Click for next page ( 68


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 67
67 APPENDIX C TriMet Case Study TriMet provides fixed route bus and light rail service to the Government Affairs Portland, Oregon, metropolitan area. The transit system Operations comprises 91 bus routes, 77 of which connect to a 4-route, Communications and Technology 44-mile light rail network. Its fleet of 614 buses and 105 rail cars carried over 95 million boarding rides in 2006. In addi- The agency's market research unit is housed in the tion, TriMet is contracted to operate the downtown Portland Communications and Technology (CT) division, as is the in- Streetcar. A noteworthy feature of the transit system is formation technology (IT) unit. The service planning and Fareless Square, a downtown zone within which travel is free. scheduling units reside in the Operations division. Over time, Preliminary analysis pointed to the following reasons for there has been a centralization of functions within IT that had selecting TriMet as a case study for this Guidebook: previously been maintained in other units throughout the agency, consistent with the development of TriMet's enter- TriMet was an early adopter of APC technology, and lever- prise data system. The IT unit is also responsible for develop- aged its APC experience in the design of its AVL system. The ing enterprise applications utilizing archived ITS data, many agency has a strong reputation in the industry for its innova- of which involve GIS. tive uses of archived AVL and APC data in market research, performance monitoring, scheduling, and service planning. TriMet's Experience TriMet's market research program is comprehensive, and the With ITS Technology agency has pioneered the development of new attitudinal metrics for understanding customer perceptions. Market re- TriMet's experience with APC technology dates from a searchers at TriMet have also used ITS data to leverage tradi- demonstration project in the early 1980s. In the course of this tional market research practices, and have linked customer project, a severe economic downturn in the region forced sig- perception and satisfaction research to AVL and APC data. nificant service cuts and the release of the agency's entire ride TriMet has developed its GIS capabilities within a fully in- checking staff. Thus the APCs offered the only means of re- tegrated enterprise data environment that includes covering passenger data for internal and NTD reporting. A archived data from ITS technologies. There are a number staff person was assigned responsibility for "fixing the bugs" of high-end GIS users employing ITS data in market re- in the system, including problems with maintenance, trip and search and planning applications. stop referencing (which were based on odometer and time clock stamps in pre-AVL days), and balancing block level The research team met with TriMet staff in January 2007. boardings with alightings. Although considerable success was achieved in overcoming operational problems with the APCs prior to AVL deploy- Organizational Structure ment, usable data recovery rates rarely exceeded 25%. With a TriMet's organizational structure includes five major fleet penetration rate of about 15% in 1990, the APCs were divisions: thus barely able to satisfy TriMet's data needs for NTD reporting. A study was done to validate the accuracy of pas- Finance senger counts and to design a vehicle/APC assignment plan to Capital Projects satisfy NTD sampling requirements (Strathman and Hopper

OCR for page 67
68 1991). This plan was implemented for several years, but prob- prior to deployment (Strathman et al. 2000). Although there lems often occurred at the garage level in assigning APC- were no changes in operations practices during the study pe- equipped buses to NTD sample blocks. With improvement in riod, improvements were observed in on-time performance, the agency's financial condition in the 1990s, passenger count- running times and running time variation. It was hypothe- ing for NTD reporting was contracted out and manually sized that the improvements resulted from the "instrumenta- recovered. At that point the APC buses were redeployed to tion effects" of the new system associated with operators support service planning information needs. having better knowledge of their schedule status and dis- Despite the initial problems, TriMet's experience with patchers having better knowledge of vehicles' operating status. APCs proved valuable in planning its AVL system. The AVL TriMet's estimated breakdown of fare revenue in fiscal year project team was headed by the staff person who had been 2008 is 39% farebox (cash), 9% pre-paid ticket sales, and 52% responsible for NTD reporting and also included the staff per- flash pass sales. TriMet does not have advanced electronic son who had been responsible for maintaining the APCs. They fare payment technologies, such as magnetic stripe or smart recognized that AVL's improved location referencing capabil- cards. Electronic registering fareboxes have been in use since ities would, when integrated with APCs, yield much higher 1989. The fareboxes are not integrated with the AVL system, rates of successful passenger data recovery. Specifications for and reporting is thus limited to daily cash receipt totals. the AVL system thus included a capability to produce and Ticket vending machines have also been in service since 1986, store passenger and operating data records at the stop level. with reporting limited to time, location, and fare instrument Bandwidth limitations prevented transmission of the large purchased. Most of the ticket vending machines are located volume of stop level data records over the radio system, at light rail stations. Generally speaking, TriMet's electronic requiring storage on an on-board computer. fare payment technologies are not used for service delivery or The requirement of location-defined data recovery distin- market research analysis. guished the specification of TriMet's AVL system from other The remaining technologies include automated phone sys- GPS-based systems that were deployed in the 1990s. All AVL tems and the Web. Tracking software has been used to log systems featured transmission of real time vehicle location Web visits since 2002. Information is thus available on Web status data to dispatchers on a 30-90 second polling cycle. links providing trip planning services and real time vehicle ar- While this information had considerable value in support of rival times at stops ("Transit Tracker"). Customers can also operations management, and has since been extended to cus- report compliments, suggestions, and complaints on the tomer information applications (e.g., in posting estimated Web. The automated phone system provides trip planning, vehicle arrival times at selected stops and on the Web), it was arrival time, and customer feedback services, all of which are incompatible with applications requiring information about logged. specific locations (e.g., stops, time points, route origins and destinations). ITS Data Management A side benefit from on-board storage of AVL data was a re- duction in the cost of APC units. With data storage comput- A central "core" set of data is maintained and managed as a ers already on board to serve the AVL system, the cost of foundation for all TriMet ITS initiatives. The architecture of adding APC units in the late 1990s fell to about $1,000/ the agency's enterprise data system is illustrated in Figure C-1. vehicle. Thus, TriMet decided to specify APCs in all new bus The diagram reflects the holistic nature of the data environ- acquisitions. Presently, about 75% of the bus fleet and 25% ment and the importance of integration of ITS systems and of the light rail fleet are APC-equipped. While this level of applications. This ensures that the publication and dissemina- penetration is higher than what is needed for NTD and inter- tion of data are accurate and consistent across all systems. nal ridership reporting, it allows staff to analyze operations Base detailed features of the transit system (e.g., routes, and evaluate service in ways that are unique in the industry. stops) are represented and maintained in Definition & Main- A final feature of TriMet's AVL system that distinguished it tenance tables. This information is used in conjunction with from others deployed in the 1990s was a control head installed vehicle itineraries, produced by HASTUS scheduling soft- near the vehicle operator. The control head displays the vehi- ware, to prepare the published schedule. After the data are cle's status in relation to its schedule (minutes early/late), and a validated, the prepared schedules are written to cards that are keypad allows the operator to send pre-programmed text mes- inserted by vehicle operators in the log in process. sages to the dispatch center (e.g., 74 = "Mechanical Problem-- Vehicles in service recover and store data records from the Out of Service"). The text message records are stored for offline AVL and APC systems. The data records (including schedule analysis. and AVL-APC operating data) are downloaded at the end of The AVL system was deployed in 1998. A study compared each day at the garages. Post processing is then undertaken by system performance a year after deployment to performance Operations staff to ensure the validity of the recovered data.

