement screening of data recovered are also available for maintenance data to unique locations (stops), and also allow AVL actual personnel. time-at-location data to be related to the corresponding Consistent and dependable service is highly valued by scheduled time-at-location data, facilitating analysis of transit customers (Morpace International and Cambridge schedule and headway adherence. The recovered AVL and Systematics 1999). From a service delivery perspective, con- APC data are downloaded once daily through a wireless link sistency and dependability cannot be delivered when sched- to a server located at each garage, and then forwarded for ules are incompatible with actual operating conditions. The

OCR for page 50
54 Mode Preference, Chicago Card Plus Customers Figure A-5. CTA rail and bus mode preferences. Figure A-6. Data flows in AVAS.

OCR for page 50
Figure A-7. AVAS intranet reporting and analysis tools.

OCR for page 50
56 development of schedules that provide consistent and de- times substantially exceed optimal running times during the pendable service relies on detailed knowledge of the pattern Midday and PM Late periods. of vehicle running times on a route. The AVAS recovers a Although available systemwide, the running time tool has large volume of running time data that can be used to in- not yet been applied to all routes due to time demands on form the HASTUS scheduling system. A new analysis tool scheduling staff. There is an expectation that a systemwide on the intranet web site facilitates analysis of running time evaluation would identify more instances where running patterns for routes and route segments. The tool provides time could be reduced than where additional running time plots of actual running times in relation to scheduled run- would be needed. This expectation reflects traditional sched- ning times for visual examination. A related tool provides uling practices that respond to operator and customer feed- estimates of optimal running and recovery times based on back that tends to focus on circumstances where running the observed patterns. time is inadequate, while being less attuned to circumstances Figure A-8 illustrates the end-to-end pattern of actual run- where running time is excessive. If so, there would be an ning times in relation to scheduled running times for one of opportunity to improve both schedule reliability and sched- the CTA bus routes. Although the schedules provide greater ule efficiency. running times during the morning and evening peak periods, it is apparent from the plot that the actual running times of Reconsidering Performance buses serving the route are notably greater than the scheduled Measures times, except during the 6:00-8:00 pm period. The variation in actual running times also appears to be much greater be- As the CTA board and senior management have become tween 2:30 and 4:30 pm than at other times. more familiar with applications of ITS data in the Transit A related tool translates the pattern of observed running Operations division, they have begun to encourage develop- times into optimal scheduled running and recovery times (see ment of new performance measures. The system level Figure A-9). This calculation is based on the objective of pro- performance metrics that have been tracked over time by CTA viding running time that is sufficient to accommodate 65% include ridership, on-time performance (bus), delay (rail), of scheduled trips, and adding recovery time that is sufficient vehicle and station cleanliness, safety, customer complaints/ to accommodate 95% of trips serving a route. In the example commendations, and affordability. shown in Figure A-9, the optimal and scheduled running Surveys indicate that reliable service is a high priority among times are fairly close for northbound trips during the Owl, customers. While on-time performance is a reliability indicator, Early AM, and AM Peak periods, while the scheduled running it is more relevant in lower frequency service environments, 160 140 120 Trip Travel Time (minutes) 100 80 60 40 20 0 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Hour of Day (24 hour time) AVAS Trip Travel Time Observation Scheduled Travel Time Figure A-8. Actual versus scheduled running times.

OCR for page 50
57 Figure A-9. Optimal versus scheduled running time. where customers plan their travel to coordinate with schedules. of bus bunching, as shown in Figure A-11. In this example, In high frequency service environments, customers tend to dis- the incidence of bunching is clearly greater during peak peri- regard schedules and pay more attention to the consistency of ods, when passenger volumes are larger and traffic congestion service. Nearly 90% of CTA customers access the transit system poses the greatest challenge to maintaining consistent transit when headways are 12 minutes or less, suggesting that an indi- operations. cator based on headway consistency would serve as a better in- dicator of reliability than one based on schedule adherence. Other ITS-Related One headway-based reliability indicator under considera- Market Research Activities tion at CTA is related to the wait assessment measure devel- oped by New York City Transit (NYCT). This indicator The CTA is planning to execute an on-board O-D survey measures the percentage of customers whose actual wait is in spring 2007. This will be the first systemwide O-D survey less than the scheduled headway plus three minutes (during undertaken since the 1970s. Fare card and APC data will be peak periods) or five minutes (during off-peak periods). Un- used in designing the sampling plan for the survey and will like the NYCT measure, which relies on manually collected provide the expansion factors for the survey results. The dis- data from selected high-volume locations, the CTA wait tribution of the surveys will also include an advanced tech- assessment measure can be calculated systemwide from nology feature. Surveyors will use special pens that will record archived AVL data. A prototype of CTA's wait assessment in- the serial number of each survey and will be linked to bus dicator is presented in Figure A-10. AVL data via a time stamp, permitting geocoding of the loca- A related headway consistency indicator developed at CTA tion where it is distributed. relates to "bus bunching," which occurs when the actual The 2007 O-D survey will provide the first comprehensive headway separating buses becomes very small. Bunching is a basis for validating the rail and bus system trip tables that had source of irritation to customers, who usually face longer been previously developed from fare card transactions and waits only to find the lead bus of a bunched pair overloaded. APC data. Of particular interest to CTA staff is the validation It is also a waste of effective capacity for the service provider, of path choices in the rail system and transfers within the bus because the trailing bus usually carries a much smaller pas- system. While O-D surveys traditionally provide only a snap- senger load. AVL data can be used to document the incidence shot of activity in a transit system at given point in time, the

