Minutes Matter: A Bus Transit Service Reliability Guidebook (2020)

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.

43 5.1 Introduction This chapter provides information regarding Step 5 of a reliability improvement program: diagnostic assessment. It first presents an overview of a few broad elements of reliability before presenting a discussion of the many internal and external factors that have an impact on reliability. It then presents findings from the case studies regarding how agencies react to reliability issues as they are identified. The chapter concludes with a discussion on identifying the root causes of specific cases of unreliability so that reliability improvement treatments based on the understanding of the specific causes can be identified, as discussed in Chapter 6. 5.2 Causes of Unreliability It is one thing to measure reliability, but it is another thing to understand why a service is unreli- able so that actions can be targeted to address factors that can result in improved reliability. To understand why a service is less reliable, it is important to consider how transit service operates and to analyze how various factors affect bus operations. The operation of fixed-route bus transit service consists of a group of buses and operators leaving a depot, with each bus/operator operating a set of scheduled trips between route end- points while traversing a series of streets and stopping at a series of stops to pick up and discharge customers on each trip. Each of these elements can introduce unreliability experienced by customers when they do not occur per the promised schedule. • Non-Operation occurs when either the bus or the operator is unable or unavailable to leave the depot to provide the scheduled service. Non-operation means that customers are either left stranded or forced to wait for the next bus. The added wait time can vary from a few minutes for a short headway route to an hour or more for infrequent service. Non-operation of a scheduled trip can also occur if a bus is delayed or breaks down while in service, causing it to miss one or more scheduled trips. • Early or Late Start of a trip from the route endpoint affects the punctuality of service and can result in customers missing the bus or having to wait longer than expected. Late-starting buses, especially on high-frequency, high-ridership routes, will often be further delayed as they travel along the route as more customers will accumulate waiting for the late bus, and the extra time spent boarding the extra customers will delay the bus even further. • Inconsistent Travel Speeds can occur while traversing a route. Sometimes congested streets or other factors will introduce variability into bus travel times, which can delay customers and affect the punctuality of service. Operators have some control over the speed of the bus, but traffic signals, congestion, emergency vehicles, and weather can result in variable travel speeds between stops. C H A P T E R 5 Reliability Diagnostic Assessment

44 Minutes Matter: A Bus Transit Service Reliability Guidebook • Inconsistent Dwell Times at stops will also introduce variability into bus travel times, affect- ing punctuality. Dwell times are closely related to the number of customers boarding and alighting at a stop but are also affected by fare collection procedures, boarding and alighting policies, the level of crowding on the bus, and wheelchair and bicycle securement. • Inconsistent Transfer Times can greatly affect the reliability of a journey as experienced by a customer. While not directly observed as part of bus operations, the variability in the amount of time a customer spends connecting from one route to another can be caused by the vari- ability of bus travel time and punctuality of service. With lower-frequency routes, the transfer time can be highly variable if schedules are such that a small variation in running time can cause a connection to be missed. The measures discussed in Chapter 4 each capture different aspects of reliability, and each points to a different set of elements of unreliability that may be addressed to improve perfor- mance. Table 5.1 shows which elements of unreliability may be the cause of unreliability as observed by each of the measures identified in that chapter. 5.3 Factors Affecting Reliability Research performed for this project identified numerous factors as having an impact on reli- ability. Factors are categorized as being either internal or external, although some involve a combination of internal and external influences. Internal factors are those that are entirely or Measure with Poor Results Element of Unreliability N o n -O p er at io n E ar ly /L at e S ta rt V ar ia b le T ra ve l S p ee d V ar ia b le D w el l T im e In co n si st en t T ra n sf er T im es Terminal departure times ● Arrival times ● ● ● Bus running times ● ● Dwell times ● Customer travel time ● ● ● ● Buffer time ● ● ● ● Headways (terminal departure) ● Headways (mid-route) ● ● ● Customer wait times ● ● ● ● Missed service ● Service disruptions ● Customer survey data ● ● ● ● ● Table 5.1. Relationship of reliability measures to elements of unreliability.

