All opinions in this post are solely my own, and do not represent the positions of my employer or any organizations of which I am part.
About two months ago, I found myself waiting for a 77 bus in a Boston rainshower. Like billions of others on this planet, I do not own a car, and rely on mass transit for travel. That day, transit service failed to meet my needs: despite the 77’s advertisement as having frequent service throughout the day, the next bus appeared to be over 30 minutes away. So, much to my chagrin, my bus trip ended up becoming a soggy walk through to a nearby train station.
The point of recounting this story is not to highlight my own inconvenience—my walk was not particularly far and I had no urgent constraint binding my arrival time. Rather, the significance of this story lies in the fact that my experience was deeply unremarkable. Every day, millions of people the world over board buses, and every day, millions of them are disappointed. A 2019 survey of American bus riders found that reliability was second only to frequency in its potential to increase usage; a 2018 study of transit satisfaction survey data from Santiago de Chile found that wait time reliability was the most important predictor of public transit user satisfaction. To make transit a reliable public service—let alone the bulwark against climate change and urban congestion that it needs to be—we need to make buses run reliably.
In recent years, discussions around bus reliability planning have focused on:
- Service management technology and techniques that would fine-tune the frequency and precision of dispatching interventions.
- Street designs that might reduce conflicts between buses and other road users (especially cars).
- Bus route design that could to reduce exposure to turns and road segments known to cause extensive delay.
This work has been impactful. Though the political landscape for buses remains challenged, many bus networks can point to material improvements in system performance related to their implementation of these practices. But in discussions around those general strategies, the details of how delays materialize and propagate can be lost. That day in the rain, I found myself wondering: how did this gap appear?
In the weeks since my trip, I have often found myself playing with the MBTA’s bus performance data. The agency has made an admirable commitment to transparency by releasing raw bus performance data. These data allow interested researchers such as myself the opportunity to dig deep into the inner workings of MBTA service, and to take an honest crack at answering quandries such as the one I posed above about the 77.
Though these data capture only a small slice of the full complexity of the MBTA’s operations, the patterns one can find them are multiple and meaningful. Like many other bus routes, the 77 suffers a spate of headwinds endemic to transit operations in a car-centric transport regime. With limited priority and high exposure to environments dense with traffic, buses experience delay across their entire trip, and rapidly translate that delay into inconsistent service. But the 77’s performance data also reveal numerous operational factors that compound the effects of this structurally disadvantageous environment. Whether it be the design of the route’s limited bus lanes, its lack of schedule resiliency, or the consistent underperformance of handful of trips, riders’ experiences of the 77 are inseparable from the operational details of the route and the streets on which it runs. The 77 can thus begin to illustrate how the finer details of transit operations ramify the health of a service—and the networks of which each of them are a part.
Three notes:
- Due to my relatively greater level of knowledge on the topic, I am going to spend more time in this post discussing service management issues on the 77 than street design challenges. This is not meant as an implicit suggestion that streets factors in bus service are secondary to service management, but rather an attempt to spend my words where I think they might provide the most value, and to highlight the interconnectedness of the two problems.
- All analyses of operational performance data in this post focus on the weekday portion of the 77’s Winter 2025 timetable (December 2024-April 2025), which was the most recent completed timetable when I began writing this post in April. I focus on weekdays to simplify analysis—weekdays and weekends have different timetables, so writing this post to include all timetables would have made its length even more unwieldy while likely also being repetitive.
- The T’s open data is extensive, but not comprehensive. Innumerable institutional factors, pieces of operational context, and types of performance data are not available to the public. As such, while I have tried to dive deeply into the problems of the 77, my writings should be understood as a first look at an incredibly complex set of problems, rather than an authoritative account.
What is the 77?