OCR for page 67
Figure C-1. TriMet's enterprise data system.

OCR for page 67
70 Post processing involves a number of steps: (1) matching from the Work Order, CSI, Event History, and Definition & vehicles' AVL data records to their schedules and the base Maintenance tables. map of stops and time points associated with the assigned GIS staff prepare a basemap for systems throughout the work; (2) event data recorded at locations other than those agency for import, including HASTUS, the Automated Trip represented in the base map are assigned to base map loca- Information System (ATIS), and the Trapeze Paratransit sys- tions (examples include instances where a vehicle drops off or tem. GIS data collection is used to collect location and picks up passengers between scheduled stops, records for amenity information for features of the transit system. temporary re-routes, the record of a vehicle's maximum As the description of post processing activities indicates, speed between stops, or an event record); (3) screening to units throughout TriMet are directly responsible for main- identify extreme values, which may indicate a malfunction- taining data in Oracle tables within the enterprise data sys- ing unit (records with extreme values are retained, but flagged tem. GIS staff adhere to the same IT standards (Oracle, Java, to indicate that the data may be suspect); (4) "zeroing out" Linux) and the same Systems Engineering approach to appli- passenger loads at the service block level, and proportionately cation development. adjusting boarding and alighting data to correspond to the The evolution of the enterprise data system is guided by block load adjustment. system integration strategies that include a database design After post processing, data are forwarded to BDS (Bus review process and procurement policies for new systems. Dispatch System) Event History tables. Event history data, ac- cumulating at a rate of about 600,000 records per day, are Service Delivery Monitoring then ready to be accessed by staff for analysis. and Analysis Selected data, such as operator-keyed event records related to stop maintenance and repair, automatically generate Few, if any, transit properties in the U.S. have developed a Service Requests that are then evaluated by staff and (if re- service delivery analysis and monitoring capability compara- quired) converted into Work Orders for work. The status of a ble to what TriMet has achieved with its archived AVL and work order is monitored by the Stop and Amenities Work APC data. These achievements are a consequence of the fore- Order Tracking System (SAMW), and the history is retained thought given to the specification of ITS data recovery and for future reference. archiving, the high penetration of APC units in the fleet, and Following the stop maintenance and repair example, the the efforts of highly skilled staff analysts. process might also be initiated through the Customer Service The service delivery monitoring and analysis activities de- Information (CSI) program. In this case the service request scribed in this section can be distinguished between regular originates from a customer Web or phone contact. In addi- performance reporting, which usually occurs quarterly, and tion to directing the information to Service Requests, the cus- ad hoc analysis, which is undertaken to address specific needs tomer contact would also be logged in a CSI table. or issues. However, it should be emphasized that regular Data from the Prepared Schedule tables are also linked to reporting activity and ad hoc analysis have been strategically several other customer information systems. The Automated related in that the content and coverage of performance re- Trip Information System (ATIS), supports the trip planning ports have evolved to incorporate what has been learned from program offered through the Web and by phone. Transit ad hoc analysis. Tracker supports the phone and Web-based program that Quarterly performance reports are posted on the agency provides next arrival times. Data are also generated in the Intranet in both pdf and Excel format. The trip level sum- Google Transit Feed Spec (GTFS) format and made available mary report serves as the cornerstone of quarterly perfor- for public access for external applications, such as Google mance reporting. The report provided detailed information on Transit. passenger activity, on-time performance, and vehicle running The final operations-related component of the enterprise times for every scheduled trip in the system. Selected infor- data system is the Accident Incident Response System (ACID) mation from the spring 2007 trip performance report for and tables, which contain data from incident reports. scheduled AM Peak period inbound trips on the Route Spatial data are directly maintained in Oracle tables to en- 14-Hawthorne is shown in Table C-1. Capacity utilization is sure consistent integration of ITS data across the enterprise represented by the average maximum passenger load per trip. system and to facilitate access for GIS applications. A number Information on the percentage of trips where the maximum of GIS applications have been developed to support desktop passenger load exceeds 80% of a vehicle's maximum design mapping and analysis of data from various Oracle tables. For capacity provides an indicator of both the variance in capac- example, the Real Time Bus Mapper is a Web-based applica- ity utilization and the extent of excess demand. tion that displays the current location of buses in service. The scheduled running time for each trip serves as a bench- Other GIS applications have been developed to map data mark against which actual running times are compared.