Reliability Diagnostic Assessment 45 primarily under the control of the transit agency and can, therefore, be directly modified to enhance reliability. External factors are not under the direct control of the transit agency and, therefore, can only be mitigated by agency actions or only be directly affected by partnering with other organizations. This distinction is important to note because it may be easier to first address the causes of unreliability under agency control before attempting to mitigate factors that cannot be directly controlled. However, external factors should not be ignored since some of the most effective treatments to address unreliability may be those that, when implemented in cooperation with regional partners, can influence or mitigate the external factors. 5.3.1 Internal Factors Internal factors are those identified as being entirely or primarily under the control of the transit agency. Internal factors can be classified as relating to (1) operations and maintenance staff, (2) service design and scheduling, (3) the bus type and fleet maintenance, and (4) fare payment. Operations and Maintenance Staff There is no single factor more critical to bus service reliability than the bus operator. If operators feel that their work is appreciated by agency management and customers, the bus fleet is in good mechanical condition, and schedules are realistic, it is far more likely that the strategies described in this guidebook will be successful. The behavior of the bus operator, which is based on experi- ence and hours worked per week, has been shown to have a significant impact on bus service reliability. The more years of experience for a bus operator, the more likely the operator is to drive in a dependable, on-time manner. Furthermore, research has shown that early arrivals are more likely during the first 2 weeks of a new bus operator sign-up [16]. Another study found that working overtime hours is not positively related to absenteeism. On the contrary, operators who worked overtime had a lower rate of absence [17]. Bus operators’ lifestyles at home and at work play a critical role in their physical and psycho- logical well-being as well as their job performance. Stress factors such as traffic, violent customers, and tight running schedules may have negative impacts on bus operators, which can result in absenteeism, vehicle crashes, and other factors that may affect bus service reliability [18]. Now that AVL/APC systems and real-time dispatching technologies are increasingly available, transit agencies can better monitor and make adjustments to reduce the impacts of bus operator behavior on bus service reliability. In addition to the bus operator, other staff can also have an impact on reliability. Maintenance workers affect the availability of buses to provide the service and can have an impact on the number of breakdowns and road calls. Field supervisors and control center staff can directly affect reliability through the actions they take to adjust service when a problem arises. Short-staffing, high attrition, low morale, or absenteeism in any of these positions can affect reliability. Service Design and Scheduling Research has shown that variability in time associated with making a transfer, which is directly influenced by frequency of service, appears to have a significant impact on transit service reli- ability [19]. Grid networks were found to be less disrupted by transfers than radial networks were, but service reliability was found to be more influenced by frequency of service than by route density. On-time arrival of vehicles at major transfer stations was also found to be important, especially for radial networks. The number, spacing, and utilization of bus stops also affects the variability of running time. Closely spaced stops that are often lightly used can result in a wide variation in the number of stops made on any given trip, leading to high variability in running times. Having fewer stops, but stops that are used consistently on every trip, will provide more consistency in running times.

46 Minutes Matter: A Bus Transit Service Reliability Guidebook The probability of late arrivals has been found to increase as buses travel along their routes, and there is a trade-off between running times and reliability when scheduling buses. Adding slack time to a schedule helps to improve reliability but can lower operating speed, which can lead to other negative consequences. Transit planners can increase the cycle time by adding travel time, recovery time, or both. Keeping travel time constant but increasing recovery time keeps vehicles moving on route, and increased recovery time serves as a reward for the bus operator for on-time running. Recovery time, however, does little to address on-route reliability other than stopping the progression of delays past the control point. The effects of late depar- tures from the origin terminal are illustrated in Figure 5.1. The number of stops on a route can also have an impact on reliability. With more stops, espe- cially with many that are less frequently used, the number of stops made on any one trip can vary significantly, which can introduce variability in running times. Routes with fewer stops, but ones that are used on nearly every trip, will experience less variability in running time. Bus Type and Fleet Maintenance Factors related to the design of the bus itself, such as its floor height, vehicle length, and bus type, have been demonstrated to have a significant effect on dwell times and thus overall bus service reliability. One study found that while articulated buses increase running times, they helped reduce running time variation due to their greater passenger capacity and more consistent boarding times [21]. The condition and availability of the bus fleet can also have an impact on reliability. Buses that are constantly needing repairs may not be available for service, resulting in fewer buses than are called for in the schedule being put into service and scheduled trips being missed. Buses that experience mechanical problems during service can also be a cause of missed service. Fare Payment Several studies have found that fare collection and payment methods have a significant rela- tionship with bus service reliability. Use of contactless fare cards, smart cards, flash passes, and off-board fare collection generally help to expedite the boarding process compared to use of magnetic strip cards or cash fare payment, thereby reducing dwell times. Fare payment policies can also have a significant impact on bus service reliability since they can be a major determinant of dwell times at stops. With off-board fare collection, agencies can adopt all-door-boarding policies. Faster, more efficient, and consistent forms of fare payment are preferable from a service reliability perspective. An agency may opt to eliminate cash fare payment from its system as a policy to improve reliability while also reducing the cost of fare collection for the transit agency. In locations where there is heavy boarding at congested downtown stops in the PM peak period, some agencies have adopted a policy where customers pay when they exit the bus. Source: Cham 2006 [20] Figure 5.1. Effects of late departures from the origin terminal on bus service reliability.