OCR for page 50
58 80% 79% 78% Average Percent Acceptable Wait 77% 76% 75% 74% 73% 72% 71% 70% First Half Second Half 2005 2006 Figure A-10. Wait assessment indicator for CTA bus service. DRAFT DRAFT Weekday July 2006 Bus Bunching Analysis does not include terminal timepoints, but all other timepoints on all (scheduled) routes are included. 7.0% 6.0% % of Headway Records Considered "Bunched" 5.0% 4.0% 3.0% 2.0% 1.0% 0.0% 00:00 01:00 02:00 03:30 04:30 05:30 06:30 07:30 08:30 09:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 18:30 19:30 20:30 21:30 22:30 23:30 Mike Haynes 12-11-2006 % :15 seconds or less % :30 seconds or less % One Min or less Figure A-11. Incidence of bus bunching by time of day.

OCR for page 50
59 2007 survey is viewed by CTA staff as providing a "ground Issues, Observations, truth" starting point that will be regularly updated through and Challenges time using current transaction and passenger movement data. There is an evident transformation underway in the mar- In another market research application, CTA staff used ket research and planning activities at CTA, wherein the use AVL and APC data to evaluate the effects of providing real of ITS data is playing a more central role in the practices em- time arrival information (on the Internet and at one bus shel- ployed by staff. This was clearly demonstrated in the inter- ter) on the #20 Madison bus route. APC data were used to se- views conducted during our site visit, and in the subsequent lect the shelter where arrival information would be provided review of materials and publications provided to us by CTA and a second shelter with similar passenger volumes, which staff. When we queried staff on the transformation they em- served as a control. Riders were surveyed on perceptions of phasized that it has not been seamless and that many chal- wait times before and after implementation of posted arrival lenges (financial, technical, entrenched practices) remain. time service. Surveyed riders at the control stop perceived no Nevertheless, it was generally acknowledged by those who we change, while those surveyed at the stop where arrival times interviewed that they were involved in developing innovative were posted perceived a 27% improvement in wait times. The applications of ITS data, and that they were "ahead of the corresponding AVL data indicated that the control stop riders curve" in relation to most other transit properties. Staff noted were "right": wait times did not change. Similar findings have that the success that they have experienced is not a direct con- been obtained in London (Nelson 1995, Schweiger 2003), sug- sequence of the size of their property, given that market re- gesting that wait time uncertainty rather than actual wait time search and planning staff numbers at CTA are roughly com- is what customers find more onerous, especially when service parable to staff numbers at properties serving half their is fairly frequent, and that the main benefit from posting ar- customer base. We asked "what specific things can you point rival times is in reducing uncertainty. to" that have made a difference in achieving more effective Evaluation of fare card transactions data has provided in- use of ITS data? This question drew the following responses. sights on travel activity related to special events and user First, the CTA has engaged in a formal arrangement with the groups. Convention passes were tested at events such as the transportation and planning programs at the Massachusetts Gay Games and the Rail~Volution conference. Data from Institute of Technology and the University of Illinois at UPasses (issued to college students) are used in revenue esti- Chicago to employ masters program students as interns. The mation, building new markets and planning for service dis- interns spend two summers in residence at CTA, familiarizing ruptions. Student travel also differs seasonally and by time of themselves with practices and identifying a thesis problem day, and having card transactions data helps in planning ser- during the first summer, and completing their thesis research vice. Surveys also found that college student riders were per- during the second. The research undertaken by interns is ceived by others to enhance safety (especially at night) and viewed as the development of prototype applications of ITS that they tended to have a calming influence on high school data that can eventually evolve into adopted practices. The riders. internship program has also served as a career pathway be- The CTA is currently in the process of validating its APC tween the participating institutions and the CTA. Two of the data against data on actual passenger movements from on- six staff that we interviewed identified themselves as veterans board cameras. This is the first step in transitioning to a new of the internship program. sampling plan using automated data for NTD reporting. Second, the U.S. DOT/FTA-sponsored Peer-to-Peer Devel- Alternatives evaluation in connection with development opment Program was mentioned as a useful means of sharing of the Pink Line rail project was cited by staff as an example experiences and practices related to the development of tools of the "new approach" to using ITS data for market research and applications that draw on ITS data. For the CTA, the ex- at the CTA. The Pink Line was created by splitting off and change involved staff from TriMet (Portland) and King realigning a trunk of the Blue Line. The alternatives analysis County Metro (Seattle). One reason the Peer-to-Peer Program first drew on rider survey data to validate O-D flows inferred was considered beneficial is that the development of new ap- from card transaction data. A network model was then used plications was evolving so rapidly that it was not being well to assign trips and estimate travel time savings associated documented in the professional literature (Gross et al. 2003). with the alternatives. Further analysis examined impacts on Third, one of the challenges to effective use of ITS data is negatively affected riders by origin and destination. The al- the gap separating development and use of new applications. ternative selected was shown to produce an overall estimated This is being addressed at the CTA through a series of internal 2.3% reduction in travel times relative to the previous align- "what's new" brown bag seminars. The seminars have proved ment, with an improvement in travel time for 98% of af- to be effective in disseminating information about new appli- fected riders. cations, and they allow market research and planning staff to