OCR for page 67
71 Table C-1. Route 14 Hawthorne trip level summary report, spring 2007 (AM peak inbound trips). Start Avg. Max Percent Over Scheduled Median Run 60th% Run 80th% Run Percent Percent Percent Time Load Capacity Run Time Time Less Idle Time Less Idle Time Less Idle On Time Early Late 7:02 AM 45 0% 0:36:00 0:31:26 0:32:18 0:33:19 90% 5% 4% 7:06 AM 27 0% 0:36:00 0:35:10 0:35:44 0:36:48 97% 1% 2% 7:10 AM 43 18% 0:37:00 0:36:14 0:36:38 0:38:20 93% 6% 1% 7:13 AM 46 0% 0:38:00 0:36:37 0:37:18 0:37:58 94% 2% 4% 7:16 AM 31 0% 0:38:00 0:35:49 0:36:14 0:37:36 97% 2% 1% 7:19 AM 31 0% 0:38:00 0:35:39 0:36:12 0:37:38 98% 1% 0% 7:24 AM 49 24% 0:38:00 0:36:25 0:37:08 0:38:14 97% 3% 1% 7:29 AM 32 5% 0:39:00 0:35:43 0:36:32 0:37:21 95% 1% 4% 7:34 AM 38 8% 0:39:00 0:36:40 0:37:32 0:39:18 99% 0% 1% 7:38 AM 19 0% 0:39:00 0:34:44 0:35:02 0:36:08 86% 12% 2% 7:43 AM 40 4% 0:39:00 0:36:46 0:37:22 0:38:10 96% 4% 0% 7:47 AM 33 2% 0:39:00 0:35:56 0:36:36 0:37:58 98% 2% 0% 7:52 AM 40 10% 0:39:00 0:35:48 0:36:12 0:37:32 99% 1% 0% 8:04 AM 45 17% 0:39:00 0:35:28 0:35:59 0:37:47 89% 3% 7% 8:10 AM 40 50% 0:39:00 0:33:28 0:34:00 0:35:42 95% 3% 2% 8:17 AM 35 8% 0:38:00 0:36:14 0:36:44 0:38:20 95% 2% 3% 8:26 AM 49 31% 0:38:00 0:33:14 0:34:25 0:35:49 98% 1% 0% 8:33 AM 31 2% 0:38:00 0:35:38 0:36:14 0:37:42 91% 7% 2% 8:44 AM 35 5% 0:36:00 0:35:06 0:35:25 0:37:27 96% 2% 2% 8:53 AM 40 9% 0:36:00 0:34:48 0:35:36 0:37:18 97% 2% 1% 9:04 AM 42 15% 0:36:00 0:37:22 0:37:38 0:38:56 98% 0% 1% Traditional scheduling practice has been to set scheduled For example, when scheduled running is inadequate, the running times to approximate the median, although stan- percentage of late departures increases and the percentage of dards exceeding the median have become more common. on-time and early departures declines. In the case of the trips Actual running times in the quarterly trip report are given for covered in Table C-1, scheduled running times are generally the median, 60th, and 80th percentile of the distribution. A greater than the actual median, 60th percentile, and 80th per- noteworthy feature of the running time statistics in the report centile values, indicating that the schedule provides more than is that they have been adjusted to remove excess dwell time, adequate time for operators to complete their trips. Conse- using threshold dwell times required to service stops on the quently, on-time performance is quite high, and the incidence route. Dwells exceeding the threshold are interpreted as likely of early departures generally exceeds the incidence of late to reflect holding actions associated with maintaining the departures. schedule. Although trip report information on mean maximum pas- Systematic departures from scheduled running times sig- senger loads indicated that there was generally sufficient nal a need to add or trim running times from the schedule. capacity, the variance in maximum passenger loads among When a decision is made that there is a need to adjust the blocks of trips on routes during peak service periods also sug- schedule, plots of actual running times are also produced to gested that deviations from scheduled headways were con- aid the process. These plots are similar to those described in tributing to instances of overloading. This was confirmed by ad the CTA case study. Staff commented that three schedule hoc statistical analysis of the relationship between headway de- writers can now do the work that used to require six, and that viation (i.e., "bus bunching") and passenger loads. Figure C-2 schedule adherence has improved over time despite worsen- illustrates the relationship for AM Peak inbound trips on Route ing traffic congestion. Also, the running time reports are 14-Hawthorne. In this plot, the average load for vehicles main- revealing opportunities to remove time from schedules, taining their scheduled headway (0.00 headway delay) is where operator and customer feedback in the past mostly approximately 45 passengers. For vehicles whose actual head- focused on the need to add time. way has grown to two minutes greater than schedule, average On-time performance information is presented in the last loads are about 55 passengers. Conversely, for vehicles whose three columns of Table C-1. "Late" is defined as departures actual headway has decreased to two minutes less than sched- that are more than 5 min after the scheduled departure time; ule, average loads fall to about 35 passengers. Thus, it was "early" is defined as departures that are more than 1 min concluded that the solution for managing (leveling) passenger before the scheduled departure time; and "on time" is defined loads lies in improving headway management. as departures that are between 1 min early and 5 min late. Further statistical analysis exploring the root causes of There is a readily apparent relationship between scheduled headway deviations identified late departures from garages running time, actual running time, and on-time performance. and route origins as principal contributors to the problem

OCR for page 67
72 80 70 60 50 Passenger Load 40 30 20 10 0 -8.00 -6.00 -4.00 -2.00 0.00 2.00 4.00 6.00 8.00 Headway Delay (minutes) Figure C-2. Passenger loads and headway delay on Route 14-Hawthorne. (AM peak inbound trips at peak load point) (Strathman et al. 2003). Tabulations of AVL data exhibited adherence and excess wait are generally, but not directly, re- instances of habitual late departures from origins (following lated. For example, for a given percentage value of headway adequate layover) and from garages. Regular reporting of adherence, excess waiting time tends to be greater for larger these tabulations to field supervisors and garage managers than for smaller scheduled headways. followed. (It should be emphasized that TriMet supervisors It is apparent in Table C-2 that excess wait time and on- and managers cannot pursue disciplinary action on the basis time performance are related. In contrast to the high on-time of evidence from AVL and APC data. Uses of the data are performance reported in Table C-1, the Table C-2 values as- limited to facilitating accepted operations management sociated with the worst excess wait situations are much lower. practices.) Also, it is apparent that excess waiting and the gap between For customers, deviations from scheduled headways trans- scheduled and actual headways are the result of lateness. late into increased waiting time and a greater likelihood of TriMet's excess wait measure is based on the same headway being passed up by an overloaded vehicle. It has long been and boarding data as the CTA's wait index measure. In addi- understood in the transit industry that waiting time at stops tion to AVL data, which are recovered fleet-wide, both meas- and stations is considered by customers to be much more ures require extensive recovery of passenger data. At the CTA, onerous than their line haul time or time spent in accessing these data are recovered from smart and magnetic stripe card stops or final destinations. usage, while at TriMet the data are recovered by APCs. Both TriMet staff developed a measure to capture the additional CTA's wait index and TriMet's wait time measures are easily time passengers spend waiting for scheduled service when understood. However, an advantage of TriMet's wait time vehicles' actual headways deviate from their scheduled head- measure is that, given a monetary value of waiting time (e.g., way. This measure, termed "excess wait," is reported quarterly Mohring et al. 1987), one can directly determine the impact by route, direction, and time period. Table C-2 presents sorted on customer welfare from TriMet's measure. values of excess wait time for the spring 2007 signup. In this Another benefit from access to comprehensive passenger case the excess wait value in the first row of the table indicates data at TriMet is evident in the agency's approach to evaluating that passengers on outbound PM Peak trips on Route 8-NE the level of service supplied to routes. Analogous to standard 15th Ave. waited three and a half minutes longer for service practice in evaluations of highway capacity utilization, TriMet than they would have had headways been maintained. calculates passenger volumes in relation to vehicle capacities Headway adherence values are also reported in Table C-2. over varying hour-long time periods. An example is shown in These values represent the percentage of instances in which Table C-3, which presents sorted weekday volume/capacity actual headways are within 50% of their scheduled values at measures for the spring 2007 signup. In this example, the high- time points for the corresponding service blocks. Headway est capacity utilization in the system occurred with outbound