Reliability Diagnostic Assessment 47 5.3.2 External Factors External factors are those that are not under the direct control of transit agencies; they relate to concerns such as roadway infrastructure, unpredictable customer behavior, traffic conges- tion, and other random events. While transit agencies cannot themselves control external factors leading to unreliability, they need to understand, anticipate, and monitor these factors and adapt to them. By partnering with other entities such as city and regional traffic management agencies and highway departments, transit agencies can often help to bring about changes to external factors that can improve reliability. External factors can be classified as being related to (1) traffic congestion; (2) traffic signals; (3) temporal factors such as weather, detours, and incidents; and (4) customer activity, including uneven or unpredictable boardings and alightings. Traffic Congestion Studies have found that traffic conditions are a significant factor in bus service reliability and travel times. Heavy traffic tends to have a negative impact on reliability and travel times in the absence of priority treatments. Time of travel, direction, and location, such as in the central business district, all affect traffic conditions and thus bus service reliability. Traffic Signals Traffic control devices, especially traffic signals, have been shown to have an impact on bus service reliability. The number of traffic signals and the signal timing, signal progression, and orientation of bus stops to signalized intersections all affect the variability of running times on bus services. Long signal-cycle times can introduce long delays if buses arrive just as a signal turns red. Signal progression timings designed to speed cars through an area can introduce delays for buses that stop to pick up passengers and drop out of the progression pattern. The placement of bus stops before traffic signals can result in buses missing a green phase and dwelling through an entire red phase. TSP treatments (both active and passive) implemented through cooperation with local traffic control agencies can address many of these concerns. Temporal Factors Time-related and seasonal factors have also proven to be significant influencers of bus service reliability, especially as they relate to peak travel periods, time of day, and seasonal travel patterns. Peak-period travel has been estimated in some research studies to have a negative impact on bus service reliability in comparison to off-peak times, but not in every case. The perception of reliability is also affected by the same temporal conditions that cause disruptions in operations. Inclement weather conditions, such as precipitation, have been shown to have a significant negative impact on bus service reliability. Among the reasons that snow makes bus service unreli- able is slower traffic due to slippery roads and the need for customers to negotiate around snow- banks at bus stops. Some transit agencies in regions where heavy snow can be expected maintain snow emergency plans that dictate when service on steep hills will be abandoned and when all service will be suspended. Some large-city transit agencies have their own snow-fighting equip- ment to keep critical links open and steep hills passable. The issue of shoveling bus stops and shelters, however, is a challenging one. A much greater impact can be expected when, for example, it snows in the southern portion of the United States, where transit and highway agencies are not equipped for the event. Customer Activity Customer activity, including boardings, alightings, loads, and customer demand or rider- ship, have been shown to significantly influence bus service reliability. Evidence indicates that

48 Minutes Matter: A Bus Transit Service Reliability Guidebook increased customer demand and activity often correspond to declining service reliability, due largely to increases and fluctuations in dwell times. Research has shown statistical correlations between customer load and headway variation [22] and travel time variation [23]. Headway variation is often cited as a primary cause for customer overloads. Another important factor influencing reliability is the use of wheelchair lifts or ramps and bike racks, which relate to bus characteristics as well as the proportion of customers using these devices. Low-floor buses, which use ramps rather than lifts, minimize dwell time impacts from customers in wheelchairs. When considering both passengers with and without disabilities, a low-floor bus can be expected to reduce dwell time by approximately half a second to 1 second per boarding customer [6]. 5.3.3 Factors and Elements of Unreliability Each of the factors discussed previously influences different elements of unreliability. Table 5.1 groups the various reliability measures by the element of unreliability the measure influences. By using a combination of Table 5.1 and Table 5.2, it is possible to identify possible factors affecting reliability for a given measure of poor reliability. For example, if mid-route headways are too variable, Table 5.1 says that start times, travel speeds, and dwell times should all be considered, and Table 5.2 provides a list of factors to examine. 5.4 How Transit Agencies React to Reliability Problems There is little information in the literature on reliability that illustrates how transit agencies react to reliability problems, especially on how they choose which strategies to use to address these problems. The 10 case studies completed for this research explored the actions both that agencies took to address reliability in real time and to understand whether there were ongoing issues that needed to be addressed to improve a specific instance of unreliability. The case studies showed that most agencies monitored their bus service in real time using CAD/AVL data, and most acted in real time as reliability issues developed. However, most lacked explicit rules that triggered an immediate response; rather, supervisors relied on their experience to identify and address reliability issues. Throughout the day, supervisors viewed activity regu- larly and received anecdotal information from operators through their daily logs. Interaction with agency dispatch centers allowed for supervisors to monitor buses in real time for increased insights into potential causes of reliability issues. Once reliability data were collected, most agencies used their service planning department to analyze available CAD/AVL and APC data, although in many places, the IT or operations departments were heavily involved. The service planning department evaluated the APC data from the route and operator perspective to track individual performance and patterns as they developed. Some agencies evaluated customer, operator, contractor, and social media input to evaluate the need for a running-time analysis on routes, and often they verified such input using road supervisors to perform manual on-time performance checks along routes and at stations. At most agencies, the first step to address a persistent recurring reliability problem was to conduct a running-time analysis. Running-time data were evaluated, and schedules were updated to better reflect actual running times. However, schedule adjustments were typically only imple- mented on the three or four occasions per year when drivers would pick new work assignments. When evaluating route reliability data, many of the larger transit agencies focused efforts on higher-frequency routes that had headways of 15 minutes or less. Since those are the routes most