The MBTA’s 77 is not a particularly interesting bus route. Like many of this country’s surface transit services, the route got its start as a streetcar connecting the Boston Elevated Railway’s Harvard subway terminal to the (then-peripheral) suburb of Arlington. Over the years, it has changed little; the route today is a frequent subway feeder for T’s Red Line running between a bus loop in Arlington Heights and Harvard Station (or rather, the bus terminal area adjacent to Harvard in Bennett Alley).
Unlike other MBTA routes, the 77’s timetable pattern is also straightforward. While many of the T’s routes share buses (“interline“), and some run trips that cover only part of each route (“short turn“), the 77’s schedule makes only limited use of either technique. Nor is its service profile particularly remarkable: though the scheduled end-to-end travel time of buses varies significantly with the daily profile of congestion and ridership, scheduled headways (transit slang for the time between vehicles) are relatively stable. This post analyzes weekday data from before the 77’s Spring 2025 service increase, so during peak hours the route was scheduled to run every 10-12 minutes (morning) or 12-14 minutes (evening). During middays and evenings, that frequency fell to 15-16 minutes.
Being a subway feeder route, the 77’s ridership patterns are also simple. Most riders board in either Arlington (heading inbound) or at Red Line stations (going outbound), and then ride through to the opposite end. While some riders do appear to make mid-route trips to and from schools and local shopping districts, the general rate of ridership turnover is low, while the long length of passengers’ trips means that bus loadings can be fairly high.
Operationally, the 77 is challenged. The MBTA’s service delivery policy defines bus performance in a two-part framework, focusing on service’s adherence to certain standards at “timepoints”—key bus stops used for scheduling—along each line.
- On routes that run less frequently than every 15 minutes, buses must depart their origin 0-3 mins late, pass mid-route timepoints between 1 min early and 6 mins late, and arrive at their destination less than 5 mins late.
- On routes that run more frequently than every 15 minutes, buses must depart origin and mid-route timepoints no less than 1 scheduled headway, plus 3 minutes, after the preceding bus. They also must reach their destinations with an end-to-end trip time of less than 20 percent greater than schedule.
The 77 falls into the former category, and sees middling performance: in 2024, only about 76 percent of the route’s weekday timepoint arrivals met the T’s standard, matching the average for the city’s frequent routes. Normal as it might be for buses in Boston, the line’s current level of performance leaves something wanting. During the route’s Winter 2025 timetable, a weekday rider at any one of the 77’s timepoints was liable to wait 1-2.5 minutes longer than scheduled, on average, due to trip cancellations, bunched buses, and variability—and one out of every twelve 77 trips followed a gap between buses at least 10 minutes longer than scheduled. Nor were trip times all that predictable, either: only about 81 percent of 77 trips meet the MBTA’s “scheduled runtime plus 20 percent” benchmark. The 77 is not a performant service.
Lateness, Variability, and Gaps
Understanding underperformance in transit services requires a working theory of operations. For most transit services, any such theory must begin with a route’s timetable. Timetables animate transit service: from policy, modeling, and budget inputs that determine the frequency, running time, and design of a transit network, timetables produce a detailed plan of action for daily service. They delineate how vehicles will enter and leave service, how they should move between terminals, and how they should be crewed. Yet important as timetables may be, few riders of frequent transit services like the 77 think about them. As I did that fateful afternoon, they turn up to a bus stop and expect the next bus to be some reasonable distance away. But here is the catch: in a timetable-based system like the MBTA’s, schedule adherence is what mediates regularity.
On transit routes with even headways (as opposed to timetables that schedule bunching), there are essentially two varieties of service gap: cancellation gaps and variability gaps. Cancellation gaps are fairly self-explanatory—if you cancel a trip, you are doubling the time between buses. Variability gaps are slightly more complex. On most transit routes, lateness can be thought of as having a “trend” and “swing” component. Some line segments or hours will tend to see more lateness due to schedule design or service environment challenges that shift all trips’ schedule adherence. This type of structural misalignment is the trend component; in the 77’s case, an example of a lateness trend might be the fact that the median inbound 77 bus is about 4 minutes late at Porter Square. Lateness trends do not have direct gap impacts. After all, if every bus at a point is 4 minutes late, the rider experiences service whose spacing is approximately the same as it would be if they were on time.