OCR for page 67
73 Table C-2. Headway adherence and excess wait time, spring 2007 (sorted by excess wait time). Daily Scheduled Headway Excess Wait Route, Direction, Time of Day Trips On Time Early Late Headway Adherence (min.) 8-NE 15th Ave - Outbound - PM Peak 14.0 45% 3% 52% 8:58 48% 3:27 4-Fessenden - Outbound - PM Peak 10.0 46% 4% 50% 12:37 49% 3:27 15-Belmont - Outbound - PM Peak 20.0 63% 2% 34% 6:14 49% 3:15 8-Jackson Park - Inbound - PM Peak 15.0 57% 1% 42% 8:19 47% 3:10 96-Tualatin/I-5 - Outbound - PM Peak 10.0 60% 0% 40% 10:57 67% 2:47 64-Marquam Hill/Tigard TC - Inbound - AM Peak 5.0 96% 0% 4% 19:50 90% 2:43 4-Division - Outbound - PM Peak 15.0 65% 6% 29% 7:58 54% 2:33 4-Division - Inbound - PM Peak 10.0 67% 4% 29% 12:43 59% 2:23 51-Vista - Inbound - AM Peak 8.4 89% 8% 2% 17:05 91% 2:11 20-Burnside/Stark - Outbound - PM Peak 7.0 66% 6% 27% 15:14 75% 2:07 17-Holgate - Outbound - PM Peak 11.0 59% 7% 33% 11:00 65% 2:07 94-Sherwood/Pacific Hwy Express - Outbound - PM Peak 14.0 73% 0% 26% 9:21 66% 2:05 72-Killingsworth/82nd Ave - Inbound - PM Peak 16.0 73% 7% 20% 7:42 50% 2:02 99-McLoughlin Express - Outbound - PM Peak 8.0 86% 0% 14% 17:09 84% 1:59 17-NW 21st Ave/St Helens Rd - Outbound - PM Peak 8.0 42% 3% 55% 15:00 75% 1:51 15-NW 23rd Ave - Outbound - Midday 31.0 53% 8% 39% 13:27 65% 1:48 4-Division - Inbound - AM Peak 15.0 89% 2% 9% 7:48 59% 1:43 32-Oatfield - Outbound - PM Peak 6.0 74% 4% 22% 24:34 88% 1:43 9-Powell - Outbound - PM Peak 14.0 62% 7% 32% 8:03 64% 1:42 12-Barbur Blvd - Outbound - Night 19.0 64% 3% 33% 20:08 81% 1:39 71-60th Ave/122nd Ave - Inbound - Midday 27.0 82% 5% 14% 16:18 76% 1:39 15-NW 23rd Ave - Outbound - PM Peak 8.0 48% 8% 44% 15:00 70% 1:38 4-Fessenden - Outbound - Night 23.0 71% 5% 24% 19:52 83% 1:34 72-Killingsworth/82nd Ave - Outbound - PM Peak 16.0 80% 5% 15% 7:43 58% 1:34 8-Jackson Park - Outbound - Midday 32.0 71% 3% 27% 13:11 80% 1:33 15-NW 23rd Ave - Outbound - AM Peak 22.0 72% 5% 24% 5:42 49% 1:33 Table C-3. Weekday peak hour capacity, spring 2007 (sorted by maximum achievable capacity utilization). Hourly Load/Seat Load/Achievable Route Dir. Begin Time End Time Location Trips Load Capacity Capacity 8-Jackson Park Out 5:02 AM 6:01 AM SW 3rd betw. Oak & Stark 1 53 123% 95% MAX Blue Line Out 4:10 PM 5:09 PM Pioneer Sq. South MAX Sta. 6 1,470 191% 92% 61-Marq. Hill/Beaverton TC Out 3:42 PM 4:41 PM SW Campus Dr & Terwilliger 1 52 120% 92% 72-Killingsworth/82nd Ave In 12:08 PM 1:07 PM NE 82nd & MAX Overpass 5 256 131% 92% MAX Blue Line Out 4:20 PM 5:19 PM Pioneer Sq. South MAX Sta. 7 1,694 189% 91% 72-Killingsworth/82nd Ave Out 4:18 PM 5:17 PM SE 82nd & Powell 6 303 129% 90% 72-Killingsworth/82nd Ave In 12:18 PM 1:17 PM NE 82nd & MAX Overpass 5 250 128% 89% MAX Blue Line Out 4:38 PM 5:37 PM Pioneer Sq. South MAX Sta. 7 1,650 184% 89% 72-Killingsworth/82nd Ave Out 7:59 AM 8:58 AM SE 82nd & Powell 5 248 127% 89% 72-Killingsworth/82nd Ave Out 4:09 PM 5:08 PM SE 82nd & Powell 7 345 126% 88% 61-Marq. Hill/Beaverton TC Out 4:12 PM 5:11 PM SW Campus Dr & Terwilliger 2 98 114% 88% 88-Hart/198th Ave Out 12:16 PM 1:15 PM Beaverton Transit Center 1 28 100% 88% MAX Blue Line Out 4:47 PM 5:46 PM Pioneer Sq. South MAX Sta. 6 1,395 182% 87% 72-Killingsworth/82nd Ave Out 7:41 AM 8:40 AM SE 82nd & Powell 6 293 125% 87% 72-Killingsworth/82nd Ave In 2:45 PM 3:44 PM NE 82nd & MAX Overpass 8 391 124% 87% 6-M. L. King Jr Blvd Out 2:58 PM 3:57 PM NE Grand & Pacific 3 146 117% 87% 15-Belmont Out 6:20 PM 7:19 PM SW Salmon & 5th 4 194 125% 87% 72-Killingsworth/82nd Ave Out 7:49 AM 8:48 AM SE 82nd & Powell 6 291 125% 87% 8-Jackson Park Out 5:17 AM 6:16 AM SW 3rd betw. Oak & Stark 2 97 118% 86% 4-Division Out 2:13 PM 3:12 PM SE Division & 12th 4 193 121% 86% 72-Killingsworth/82nd Ave Out 7:05 PM 8:04 PM SE 82nd & Powell 4 193 124% 86%