Reliability Diagnostic Assessment 49 Reliability Elements and Factors Internal/External Factor Category Non-Operation Operator availability Internal Staff Vehicle availability Internal Fleet Breakdowns Internal Fleet Early/Late Start Insufficient recovery time Internal Planning Operator restroom breaks Internal Staff Holds for late connections Internal Service Poor operational control Internal Staff Mechanical issue Internal Fleet Variable Travel Speed Insufficient/excess scheduled time Internal Planning Too few/too many time points Internal Planning Overly long route Internal Planning Lack of adherence to time points Internal Staff Operator skill/behavior Mixed Staff Delays merging into traffic from stops Mixed Traffic Incidents, special events, construction Mixed Temporal Traffic congestion External Traffic Signal delay External Traffic Weather External Temporal Variable Dwell Times Too many stops/poorly located stops Internal Planning Poor transfer connections Internal Planning Uneven loading due to variable headway Internal Service Demand in excess of capacity Internal Customer Variable passenger demand Mixed Temporal Fare payment delays Mixed Service Access for cyclists Mixed Customer Access for mobility impaired Mixed Customer Inconsistent Transfer Times Insufficient recovery time Internal Planning Poor schedule coordination Internal Planning Poor route connectivity Internal Planning Table 5.2. Relationship of reliability measures to elements of unreliability.

50 Minutes Matter: A Bus Transit Service Reliability Guidebook dependent on reliable fast service, it is important that the transit agency guarantee performance and address any route issues. While most of the case study agencies actively addressed unreliability in real time, none described a formal process for diagnosing specific causes of unreliability; rather situations were generally addressed in real time using enhanced supervision and applying operational strategies (such as inserting spare buses, holding buses, deadheading, and introducing short turns). If situ- ations persisted over time, schedules were changed to provide enough running time to absorb the more extreme cases of late buses so that delays did not propagate through the schedule. This is not to say that other strategies to improve reliability were never enacted, just that improvements were often made through a more informal process to go from analysis of reliability measure data to diagnosing causes of unreliability, which could in turn lead to improvement strategies specifically targeting the identified causes. 5.5 An Approach to Identifying Causes of Unreliability In recent years, transit agencies have become accustomed to using real-time data generated from AVL/APC systems to monitor their operations in real time. Many have also begun to mine those data to develop better schedules and to develop key performance indicators, including reliability measures. Measuring and reporting reliability performance are the first step in improving reliability. But to make effective improvements in reliability, agencies need to understand why a service is falling below its reliability targets. Are the causes internal, external, or a combination? Which elements of unreliability are the problem? What are the root causes? The order in which the five elements are listed in Section 5.2 and Table 5.2 provides a logical sequence for examining causes of unreliability and developing improvement strategies. Use of this sequence begins with examining reasons why service fails to operate, then examining why service does not start on time, and then examining why service fails to stay on time. Use of this sequence allows agencies to look first at internal factors under their control and then move to factors where they have little to no control and may need to collaborate with other agencies to develop improvements. An approach to using this sequence is described in the following: • Are there enough buses and operators available to provide the scheduled service? Non- operation of scheduled service not only results in customers left behind, it can also lead to reliability problems on the services that do operate. Agencies should first consider whether employee absences (including operators and maintenance and supervision personnel) or vehicle maintenance issues are causing an insufficient number of buses to pull out of the depot each day. Providing the number of operators and buses needed to operate the schedule is the first step in providing reliable service. Also, are road calls reducing the number of buses on the road throughout the day? • Are vehicles/operators available to start each trip on time? Starting each scheduled trip on time is the next step in providing reliable service. Operators are not able to start a trip on time if they have not been able to complete the preceding trip on time so as to be at the route terminal ready to begin the next trip. Schedules must be able to accommodate the normal variability of running time such that the agency’s on-time performance target for terminal departure can be met. In addition, schedules need to allow time for operator restroom breaks at terminals. • Are operators starting each trip on time? It is not enough for vehicles and operators to be available to start each trip on time; they must do so. Once schedules adequately reflect the amount of running time and layover needed, operators and supervisors must ensure that buses depart on time so that late departures do not cause further delays due to crowding resulting from the initial and subsequent delays, potentially resulting in bus bunching.