The other variety of lateness is “swing,” or variability. From trip to trip, differences in loading, traffic conditions, operating style and more will impact just how late (or early) each vehicle is along its route. This type of variability can and will produce service gaps, especially on frequent routes. For example, on a bus route with an 8 minute headway (time between buses), one bus running 2 minutes early and the next running 2 late will create a 12 minute service gap—or, in other words, a 50 percent increase in headway impacted stops. Once seeded, these gaps tend to grow. A bus following a 12 minute gap is carrying 12 minutes worth of accumulated ridership, rather than 8. The extra load causes an increase in dwell time at stops, as both the volume of boarding or alighting riders increases, and the difficulty of squeezing onto a crowded bus grows. This effect progressively slows buses behind gaps down, eventually to the point where they bunch up with the buses behind them. Needless to say, this behavior is highly impactful to riders: not only do passengers awaiting these buses receive uneven service, but those aboard gap-following buses experience significantly lengthened travel times.
Dissecting the 77
If we accept that variability is important and that service gaps are critical drivers of both customer experience and vehicle delay, the path to understanding the 77 then lies through understanding its schedule adherence and gap rates. These aspects tell a remarkably simple story about the 77’s performance. Beginning with schedule adherence, the line clearly suffers from two major problems:
- In both directions, the spread of lateness values increases dramatically as buses proceed along the route. This is an indication of a lateness swing problem that would drive service inconsistency.
- Especially in outbound (i.e. Harvard to Arlington) service, the spread of lateness values is significant from origin terminals. In other words, buses are frequently being dispatched from Harvard late.
Gap rates tell a similar story. Breaking 77 headway data down to separate cancellation gaps from variability gaps, one can readily see that:
- Cancellation gaps are a non-negligible but relatively insignificant fraction of the overall gap challenge.
- Especially in the inbound direction, gap rates escalate as buses travel down the line.
- As one might expect given the lateness spread pattern in the schedule adherence charts, the 77’s outbound service suffers extensively from variability-related gaps that begin at Harvard.
These charts produce two simple questions: what causes the variability in 77 service? And what makes its terminals susceptible to dispatching service gaps?
Managing Traffic
The first step to understanding any bus route is understanding where and why its travel times tend to jitter. The 77’s challenges with variability are significant, and a structural feature of the line’s service environment. But, as ever, this general problem expresses itself through a handful of specific bottlenecks, whose challenges are broadly representative of the street planning tradeoffs that so commonly afflict bus service in urban environments.
Some of the 77’s variability challenges are straightforward products of traffic. Between Harvard and Walden Street, Massachusetts Avenue traffic volumes are high and bus stops tend to be busy. As a result, buses suffer traffic-related variability between stops and loading delays at stops. Similarly, at the northern end of the 77 in downtown Arlington, high ridership and apparent traffic interference from cars pulling in and out of local businesses increase variability.
Alongside these simpler traffic volume delays, a handful of complex intersections wreak havoc in 77 service. The most impactful of these are the intersections of Massachusetts Avenue and Mystic Street in Arlington, and the intersection of Massachusetts Avenue and Alewife Brook Parkway on the Arlington-Cambridge line. In both cases, the high volume of turning traffic (and the attendant complexity of traffic light phasing and lane design) makes for variable levels of delay. Buses passing through these intersections are more likely to be cut off by other drivers, and are also more likely to be caught by a red light given that additional signal phases tend to extend the total cycle time of each traffic light.
To the credit of involved municipalities, some of these congestion hotspots have received (or soon will receive) bus priority treatments. On the Cambridge-Arlington line, bus lanes carry 77 trips into the Alewife Brook Parkway intersection, and Cambridge is planning lanes for the lower portion of Massachusetts Avenue. While these bus priority measures have been unquestionably effective at reducing travel times, they are imperfect—and broadly emblematic of the North American circulation planning deficit so meticulously illustrated by Marco Chitti.