OCR for page 67
74 service on Route 8-Jackson Park at SW 3rd Ave. between Oak primarily for stop planning, determining where amenities and Stark St. during the hour period beginning at 5:02 AM. A should be added, and identifying when and where customers single scheduled trip provided service during the period and its with disabilities are connecting with the system. A recent large average hourly load was 53 passengers. This load was 123% of scale planning effort that focused on stop placement and stop the seat capacity of service supplied and 95% of the maximum consolidation also drew heavily on passenger census informa- achievable vehicle capacity. tion. Passenger census information is also important in Examining service utilization over blocks of trips gives market research, as will be further discussed in the following service planners a better sense of when and where additional section. capacity is needed. Probably the greatest utility from this ap- Conducting the passenger census using APC data has a proach comes in evaluating capacity utilization in corridors number of advantages over the manual recording process served by multiple routes. Here, passenger loads and vehicle that had been previously used. First, in the present system, capacities can be easily aggregated over routes and trips in the data are current in comparison to the manual approach, corridor, providing a capacity utilization measure that is which operated on a five-year cycle. Second, in the present more consistent with what customers see and with the per- system, data are comprehensive and precise in comparison to ceptions of crowding they report in satisfaction surveys. the old system, where each stop was sampled once in a six- The extensive deployment of APCs also allows TriMet to month data collection period. Third, the new system replaces produce a stop level passenger census on a twice-yearly cycle. a manual process that was expensive, costing about $250,000 An example of passenger census information is shown in when it was last done in the 1990s. Table C-4 for inbound stops on a section of Route 8-NE 15th Operator-keyed event data are a key feature of TriMet's Ave. in weekday service. Average daily boarding and alighting AVL system, with 51 preprogrammed messages that operators statistics are reported, as are monthly lift deployments (from can transmit to dispatchers by pressing selected numbers on AVL vehicle monitoring data records). Passenger census in- the vehicle control head. Events include incidents and cir- formation is also organized and reported by stop, summariz- cumstances that are directly or indirectly related to customers' ing passenger movements across all routes serving given stops. riding experience. Table C-5 shows a frequency breakdown Passenger census information is widely used throughout for 16 event types recorded between June 3 and July 22, 2007. the agency. Within Operations the information is used As previously discussed, events requiring maintenance or Table C-4. TriMet passenger census, spring 2007 signup (weekday average). Route No. Route Description Direction Stop Location Ons Offs Total Monthly Lifts 8 8-NE 15th Ave 1 NE 15th & Killingsworth 93 22 115 23 8 8-NE 15th Ave 1 NE 15th & Sumner 37 5 42 1 8 8-NE 15th Ave 1 NE 15th & Alberta 254 105 359 32 8 8-NE 15th Ave 1 NE 15th & Going 50 8 58 15 8 8-NE 15th Ave 1 NE 15th & Prescott 68 11 79 3 8 8-NE 15th Ave 1 NE 15th & Mason 40 13 53 1 8 8-NE 15th Ave 1 NE 15th & Failing 31 9 40 2 8 8-NE 15th Ave 1 NE 15th & Fremont 106 49 155 21 8 8-NE 15th Ave 1 NE 15th & Siskiyou 21 6 27 0 8 8-NE 15th Ave 1 NE 15th & Knott 40 8 48 7 8 8-NE 15th Ave 1 NE 15th & Brazee 25 11 36 0 8 8-NE 15th Ave 1 NE 15th & Thompson 17 3 20 0 8 8-NE 15th Ave 1 NE 15th & Tillamook 34 20 54 29 8 8-NE 15th Ave 1 NE 15th & Weidler 48 97 145 15 8 8-NE 15th Ave 1 NE 15th & Halsey 27 52 79 7 8 8-NE 15th Ave 1 NE 16th & Multnomah 17 18 35 1 8 8-NE 15th Ave 1 NE Multnomah & 13th 122 178 300 20 8 8-NE 15th Ave 1 NE Multnomah & 11th 97 49 146 8 8 8-NE 15th Ave 1 NE Multnomah & 9th 80 29 109 16 8 8-NE 15th Ave 1 NE Multnomah & 7th 28 24 52 23 8 8-NE 15th Ave 1 NE Multnomah & Grand 24 39 63 6 8 8-NE 15th Ave 1 NE Multnomah & 3rd 22 13 35 3 8 8-NE 15th Ave 1 Rose Quarter Transit Center 241 169 410 68 8 8-NE 15th Ave 1 NW 3rd & Flanders 99 78 177 16 8 8-NE 15th Ave 1 NW 3rd & Couch 125 117 242 31 8 8-NE 15th Ave 1 SW 3rd between Oak & Stark 14 204 218 44 8 8-NE 15th Ave 1 SW 3rd & Yamhill 2 9 11 3 8 8-NE 15th Ave 1 SW 3rd & Madison 1 9 10 1

OCR for page 67
75 Table C-5. Customer-oriented operator-keyed Changes in running time and running time variation have events, JuneJuly 2007. been evaluated following the implementation of signal prior- ity technology at selected intersections in the city of Portland Event Type Number of Events Fare evasion 19269 (Kimpel et al. 2005). And changes in passenger activity, run- Passup-overload 827 ning time, and running time variation have been evaluated Mechanical problem-out of service 688 following the relocation and consolidation of stops on se- Skip stopping per dispatch 439 lected bus routes (El-Geneidy et al. 2006). Route blocked 364 Lost/found item 349 Emergency-police 294 Use of ITS Data in Market Research Sign/pole problem @ stop 282 Litter problem 231 TriMet maintains a comprehensive market research pro- Verbal passenger 225 gram. The most recent systemwide on-board survey of origins Lift-rolling/passup 142 and destinations was completed in 2000. Since the 2000 sur- Shelter glass broken 114 vey, O-D data has been updated periodically on routes where No-injury accident 106 Emergency-medical 99 major service changes have occurred. An annual attitude and Shelter graffiti 92 awareness survey of riders and non-riders in the Portland re- Report hazard 76 gion is conducted by telephone. Regional telephone surveys of customer satisfaction are also conducted periodically. Lastly, TriMet conducts an annual on-board survey to determine the repair are automatically forwarded as work requests, ensuring mix and usage of fare instruments. The data from this survey a more timely response to these problems. are used to understand average pass use rates, employer pass Given that the event messages are operator initiated, it is usage, and who uses which method of fare payment. likely that event counts understate the actual incidence of TriMet's integration of ITS data with advanced market re- most of the phenomena that are covered in event records. search practice is best illustrated in two research projects. The Nevertheless, the event data provide important information first used surveyed customer satisfaction data to construct in support of customer service and operations functions. A satisfaction impact scores for twenty-seven service attributes. customer complaint about being passed up, for example, can Customer perceptions of service attributes were then selec- be more readily resolved when event records show that the tively compared to "reality" (represented by ITS data) to vehicle had been instructed to skip stops or was overloaded identify feasible improvements the agency could make that with passengers. The ability to document such facets of ser- would have the largest customer satisfaction benefit. The sec- vice delivery, along with the security enhancing features of ond project drew on surveyed attitudinal data from riders the AVL system, help to explain why many operators favor- and non-riders in the Portland region in a market segmenta- ably view the addition of this and other technologies. tion analysis that sought to identify improvements that would Fare evasions are easily the most common operator-keyed have the greatest likelihood of retaining existing riders and events. Analysis of these event records can reveal patterns in attracting new riders. time and space that can inform fare payment enforcement. TriMet's research on service attribute satisfaction is based With information such as that portrayed in Figure C-3, on impact scores constructed from two pieces of customer supervisors can assign fare inspectors to the "hot spots" where survey information. The first piece is provided by the fare evasion is concentrated. While the map in Figure C-3 por- response to a question asking whether the customer had a trays fare evasion frequency at the traffic analysis zone level, problem related to a given attribute on his/her most recent the frequencies can also be portrayed at the stop level at finer trip. Those customers who respond "yes" are placed in one geographic resolution. group and those who respond "no" are placed in a second Beyond facilitating the regular internal service delivery group. The second piece of information is the customer's monitoring functions at TriMet, AVL and APC data also con- overall satisfaction rating (on a ten-point scale) of that at- tribute to external reporting requirements and ad hoc evalu- tribute. For each attribute, a satisfaction impact score is cal- ations. TriMet now samples the AVL-APC data archive for culated as the difference in mean satisfaction between the NTD reporting for bus and light rail service. The data valida- "yes" and "no" groups multiplied by the percentage of re- tion and sampling procedures for the respective modes are spondents in the "yes" group. Thus a high satisfaction impact described in Kimpel et al. (2003) and Strathman et al. (2005). score could be the result of a small percentage of customers Archived AVL data were used to evaluate the accuracy and experiencing a problem that has a large effect on their overall precision of the real time vehicle arrival information pro- satisfaction (e.g., safety-related attributes) or, at the other duced by the agency's Transit Tracker system (Crout 2007). extreme, a large percentage of customers experiencing a