Reliability Diagnostic Assessment 51 • Are operators able to meet scheduled time points? If all the scheduled buses and operators are in service, and all trips have left the route endpoint on time but reliability at down- stream time points is still a problem, agencies should first look for internal causes within their control. These could include inaccurate scheduled running times between time points; operator performance; lack of supervision; or complex, overly long route design. Many other possible causes of unreliability may be only partially or not at all within the agency’s control. In this case, additional investigations with partner agencies may be needed to identify what improvements might be most effective. Following this process in conjunction with existing practices to monitor and adjust service in real time should lead to reliability improvements that will enable existing real-time management practices to more effectively control service. 5.6 Analysis Framework for Identifying Causes of Unreliability Following the approach outlined in the previous section, this section proposes a step-by-step framework for transit agencies to use to identify causes of unreliability. Each step is geared toward identifying the actions to be taken to assess potential causes of unreliability. While most steps involve analyzing bus operations data that may already be available, new and more detailed analysis of data often may be required, and in some instances, new focused data collection may be needed. Throughout the process, however, data analysis should be complemented by communication with operators, supervisors, dispatchers, and maintenance staff who have direct and detailed knowledge of how service is being provided and can help focus the effort to identify causes of unreliability as well as possible reliability improvement treatments. The steps outlined in this section are best considered in the order shown. By eliminating the possible causes addressed in the earlier steps, agencies will be better able to determine the true root causes of observed unreliability. By understanding the root causes, agencies will be better able to identify the most effective treatments and ensure success in improving reliability. 5.6.1 Sufficient Vehicles in Service For service to be provided that is reliable and meets the promised schedule, an agency needs to ensure that enough vehicles are leaving the depot to provide the service and that the required number of vehicles remain in service throughout the day. If the required resources are not avail- able, an agency cannot expect to operate service reliably. The two main causes of an insufficient number of buses in service are operator availability and vehicle availability. Transit agencies keep dispatch logs that track all scheduled pullouts from the bus depot and whether any runs were missed. Agencies should review dispatch processes to ensure that these logs identify why a particular pullout was missed or late. At a minimum, the logs should identify whether it was a lack of a driver or of a vehicle. The three key factors to investigate when considering operator availability are (1) operator recruitment and hiring, (2) absence rates, and (3) the size of the extra board. The scheduling and run-cutting process at an agency determines the number of operators needed on a given day to provide the service, but determining the number of operators needed on the payroll and the number scheduled to work each day is a more complex process. If operator availability is a problem, agencies should first understand whether all positions are filled and whether operator recruitment and training have been a problem or whether retiring of operators is playing a role. Next, agencies should consult attendance records and evaluate their rate of absenteeism and how that relates to the size of the extra board scheduled each day. If absence rates are high due to

52 Minutes Matter: A Bus Transit Service Reliability Guidebook sick or vacation days, then the focus may need to be on strategies to lower that rate. If rates are acceptable, then the size or scheduling of the extra board may need adjustment. Key factors in vehicle availability are more complex; vehicle age and condition, scheduled maintenance procedures, and unplanned repairs play a role, as do maintenance staffing levels and absences. The size of the fleet compared to the number of vehicles needed to provide the service, or the spare ratio, also contributes to whether enough vehicles are available to provide the service. If vehicle availability is contributing to missed pullouts, or breakdowns are causing trips to be missed, then agencies will need to conduct a thorough analysis of maintenance practices to identify specific causes of missed service. 5.6.2 Arrival at Trip End One of the most difficult aspects of identifying causes of unreliability is that the effects of one factor can cause a cascading effect on reliability. Just within a given trip, a delay caused by one factor can result in longer headway from the previous vehicle, which could result in greater boarding/alighting variability, which in turn could cause further delay. If the vehicle arrives at the end of the route too late to start its next trip, the next trip could become delayed, and reliability would suffer on that trip and on subsequent trips by that vehicle as well. Therefore, it is critical to separate the cascading effects on reliability caused by previously encountered factors from newly encountered factors affecting reliability. To identify whether previous delays are affecting reliability on subsequent trips, agencies should first examine whether buses are arriving at the route endpoint in time to start their return trips. Until recently, agencies had to rely on supervisor logs or data from traffic checkers for this information, and these data were often less complete because supervisors could be distracted by other duties when reliability was suffering and agencies eliminated traffic checkers due to budget considerations. Now AVL/APC systems are providing more complete information at route end- points that can be accessed to determine whether vehicles are ending their trips before their next trip is scheduled to start. To examine this, AVL/APC reports are needed that compare arrival times to scheduled departure time at route endpoints in addition to just the scheduled arrival time. If buses are almost always arriving in time for their next trip, then agencies can skip to examining vehicle departures at the trip starting point. If buses frequently arrive too late to start their next trip on time, agencies may need to consider adjusting schedules to eliminate the cascading effects of unreliability. The transit industry has typically addressed this issue by adjusting running times and sched- uling recovery time at the route endpoints so that a late-arriving vehicle is still able to start its next trip on time. While some scheduled recovery time will always be necessary given that some variability in running times should be expected for a bus service operating in mixed traffic, long scheduled recovery times may be masking reliability problems along the route. Nevertheless, agencies should consider whether insufficient recovery time is contributing to reliability prob- lems and causing delays to cascade throughout the day. If running times and recovery times can be adjusted to ensure that nearly all trips at least start on time, then other factors causing poor reliability will be easier to identify and treat. It may be possible that treating those factors may then improve running times and reduce the need for recovery time, but without adequate run- ning and recovery times initially, it may be difficult to identify those root causes. Most agencies typically complete some form of a running-time analysis when reliability issues have been identified. Historically, these consisted of manual running-time checks, operator and supervisor interviews, and, often, test runs. More recently, the wealth of data available from AVL/APC systems has provided much more robust samples of running-time data from which statistically valid measures of reliability can be calculated to supplement the more qualitative, but often valuable, observations by operators and supervisors.