OCR for page 50
60 communicate their needs to those involved in the develop- Program, ITS Joint Program Office, U.S. Department of Trans- ment of the applications. The seminars also stimulate ques- portation, Washington, D.C., 2003. Morpace International and Cambridge Systematics. TCRP Report 47: tions from users and encourage exploration of new uses for A Handbook for Measuring Customer Satisfaction and Service Qual- the data more effectively than can be achieved through pro- ity. TRB, National Research Council, Washington, D.C., 1999. viding system documentation (which many cannot find the Multisystems with NuStats International. Fare Structure Pricing time to read). Research and Update of Ridership/Revenue Fares Model. Final Fourth, staff turnover represents another challenge to de- Report. Chicago Transit Authority, Chicago, IL, 2000. veloping and implementing new ITS data applications. Nelson, J.D. The Potential for Real-Time Passenger Information as Part of an Integrated Bus-Control/Information System. Journal of There are a limited number of people in the transit industry Advanced Transportation, Vol. 29, 1995, pp. 1325. who are both capable of working with ITS data archives and Rahbee, A., and Czerwinski, D. Using Entry-Only Automatic Fare Col- also have a deep understanding of market research, schedul- lection Data to Estimate Rail Transit Passenger Flows at CTA. Proc., ing, and planning practices. When staff with these abilities 2002 Transit Chicago Conference, Chicago, IL, 2002. leave they are not easily replaced, and the momentum of the Schweiger, C.L. TCRP Synthesis of Transit Practice 48: Real-Time Bus Ar- rival Information Systems. Transportation Research Board of the working relationship between the analyst and practitioner is National Academies, Washington, D.C., 2003. interrupted. Zhao, J. The Planning and Analysis Implications of Automated Data Col- Finally, the creation of the Data Services unit within the lection Systems: Rail Transit OD Matrix Inference and Path Choice planning section helped to bridge the gap between ITS data- Modeling Examples. MS thesis. Massachusetts Institute of Technol- base management and administration (which resides in the ogy, Cambridge, 2004. Management & Performance Division) and market research staff. Data applications involve complex queries and joining of CTA Staff Interviewed data tables, skills that are beyond the capabilities that are typi- Jeffrey Busby, Manager, Market Research cally found in market research and planning environments. Alex Cui, Project Coordinator, Data Services Elizabeth Donahue, GIS Analyst, Data Services References Mike Haynes, Project Manager, Transit Systems Support, Technology Management Cui, A., Bus Passenger Origin-Destination Matrix Estimation Using Aimee Lee, Manager, Resource Planning Automated Data Collection Systems. MS thesis. Massachusetts Kevin O'Malley, Manager Data Services Institute of Technology, Cambridge, 2006. Jeffrey Schroeder, Manager, Transit Systems Support, Technology Gross, P., Haynes, M., and Schroeder, J. APC/AVL and GIS Pacific North- Management west Knowledge Sharing. Final Report. Peer-to-Peer Development Jeffrey Sriver, General Manager, Strategic Planning