OCR for page 67
Figure C-3. Geographic incidence of operator-keyed fare evasion events, June 2007. (Supervisor District 3)

OCR for page 67
77 problem that has a modest effect on their satisfaction (e.g., Service Delivery availability of seats on a vehicle). The twenty-six attributes Frequency/short wait times Reliable service/on schedule included in the customer satisfaction survey are presented in Vehicle not overcrowded Figure C-4. The attributes can be grouped into categories re- Courteous/quick drivers Driver assistance/special needs lated to service delivery, information provision, comfort, Adequate capacity at park & ride lots amenities, safety, and fare payment. Among bus riders, TriMet found the satisfaction impact Information Provision Availability of real time information scores for twelve of the twenty-six attributes were statistically Delays explained/announced significant (see Figure C-5). The two largest satisfaction Clearly marked/visible stops Clear/timely announcements impact scores (and four of the top twelve) are associated Availability of schedule information at stops with service delivery. Four comfort-related attributes are Availability of schedules/maps also among the top twelve, followed by three customer Comfort information-related attributes. Absence of offensive odors The impact scores help to identify the attributes that de- Smoothness of rides/stops Physical condition of the vehicle tract from customers' travel experience and can be expected Availability of seats on vehicle to influence their future travel choices. While some factors Comfort of seats on vehicle Cleanliness of vehicle exterior (e.g., odors) are beyond a transit provider's control, most are Cleanliness of vehicle interior not. With limited resources, managers cannot respond to all Cleanliness of stops/stations of the attributes that have an important influence on cus- Freedom from nuisance behavior tomer satisfaction, but the impact scores do help to prioritize Amenities actions. ITS data can facilitate this process in two ways. First, Availability of shelters the data can help to identify where given problems are most Safety acute. For example, combining schedule adherence data from Safety from crime at stops AVL with APC passenger load data, one can ask, "Where are Safety from crime on vehicle reliability problems affecting the greatest number of cus- Fare Payment tomers?" Also, using the same data, one can ask, "Is our worst Affordability of trip Ease of paying fares overcrowding due to limited capacity or to poorly main- tained headways?" Or, with AVL schedule adherence and Figure C-4. Attributes included in TriMet's APC boarding and alighting data, one can ask "What are our customer satisfaction impact analysis. highest volume/longest wait stops lacking shelters?" In short, Bus Impact Scores March 2004 30 Service Comfort Amenities Information 25 Impact Score 20 15 10 5 0 fo e e s ps ps e s s s y d cl ul op cl er y or nc de in la to to hi hi d iv od st ue w de he e /s ts ve ve dr m ro e es eq e sc ra e bl -ti k iv on on rc nc rid Fr ic si ns al te n ve qu ou s ts /o vi re el of fe at to d/ ea e sh s/ nn of of se ic s ke no ou s es rv /a of e y of of ar te nc lit in se e hn y m cl ur bi rt y a lit se t lit e pl hi fo oo la rly Co bi bl Ab bi Ex Ve m ai la Sm lia ea la Av Co ai Re ai Cl Av Av Figure C-5. Satisfaction impact scores from TriMet customer survey.

OCR for page 67
78 ITS data can help managers achieve the greatest improvement 1993, and updated in 1998 and 2004. In the 2004 study, a in satisfaction with the limited resources at their disposal. population survey was conducted by telephone with over Second, ITS data can supplement customer satisfaction 2,600 service area residents. Along with demographic and analysis over time through ongoing monitoring of service travel activity information, the survey recovered scaled re- delivery, as well as by linking to replications of customer sat- sponses to a variety of questions related to travel preferences, isfaction surveys. For example, an earlier satisfaction impact the benefits associated with transit as a travel option, TriMet's study was completed in 2001, and the attributes with the five performance as the region's transit provider, and general highest scores were, in order conditions in the region. The first step in the analysis of the survey data employed a Frequency, factor analysis of scaled responses to the battery of attitudinal Availability of a seat, questions. The factor analysis identified five underlying Explain/announce delays, principal components organized around the responses to Reliable service/on schedule, and eighteen of the survey's attitudinal questions. Next, cluster Availability of schedules and maps. analysis was performed on the eighteen attitudinal question responses. This analysis yielded the five cluster groups, or Comparing the 2001 and 2004 findings, it appears that re- market segments, presented in Figure C-6. Several things liability is an increasingly important satisfaction issue for cus- should be noted about the characteristics associated with tomers. How does the change in satisfaction correspond to these market segments. First, with respect to travel mode, re- on-time performance trends over the period? Has service spondents were defined to be transit riders if they reported reliability deteriorated or has reliability become increasingly having taken more than two trips on the system in the past important to customers? month. Second, while the clusters were generated from Further analysis of ITS data allowed TriMet market re- attitudinal data, distinctions between segments related to in- searchers to gain a better understanding of the linkages come, demography, and length of residence sometimes between service delivery trends and customer satisfaction. emerged. For example, while the trend in on-time performance over The largest market segment cluster, labeled "Transit is a the period was essentially flat, examination of AVL headway Lifestyle Choice," represents over one-third of the survey re- data revealed that the incidence of bus bunching had in- spondents, 43% of transit riders, and 28% of non-riders. creased. Bunching can affect customer perceptions in two Compared to other segments, members of this group tend ways. First, they are likely to view bus bunching as a reliabil- to be more supportive of transit (particularly bus) as a travel ity problem, independent of what on-time performance sta- option and hold more favorable opinions about TriMet's tistics say. Second, bus bunching removes effective capacity performance as a transit provider. They view transit as a from a route and thus leads to an increase in the incidence of convenient and economical means of travel, and recognize overcrowding, which was also observed in the 2004 study but transit's social and environmental contributions to the not in 2001. Thus, in this instance, the resolution of two region's livability. They tend to be newer to the Portland customer satisfaction problems depended on implementing area, better educated, and more likely to reside in urban operations control measures to better manage bus headways. neighborhoods. Given these defining characteristics, TriMet The ability to identify and respond to distinct travel market market researchers concluded that the strategy with the segments has been found to be a common feature of transit greatest likelihood of retaining current transit riders and agencies that have built high ridership systems (TranSystems et attracting non-riders from this segment should focus on al. 2007). These agencies strategically target marketing and cus- improving service. tomer information, fare options, and service improvements, The second largest market segment is labeled "I'm not ensuring that resources are invested where they will produce Comfortable Riding the Bus." This group includes over one- maximum ridership returns. ITS data offers a cost-effective, quarter of the sample and more than one-fifth of the transit timely, and comprehensive means of monitoring customer re- riders surveyed. Its disproportionately female members ac- sponses to targeted marketing investments. knowledge transit's benefits, but also express discomfort with Market segmentation research in the transit industry is being around strangers and concern about their personal evolving from the days when travelers were distinguished by safety. Retention of existing riders will likely depend on pro- age, sex, income and car ownership to more sophisticated viding enhanced security measures, and making other changes approaches that supplement the traditional metrics with in- that would reduce their waiting time at stops and make their formation on attitudes and preferences. An example of the waits more amenable. Attraction of non-riders will further new approach is a market segmentation study of TriMet's depend on effectively communicating the improvements that service area by Gilmore Research Group, first conducted in are made.