Reliability Diagnostic Assessment 53 While there are several different methods for determining how much time should be allo- cated to running time versus recovery time, agencies can adopt a simple rule of thumb that buses should arrive at the end of a route before their next scheduled departure time a specified percent of the time, with that percent based on their on-time performance standard for trip starting point departures. If that percentage is, say, 95 percent, then running time plus recovery time should be set at or above the 95th percentile running time. If operator break time must be included (i.e., if operators are guaranteed a specific break time, regardless of arrival time), then that break time should be added to 95th percentile running time. With the wealth of data provided by AVL/APC systems, the calculation of the 95th percentile can be completed for each individual scheduled trip and adjustments made to running times and recovery times at a much finer level than had been previously possible. Providing an adequate amount of running time plus recovery time uses scheduling tech- niques to reduce or eliminate the cascading effects of unreliability. It may also reduce or elimi- nate unreliability if tight schedules were causing unreliability by not adequately accounting for normal variability in running times. If unreliability persists, however, agencies will still need to go further to identify factors causing the unreliability, which could be the result of internal operational factors or external factors more difficult to control. 5.6.3 Departure from Trip Starting Point Once the scheduled running time plus recovery time is adjusted to allow the clear majority of vehicles to arrive before their next scheduled departures, agencies should examine whether vehicles are indeed leaving the trip starting point on time. Trips that depart the starting point early or late can quickly lead to bus bunching, uneven loads, and unreliable service, but these types of trips are one of the easiest causes to correct if the vehicle has arrived on time. AVL/APC data and reports, supplemented by operator and supervisor reports and interviews to determine causes, would again be the primary sources for analysis of whether departures are on time. Some of the factors causing late departures, even when vehicles arrived on time, could result from delays that occur at the end of the route. These include holds for late connections, operator restroom breaks, and mechanical issues. Supervisors and operators would be the best sources to use for understanding whether holds for connections or operator restroom breaks are causing late departures. If holds for connections are causing late departures, agencies may need to examine scheduled connection times, the proximity of bus stops or berths for connecting routes, and poli- cies such as only holding buses on routes or at times when headways exceed a specified duration. Operator restroom breaks are a necessity that is often overlooked by those not involved in bus operations. They do not usually occur every time a bus reaches the endpoint of a route, but rest- room facilities poorly located away from layover points can have a direct impact on vehicle depar- ture times. Frequent mechanical issues may be identified by operators, but maintenance records would need to be consulted to determine the specific causes of mechanical issues. Poor on-time performance leaving the starting point on a route may simply be the result of operator behavior, overall poor operational control on the part of supervisors, insufficient layover time, operators not trained on urgency to always leave exactly on time, or heavy passenger loading volumes. Early departures are a clear indication of lack of control and can result in customers missing the bus and overloading and delaying the following bus, causing bus bunching and poor reliability. Late departures not resulting from any of these causes may simply result from a lack of attentiveness to the schedule on the part of operators or supervisors. Running-time analysis is easily obscured when operators do not consistently leave on time. Agencies should consider iden- tifying patterns of late or early departures by comparing the timeliness of departures to operator and supervisor schedules to identify if issues are resulting from specific operators or supervisors.

54 Minutes Matter: A Bus Transit Service Reliability Guidebook 5.6.4 Scheduling Adjustments at Time Points Along the Route While many of the factors affecting on-time departures at the beginning of a route are internal factors, the factors affecting reliability at bus stops and time points along a route can be a mixture of internal and external factors that can be much more difficult to separate from each other. Thus, causes of unreliability may be a combination of factors, and effective ways of addressing them may involve a combination of treatments, some of which may require involvement from other agencies. Identifying causes of unreliability along a route can be difficult if vehicles are not departing the starting point on time or if not all scheduled service is being provided. Therefore, it is important to try to address the issues covered in the preceding sections prior to addressing reliability along the route. In examining on-time performance along a route, agencies may address either schedule adher- ence (in the case of longer headway, scheduled-based services), or headway adherence (in the case of shorter headway, headway-based services). Agencies may also assess running times between time points and compare the variability of running times or the percentage of trips shorter or longer than the scheduled running time by more than a set percentage. In any of these cases, AVL/APC data are essential to obtaining a sufficiently large sample of trips to perform an analysis. Agencies will need to ensure that their AVL/APC systems can produce reports examining both on-time performance at time points and running-time variability between adjacent time points. Reports from APC data that can identify the number of stops made and that can generate reports of dwell time (door open time) at stops, both per trip and per stop, can also be useful. Agencies should begin with developing an understanding of reliability using AVL data and, if necessary, delve deeper into possible causes of unreliability using APC and AFC data. For each time point along a route, agencies should first use AVL data to make sure that sched- uled running times are appropriate and adapt schedules as needed to match real-world running times. Providing operators with realistic, achievable schedules can improve their performance and address at least some of the lack of reliability. However, such measures merely adapt schedules to factors that cause unreliability while not actually mitigating those factors. Therefore, once schedules are adjusted, agencies should delve deeper into the remaining unreliability to identify causes and solutions that could mitigate the causes and lead to significant improvements in reliability beyond what traditional scheduling adjustments can achieve. A first step in using AVL data to improve schedules on an unreliable route is to determine whether trips that are not on time are largely early, late, or both (or that running times are too short, too long, or both). If sufficient data are available—that is, if there are many days of data for each scheduled trip—a detailed analysis of running-time data can be done using individual observa- tions of each trip on each date in the analysis period. Analyzing a sufficiently large sample of trips on each scheduled trip individually may lead to a better understanding of the time intervals when run- ning times are consistent and so may lead to redefinition of the periods used to set running times. If there are many early trips and few or no late trips, then excess scheduled running time could be the cause. If there are many late trips and few or no early trips, then it could be insufficient scheduled running time. However, in the case of late trips, the insufficient time could be on any of the preceding route segments, not just the immediately preceding segment, so agencies should examine all time points along the route to identify any sources of delay. Lateness could also be the result of a late departure from the route starting point. Agencies should also consider too many time points as a possible cause of lateness if time points are less than 5 to 10 minutes apart, which can cause operators to hold too often, delaying service. If there is no clear tendency toward early or late trips, but on-time performance is poor, agencies will need to examine the variability of running times. If variability is reasonably low (say a coefficient variation of 0.15 or less), but many trips fall outside the on-time performance

