Leveraging ITS Data for Transit Market Research: A Practitioner's Guidebook (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

50 The CTA is the nation’s second largest transit property, serving 1.5 million weekday passengers with a fleet of 2,220 buses and 1,190 rail cars. Preliminary analysis pointed to the following reasons for selecting the CTA as a case study for this Guidebook: • The CTA maintains a comprehensive market research program that is increasingly drawing on ITS data to monitor service delivery and to leverage traditional customer research practices. • The CTA is an industry leader in analyzing customer data from smart cards and magnetic stripe cards in support of market research and planning. • Although the CTA’s experience with AVL and APC tech- nology has been fairly brief (less than 3 years), it has developed an innovative array of intranet applications for monitoring service delivery and improving market research and planning. The research team met with CTA staff in March 2007. Organizational Structure CTA’s market research function resides within the Planning section of the Transit Operations division (see Figure A-1). The Planning section includes Strategic Planning (which houses market research), Service Planning, Scheduling, Facil- ities and Traffic Engineering, and Data Services. The creation of the Data Services unit within the Planning section was a consequence of the deployment of EFP, AVL and APC tech- nologies, reflecting the interest of Planning section units in accessing and analyzing data for their purposes. Previously, for example, EFP data served the Finance section, and Plan- ning units had to cross divisions and multiple sections to gain access to data. Until recently, Data Services staff have been responsible for post-processing and merging data from the various systems, as well as developing applications for data analysis. The sys- tems have been producing very large data volumes (3.2 mil- lion data records per day from the AVL and APC systems alone). A reorganization has shifted several staff to the IT sec- tion within the Management and Performance division to take responsibility for database management, while several analysts remained within Data Services. Experience With Smart Cards and Magnetic Stripe Cards Smart cards were introduced in 2002 and are offered in two versions. The Chicago Card (CC) is a stored value device with optional registration. Card balances can be recharged at ticket vending machines. The Chicago Card Plus (CCP) is managed through a Web-based account that is recharged by credit card. All CCPs are registered to users. Registration informa- tion includes a billing address, an optional email address, and an optional field for customers to opt in for market research 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. The demographics of smart and magnetic stripe card users are recovered from customer satisfaction surveys. The most recent count of CC and CCP cards in circulation is 530,000, of which 250,000 are registered with email addresses. Magnetic stripe cards were introduced in 1997. Following their introduction all flash passes and tokens were eliminated. There are a number of alternative purchase options. Single ride cards are sold in packs of 10 and 20. Unlimited ride passes are offered for 1, 7, and 30 day’s duration, as are unlimited ride “visitor” passes covering 1 to 5 days. The un- limited ride cards are activated on first use. Lastly, cards are sold in stored value amounts ranging from $2.25 to $20.00 (rechargeable through ticket vending machines). Transac- tions on the CTA rail system are limited to cards, while cash remains a payment option on the bus system. A P P E N D I X A Chicago Transit Authority Case Study

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

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

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

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

Figure A-7. AVAS intranet reporting and analysis tools.

development of schedules that provide consistent and de- pendable service relies on detailed knowledge of the pattern of vehicle running times on a route. The AVAS recovers a large volume of running time data that can be used to in- form the HASTUS scheduling system. A new analysis tool on the intranet web site facilitates analysis of running time patterns for routes and route segments. The tool provides plots of actual running times in relation to scheduled run- ning times for visual examination. A related tool provides estimates of optimal running and recovery times based on the observed patterns. Figure A-8 illustrates the end-to-end pattern of actual run- ning times in relation to scheduled running times for one of the CTA bus routes. Although the schedules provide greater running times during the morning and evening peak periods, it is apparent from the plot that the actual running times of buses serving the route are notably greater than the scheduled times, except during the 6:00-8:00 pm period. The variation in actual running times also appears to be much greater be- tween 2:30 and 4:30 pm than at other times. A related tool translates the pattern of observed running times into optimal scheduled running and recovery times (see Figure A-9). This calculation is based on the objective of pro- viding running time that is sufficient to accommodate 65% of scheduled trips, and adding recovery time that is sufficient to accommodate 95% of trips serving a route. In the example shown in Figure A-9, the optimal and scheduled running times are fairly close for northbound trips during the Owl, Early AM, and AM Peak periods, while the scheduled running times substantially exceed optimal running times during the Midday and PM Late periods. Although available systemwide, the running time tool has not yet been applied to all routes due to time demands on scheduling staff. There is an expectation that a systemwide evaluation would identify more instances where running time could be reduced than where additional running time would be needed. This expectation reflects traditional sched- uling practices that respond to operator and customer feed- back that tends to focus on circumstances where running time is inadequate, while being less attuned to circumstances where running time is excessive. If so, there would be an opportunity to improve both schedule reliability and sched- ule efficiency. Reconsidering Performance Measures As the CTA board and senior management have become more familiar with applications of ITS data in the Transit Operations division, they have begun to encourage develop- ment of new performance measures. The system level performance metrics that have been tracked over time by CTA include ridership, on-time performance (bus), delay (rail), vehicle and station cleanliness, safety, customer complaints/ commendations, and affordability. Surveys indicate that reliable service is a high priority among customers. While on-time performance is a reliability indicator, it is more relevant in lower frequency service environments, 56 0 20 40 60 80 100 120 140 160 5 9876 10 11 12 13 14 15 16 17 18 19 20 Hour of Day (24 hour time) Tr ip T ra ve l T im e (m inu tes ) AVAS Trip Travel Time Observation Scheduled Travel Time Figure A-8. Actual versus scheduled running times.