OCR for page 67
79 Market Segment Label Composition Demographic-Attitudinal Characteristics Attraction-Retention Strategy "Transit is a Lifestyle Choice" 43% of Riders Pro-bus & pro-TriMet Likely to respond to 28% of Non-riders See riding as convenient, economical, & service improvements 35% of Sample good for Portland's livability Newer to the area, well educated, & live in urban neighborhoods "I Use Transit When it 18% of Riders Demographics similar to the region's Likely to respond to Makes Sense" 14% of Non-riders No strong attitudinal barriers toward promotional marketing & 16% of Sample using transit service improvements Don't have a compelling reason to use transit more often "Riding the Bus Saves Money 10% of Riders Predominantly male & more Likely to respond to for My Family" 10% of Non-riders ethnically diverse service improvements, new 10% of Sample More children at home service, & more stop Transit is "a way to get around" amenities Places low value on transit's environmental & social benefits "I'm not Comfortable Riding 22% of Riders Predominantly female Likely to respond to the Bus" 29% of Non-riders Not comfortable around strangers security measures, more 26% of Sample Concerned about personal safety stop amenities, and Recognizes transit's environmental & service improvements that social benefits reduce wait time "There's no Way I'm Getting 7% of Riders Married, homeowner, high income, None on a Bus!" 19% of Non-riders longer term resident 13% of Sample Anti-transit & anti-TriMet If they ever ride, only use light rail Always prefer to drive, even in rush hour traffic Figure C-6. Portland region's transit market segments in 2004. If there is such a person as the "median citizen" in the Port- tem. Figure C-7 shows where the "Transit is a Lifestyle land region, he/she would be a member of the third largest Choice" market segment is concentrated. Service improve- market segment, labeled "I Use Transit When it Makes ments that target this group should be directed to routes serv- Sense." The demographics of this group are close to the re- ing areas where the group's concentration is highest. gional averages. They are not disinclined to use transit, but The effectiveness of the service improvements can be moni- neither are they so inclined. Promoting the advantages of tored by relating APC boarding and alighting data from stops transit use coupled with service improvements likely offers in the subject areas to overall passenger trends. It would not be the greatest prospect of providing this group a compelling surprising to observe ridership growing faster where the service reason to choose transit over other modes. improvements are implemented. Rather, the effectiveness of the The fourth largest market segment, labeled "There's no market segmentation analysis should be based on the extent to Way I'm Getting on a Bus," associates transit with everything which the observed rate of ridership increase in the target areas they see going wrong in the Portland region, and there is exceeds a benchmark rate representing the average system level likely to be little that can be done that would sway members ridership response to a change in the level of service. of this cluster from that opinion. They are relatively well off, In addition to the two research projects reviewed above, have lived in the area the longest, and are convinced that auto market research staff at TriMet draw on ITS data to facilitate is the only way for them to travel. their ongoing practices. APC ridership data provide the great- The smallest market segment, labeled "Riding the Bus est utility, and serve in the following applications: Saves Money for My Family," views transit as the most eco- nomical (and, for many, the only) option for traveling in the Stop level boardings and alightings are used for on-street region. Improvements in comfort and speed are changes that intercept surveys to determine the best times to survey, and would provide the most likely means of building ridership in when or where more than one surveyor is needed. this segment. Route level ridership data are used for on-board intercept Defining market segments provided TriMet with a basis for surveys to determine sampling rates, to determine how developing strategies to grow transit ridership in the Portland many surveys to print, and to determine weighting factors region in a cost effective way. The actual implementation of for expanding sample responses to population totals. strategies tailored to each market segment also depended on Ridership data are analyzed to assess the effects of changes knowing where each market segment was geographically con- to the system, including fare increases and service hours. centrated. Sampled market segments were inferred to zip Survey responses on satisfaction with "overcrowding" are code populations and mapped in relation to the transit sys- compared with passenger load data to identify specific

OCR for page 67
80 F11174-AppC-007.eps Figure C-7. Spatial distribution of the "Transit is a Lifestyle Choice" cluster group. circumstances where high passenger loads are affecting Market research staff have conducted Web satisfaction customer satisfaction. surveys. Pages or links with high hits that draw low survey sat- Survey vendors draw on stop and route passenger data to isfaction responses would be prime candidates for redesign. gain a better understanding of the dynamics of the system's The logs from the Trip Planner and Transit Tracker links operations. could be compared to demographic data from the origin- destination survey to identify where communication gaps In one instance, APC data have substituted for traditional exist between the agency and customer groups. These gaps practice. Prior to 2004, monthly ridership estimates were deter- could be targeted for additional promotion of available in- mined from an annual passenger fare survey, which provided the formation technology options or could be designated as areas mix and usage of fare instruments, along with monthly fare rev- that need traditional printed materials. enue receipts by type of fare (e.g., cash, pass). Changes in fares and fare instruments were increasingly complicating the esti- Issues, Observations, mates, and the traditional approach was replaced by a sampling and Challenges and estimation procedure that drew on archived APC data. Web and phone data are used by market research staff in There is substantial evidence of a progressive evolution in assessing the effectiveness of communications with customers TriMet's 10+ years experience with the recovery, archiving, and in evaluating customer communication with TriMet. and analysis of ITS data. It is noteworthy that when TriMet's Page hits on the Web proxy the visibility of messages to cus- AVL system was deployed, vendor-developed reporting soft- tomers. With a major reconstruction of the downtown transit ware did not exist. Staff developed the reports described in mall currently underway, a low number of hits on the con- this case study from structured queries of the event history struction and service information links would indicate that tables. However, as the other case studies have demonstrated, more promotion would be needed to increase the visibility of such data query skills proved essential in capturing the full this information. value of ITS data.