Reliability Diagnostic Assessment 55 window, then there may be too few time points on the route, and adding time points might help operators stay closer to the schedule. The route may also be an overly long route that is difficult to manage reliably due to the cumulative effects of factors causing variations in running time. With the significant quantities of AVL data that are typically available, AVL data can also be merged with other observed data to search for relationships that may explain variations in run- ning time or on-time performance. If individual trips can be tagged with the operator run number or operator identification, then it may be possible to determine if long or short running times are due to individual operator skill or behavior or just a general lack of adherence to time points. Agencies should also consider keeping track of incidents, special events, and construction activities, as well as severe weather days, and examine the extent to which longer running times are the result of such factors. 5.6.5 Identifying Causes of Unreliability Along the Route The factors affecting reliability along a route that are most difficult to address are mostly related to (1) the number of passengers carried, (2) number of stops made, (3) traffic signals and congestion, or (4) restricted movement in and out of bus stops. Identifying these factors will help agencies identify effective treatments to address unreliability. Quantifying the impacts of these factors will aid in cooperatively developing solutions with the agencies that control the streets, traffic signals, and sidewalks used by buses and bus riders. While AVL data can provide large samples of data to analyze reliability, identify areas of poor reliability, and provide the information needed to fine-tune schedules to adapt to causes of unreliability, AVL data have limited capability to identify most factors causing unreliability. Transit agencies are increasingly implementing APC systems to gather detailed data on stopping patterns and ridership by stop. AFC systems can also be used to provide ridership information for analysis of factors affecting reliability. APC and AFC systems can be used to better understand how ridership might affect reliability along a route. APC systems provide detailed stop-level ridership and passenger loads as well as “door-open” and total dwell times by stop; however, many agencies do not have every bus equipped with APCs, so data on every trip may not be available. AFC systems can provide trip- level boardings by fare type on every trip but lack stop-level detail unless substantial post-processing is done to connect AFC records to location from an AVL system. AFC systems also typically lack data on passenger alightings, and therefore load, along the route, unless work has been done to post-process origin/destination pairs to impute alightings at the stop level. If the variability of running time on a route or route segment is high, the agency should do a more in-depth investigation of running time data that may involve ad hoc analysis of AVL/APC data beyond standard reports if reports have not been developed for this purpose. Investigations should also include conversations with field supervisors, control center staff, and (possibly) operators to better understand issues that are being encountered in the field. Number of Passengers While excessive passenger loads on an individual trip may be a symptom rather than a cause of unreliability, APC and AFC data can be used to assess whether overall passenger demand is causing reliability problems or whether reliability problems are resulting in uneven passenger loadings. If APC data are not available on all trips, the data can be supplemented with AFC data to estimate total passenger loads over a period. If the average load over a period is sufficient to affect travel time, then overall demand in excess of capacity may be the root cause of unreliability. In this case, the scheduled supply of service is likely inadequate, and efforts to address reliability may be ineffective until additional service can be provided.