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

58 Weekday July 2006 Bus Bunching 0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 00 :0 0 01 :0 0 02 :0 0 03 :3 0 04 :3 0 05 :3 0 06 :3 0 07 :3 0 08 :3 0 09 :3 0 10 :3 0 11 :3 0 12 :3 0 13 :3 0 14 :3 0 15 :3 0 16 :3 0 17 :3 0 18 :3 0 19 :3 0 20 :3 0 21 :3 0 22 :3 0 23 :3 0 % o f H ea dw ay R ec or ds C on si de re d “B un ch ed ” % :15 seconds or less % :30 seconds or less % One Min or less Analysis does not include terminal timepoints, but all other timepoints on all (scheduled) routes are included. Mike Haynes 12-11-2006 D R A F T D R A F T Figure A-11. Incidence of bus bunching by time of day. 70% 71% 72% 73% 74% 75% 76% 77% 78% 79% 80% First Half Second Half A ve ra ge P er ce nt A cc ep ta bl e W ai t 2005 2006 Figure A-10. Wait assessment indicator for CTA bus service.

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

communicate their needs to those involved in the develop- ment of the applications. The seminars also stimulate ques- tions from users and encourage exploration of new uses for the data more effectively than can be achieved through pro- viding system documentation (which many cannot find the time to read). Fourth, staff turnover represents another challenge to de- veloping and implementing new ITS data applications. There are a limited number of people in the transit industry who are both capable of working with ITS data archives and also have a deep understanding of market research, schedul- ing, and planning practices. When staff with these abilities leave they are not easily replaced, and the momentum of the working relationship between the analyst and practitioner is interrupted. Finally, the creation of the Data Services unit within the planning section helped to bridge the gap between ITS data- base management and administration (which resides in the Management & Performance Division) and market research staff. Data applications involve complex queries and joining of data tables, skills that are beyond the capabilities that are typi- cally found in market research and planning environments. References Cui, A., Bus Passenger Origin-Destination Matrix Estimation Using Automated Data Collection Systems. MS thesis. Massachusetts Institute of Technology, Cambridge, 2006. Gross, P., Haynes, M., and Schroeder, J. APC/AVL and GIS Pacific North- west Knowledge Sharing. Final Report. Peer-to-Peer Development Program, ITS Joint Program Office, U.S. Department of Trans- portation, Washington, D.C., 2003. Morpace International and Cambridge Systematics. TCRP Report 47: A Handbook for Measuring Customer Satisfaction and Service Qual- ity. TRB, National Research Council, Washington, D.C., 1999. Multisystems with NuStats International. Fare Structure Pricing Research and Update of Ridership/Revenue Fares Model. Final Report. Chicago Transit Authority, Chicago, IL, 2000. Nelson, J.D. The Potential for Real-Time Passenger Information as Part of an Integrated Bus-Control/Information System. Journal of Advanced Transportation, Vol. 29, 1995, pp. 13–25. Rahbee, A., and Czerwinski, D. Using Entry-Only Automatic Fare Col- lection Data to Estimate Rail Transit Passenger Flows at CTA. Proc., 2002 Transit Chicago Conference, Chicago, IL, 2002. Schweiger, C.L. TCRP Synthesis of Transit Practice 48: Real-Time Bus Ar- rival Information Systems. Transportation Research Board of the National Academies, Washington, D.C., 2003. Zhao, J. The Planning and Analysis Implications of Automated Data Col- lection Systems: Rail Transit OD Matrix Inference and Path Choice Modeling Examples. MS thesis. Massachusetts Institute of Technol- ogy, Cambridge, 2004. CTA Staff Interviewed Jeffrey Busby, Manager, Market Research Alex Cui, Project Coordinator, Data Services Elizabeth Donahue, GIS Analyst, Data Services Mike Haynes, Project Manager, Transit Systems Support, Technology Management Aimee Lee, Manager, Resource Planning Kevin O’Malley, Manager Data Services Jeffrey Schroeder, Manager, Transit Systems Support, Technology Management Jeffrey Sriver, General Manager, Strategic Planning 60