OCR for page 67
81 In realizing value from ITS data, TriMet has benefited from misbehavior. TriMet's data matching practices began prior to staff resources that compare favorably to those at the CTA and the deployment of AVL, when odometer and time-stamped Madison Metro Transit. An institution-level employment APC data were matched to schedules, and have evolved to the measure does not necessarily reflect IT and research staff current heuristic programs that incorporate a variety of issues levels; it does offer one explanation for why TriMet has expe- discovered over time. Fundamentally, successful data match- rienced success in utilizing AVL and APC data for market ing requires active collaboration between programmers and research and planning while other transit properties have staff familiar with scheduling data practices. often struggled to advance beyond standardized reports. Post processing of the raw data is an important step; there In the course of our interviews, we asked TriMet staff to re- are a number of anomalies in the raw data. But keep the flect on issues, challenges, and lessons learned from working raw data because some applications need it. with ITS data. The case study concludes with a summary of Staff responsible for post processing the AVL, APC and staff observations. event data have been a valuable resource to other staff in their efforts to analyze ITS data. Be sure to get detailed documentation from vendors. For Market Research example, staff was told that a data record was created when Market Research staff operate in a statistical package for the the bus door opens, but discovered 10 years later that a social sciences (SPSS) environment. Data cannot be weighted record is created when the bus enters the GPS stop circle. within GIS for expansion, so selected fields have to be first ex- Allowing unlimited access to ITS microdata can lead to ported to Excel, manually weighted, and then imported into misuse and misinterpretation. This advocates for granting a GIS. The process is time consuming and cumbersome. wide access to summary data and restricted access to indi- Market Research staff use SPSS; Operations staff use SAS; vidual data records. and the ITS data are in Oracle tables. The software envi- ronment supporting data analysis is not very user friendly. Information Technology and GIS Metadata describing the contents of the ITS data archive are lacking. There are only a few people who fully under- The TriMet system generates 36 million data records every stand the data archive. Understanding and finding the three months. The data are complex in that both spatial exact data one needs requires multiple requests, is time and temporal features must relate to each other. A rela- consuming and uses two people's time instead of one. tional database management system (RDBMS) provides Survey data could be geocoded and placed in the data the only way to manage this data in an effective way so staff archive so that it would link to ITS data and be available for can answer complex questions that cross department agency-wide use. However, there is a desire in market re- boundaries. search to maintain control over the availability of survey There is a tendency to have a single-minded focus in spec- data to avoid situations where the data are misunderstood ifying and deploying ITS technologies in transit. It is im- or misinterpreted. portant to engage IT staff early in the process to ensure compatibility with the enterprise data model and to allow for applications beyond the designed use of the technology. Operations/Service Planning/Scheduling Metadata for the data model and data tables is usually not Errors in data or interpretation can ruin trust. created because it is very time intensive. Consequently, most It has been a challenge aligning information from ITS data analysts don't fully know what data are available in the en- with what top management needs to make better decisions. terprise system and cannot take full advantage of it. A good When managers are used to such detailed information, the technical writer is needed to create data documentation. time and effort that it takes to create, maintain, and ana- With emerging applications like Google Transit, it is be- lyze the data begins to lose value. coming increasingly important for transit data to be publicly Data matching is essential; AVL and APC data are virtually accessible in a standard format. Without an Enterprise Data useless if they cannot be matched to the schedules. ITS ven- Environment, this can be difficult and complicated. dors have very limited capabilities in this area. The data Larger transit properties need a database administrator matching process is not simple; it must be fully automated to (DBA) to manage the enterprise data system. Smaller prop- handle the large volume of data records and it must provide erties also have enterprise data system management needs, sufficient feedback to allow "malfunctions" in the data to be which could potentially be met by contracting out or pool- detected. The process must be actively monitored because ing resources among properties to hire a single DBA. there are always new twists in how vehicles are operated, An enterprise RDBMS is needed to fully support GIS ap- including reroutes, operations control actions, and operator plications using ITS data.

OCR for page 67
82 References Strathman, J.G., Kimpel, T.J., and Callas, S. Rail APC Validation and Sampling for NTD and Internal Reporting at TriMet. In Trans- Crout, D.T. Accuracy and Precision of TriMet's Transit Tracker System. portation Research Record: Journal of the Transportation Research Presented at the 86th Annual Meeting of the Transportation Board, No. 1927, Transportation Research Board of the National Research Board, Washington, D.C., 2007. Academies, Washington, D.C., 2005, pp. 217222. El-Geneidy, A.M., Strathman, J.G., Kimpel, T.J., and Crout, D.T. Strathman, J.G., Kimpel, T.J., and Callas, S. Headway Deviation Effects The Effects of Bus Stop Consolidation on Passenger Activity and on Bus Passenger Loads: Analysis of TriMet's Archived AVL-APC Transit Operations. In Transportation Research Record: Journal of Data. Center for Urban Studies, Portland State University, Port- the Transportation Research Board, No. 1971, Transportation land, OR, 2003. Research Board of the National Academies, Washington, D.C., Strathman, J.G., and Hopper, J.L. An Evaluation of Automatic Passen- 2006, pp. 3241. ger Counters: Validation, Sampling, and Statistical Inference. In Gilmore Research Group. 2004 Market Segmentation Study. Final report Transportation Research Record 1308, TRB, National Research prepared for TriMet, Portland, OR, 2004. Council, Washington, D.C., 1991, pp. 6977. Kimpel, T.J., Strathman, J.G., Bertini, R.L., and Callas, S. Analysis of TranSystems, Planners Collaborative, Inc., and Tom Crikelair Associ- Transit Signal Priority Using Archived TriMet Bus Dispatch ates. TCRP Report 111: Elements Needed to Create High-Ridership System Data. In Transportation Research Record: Journal of the Transit Systems. Transportation Research Board, National Acade- Transportation Research Board, No. 1925, Transportation Re- mies, Washington, D.C., 2007. search Board of the National Academies, Washington, D.C., 2005, pp. 156166. Kimpel, T.J., Strathman, J.G., Callas, S., Griffin, D. and Gerhart, R.L. TriMet Staff Interviewed Automatic Passenger Counter Evaluation: Implications for National Transit Database Reporting. In Transportation Research Doug Allen, Computer Technology Specialist Record: Journal of the Transportation Research Board, No. 1835, Brett Baylor, Manager, Oracle Technology Transportation Research Board of the National Academies, Wash- Terry Bryll, Systems Analyst Programmer IV ington, D.C., 2003, pp. 93100. Steve Callas, Manager, Service and Performance Analysis Mohring, H., Schroeter, J., and Wiboonchutikula, P. The Values of David Crout, Planner/Analyst Waiting Time, Travel Time, and a Seat on the Bus. Rand Journal of Rex Fisher, Schedule Data Technician Economics, Vol.18, No. 1, 1987, pp. 4056. James Hergert, Manager, Service Planning Strathman, J.G., Dueker, K.J., Kimpel, T.J., Gerhart, R.L., Turner, K., Nancy Jarigese, Senior Financial Analyst Taylor, P. Callas, S., and Griffin, D. Service Reliability Impacts of Bibiana McHugh, IT GIS and Location Based Services Computer-Aided Dispatching and Automatic Vehicle Location Tim McHugh, Chief Technology Officer Technology: A TriMet Case Study. Transportation Quarterly, Vol. Ginger Shank, Senior Research Analyst 54, No. 3, 2000, pp. 85102. Tom Strader, Senior Research Analyst