56 Minutes Matter: A Bus Transit Service Reliability Guidebook Even if there is enough capacity on a route, variability caused by other factors can be made worse by the bunching of buses and the uneven loading that results. Reducing the time per passenger it takes for passengers to board will help speed up late, over-crowded buses. APC data can be used to identify factors that affect the amount of time it takes to board and discharge passengers at a stop. APC data can include data on the amount of time the doors are open at each stop, the number of boardings, the number of alightings, and the passenger load on board the bus. Typi- cally, the processing time for each boarding and each alighting is a function of the bus design (floor height and the number and size of doors), fare collection system, and level of crowding on the bus. Using APC data, an analyst could develop the relationship between the number of boardings and alightings and the load on the bus and determine the boarding time per passenger and the point at which passenger loads result in enough standees to restrict boarding and alighting and slow down service. Armed with an understanding of passenger boarding and alighting times, agencies could consider strategies to reduce passenger service time, such as all-door boarding, off-board fare collection, low-floor buses, wider doorways, faster fare-collection techniques, and techniques to speed boarding for cyclists and mobility-impaired individuals. An analysis could also be done on a stop-by-stop basis, either for one route or combing together all routes serving a stop, to identify the stops with the longest delays that could be subject to targeted improvements in fare collection, boarding policy, or stop layout and design. Number of Stops Made The number of times a bus must stop to pick up or discharge passengers has a significant impact on the variability of travel time, and ridership has a significant impact on the number of stops made. However, if there are many closely spaced stops on a route and those stops tend to be lightly used, a high-ridership trip will make many more stops than a low-ridership trip. If there are fewer but more widely spaced stops, those stops will tend to be used on nearly every trip, and the number of stops made will vary less with ridership fluctuations, reducing the variability of running time and increasing reliability. APC data can be used to examine the number of times a bus stops on each trip. If the number of times a bus stops varies considerably, then an agency should consider whether there may be too many stops causing running times to vary more and make service less reliable. If stop spacing is less than every 800 to 1,200 feet, stop consolidation may be appropriate. Traffic Signals and Congestion Traffic signals have an impact on travel speeds for all traffic, but they often have a greater impact on bus speeds because signal timing and progressions are not normally designed to account for the slower average speed of buses and because buses most stop frequently to serve passengers. AVL data can be used to assess overall travel time. APC data can be used to assess travel time in motion by subtracting time spent at stops (typically door-open or dwell time plus about 10 seconds per stop to account for acceleration and deceleration) to filter out the impacts of passenger boarding activity and focus on travel time between stops. Low average speeds for the type of roadway could indicate that intersection-based transit priority measures, or possibly exclusive lanes, could be beneficial. Intersection-based priority could include queue-jump lanes or TSP. Agencies will need to develop justification for such measures through use of passenger volume data developed through APC and AFC systems, as well as travel time analyses, to show the benefits of such improvements versus the impacts on other roadway users and project costs. Traffic counts to determine current traffic volumes and level of service will also be needed to evaluate the feasibility of any proposed improvements and to compare the impacts on a per- person (rather than per-vehicle) basis on each mode. Movement In and Out of Bus Stops One of the more difficult things to measure, but a significant source of delay and variability, is the time spent moving in and out of bus stops. The process of serving a bus stop involves

Reliability Diagnostic Assessment 57 deceleration, entry into the stop, waiting for passengers to move to the bus (for larger stops), passenger service time (dwell or door-open time), signal delay (for near-side stops), time waiting to merge into traffic, and acceleration time. APC data typically report only door-open time at the stop, and AVL/APC systems may lack the capability to determine locations at a fine enough level to measure all elements of this process. Supplemental data collection with handheld GPS devices or portable “geologger” units temporarily attached to buses can provide more fine- grained locational data. Such data can be analyzed to identify the total time that a bus is within a defined radius around a stop. However, in many cases, manual observation may be needed to determine exact causes of delay that are not related to passenger volumes. One method that has been used by agencies is to conduct a manual time-and-delay study. In such a study, observers ride a bus and observe and record causes of delays, as well as ridership levels. Such studies now can be done using a handheld device or cell phone with GPS capabilities to timestamp and location-stamp each manual observation as well as collect detailed second- by-second location information. With such studies, sample sizes are much smaller than from analyses of AVL/APC/AFC data, but the new observations can supplement those analyses and help pinpoint specific locations and causes of delays. One thing that should be considered during observations is whether buses are delayed entering the stop; this could be the result of queued traffic at a signal blocking buses from entering the stop, in which case an agency could consider a queue-jump lane or relocating the stop to the far side of the intersection. If the stop itself is blocked by another bus, the stop may be too small. If other vehicles are blocking the stop, then greater enforcement may be sufficient, although in many cases an analysis coordinated by the municipality may be needed to determine an appropriate allocation of curb space along the roadway. If buses are delayed waiting for a green light, then stop relocation, a queue jump, or signal retiming may be beneficial. If traffic volumes prevent buses from entering traffic, then strategies to hold traffic could be considered, such as yield-to-bus laws or allowing the bus to stop in the traffic lane with passengers boarding from a curb extension. Applying this framework for identifying causes of observed unreliability is an important part of Step 5 of a reliability improvement program. This framework should assist agencies in devel- oping a clearer understanding of the likely causes of unreliability on a particular service and can help guide agencies in the selection of the most effective treatments for unreliability in Step 6 of the program.