Northbound 77 performance approaching Alewife Brook Parkway captures these challenges most clearly. During the morning rush hour, the Churchill Avenue-Gladstone Street segment of the 77’s journey has the unfortunate distinction of being the single highest-variability stop-to-stop segment anywhere on the line. The bus lane implemented in this area is a typical offset lane, buffered from the curb by a bike lane. Like many offset bus lanes, its exclusivity is regularly attenuated to permit right turns onto intersecting streets. While many of these permitted turns are low-volume (and therefore low-impact), the right turn flow onto Alewife Brook Parkway at the north end of the lane is not. Conflicts between this high turning volume and the flow of 77 service appear to cause extended and variable delays for riders, who often find their journeys muddled by a sea of turning cars.
Terminal Troubles
With each additional increment of variability, the regularity of 77 service degrades. Buses run late, bunch, and strand riders at stops amid growing gaps. But bus service on the 77 and elsewhere is not always bad; a rough morning rush hour does not instantly lock the rest of the service day into chaos. Standing as one of the most important firewalls between one trip’s delay and the next’s timeliness are a line’s terminals, and the schedules that parameterize them. Just as en-route variability can determine the fate of a trip, terminals and terminal-related variability can tip the scales against a trip before it even turns a wheel—making them an essential (if often overlooked) element of the bus performance system.
As was immediately evident from schedule adherence and gap charts of the 77, terminals are a critical part of the line’s challenges. To be specific:
- Over 35 percent of gap events on inbound trips and over 50 percent of gap events on outbound trips are associated with gaps that existed at a trip’s origin terminal.
- An additional 13 percent (inbound) and 10 percent (outbound) of gap events are associated with trips that left between 1.5 and 3 minutes after their scheduled headway—trips whose gaps may have been “seeded” at origin, in other words.
These data highlight the vulnerability of the 77 to ricochet, or the tendency for lateness and gaps to bounce through terminals. On most bus networks, schedules are assembled in “blocks” that define the movement of equipment on a line. Critically, vehicle blocks at many agencies are tightly bound to crew runs—drivers may change buses around their lunch breaks (to give one example), but otherwise will often work with a single vehicle for their entire shift. Crewing bus service in this manner simplifies operations and can reduce costs, but it also transforms terminals into simple input-output processes. If a block’s schedule provides (say) 5 minutes of recovery time between trips, and an inbound trip arrives more than 5 minutes late at the destination terminal, the outbound trip will necessarily leave late in the opposite direction, and may actually recover little of its inbound lateness at the terminal when involved drivers need to utilize their break between trips to use the bathroom or otherwise.
On the 77, this challenge with recoveries is particularly pronounced at the line’s Harvard terminal. Turning buses around at Harvard requires a fair bit of movement through congested streets, meaning that minimum turn times and turn time variability are high. Conversely, the volume of buses terminating at Harvard means that scheduled turn times cannot run much longer than minimum turn times, lest buses laying over between trips on the eight routes terminating at Harvard’s busways saturate all the street space in Harvard Square. The result is a terminal that is scheduled for stress. Even buses that arrive at Harvard early are liable to leave late, and those unfortunate enough to arrive at Harvard more than a few minutes late will almost certainly leave late as well.
As a result, the 77’s ability to recover delays and gaps on inbound trips is limited. Almost 30 percent of all trips departing Harvard are both leaving more than 90 seconds late and are doing so because of a late arriving inbound bus. In gap terms, this means ricochet: excluding gaps from cancelled trips, 50 percent of service gaps that pass Porter Square going north throughout the day correspond to a southbound gap that bounced back from Harvard. The kicker? 40 percent of those southbound gaps originated at the north end of the line in Arlington.
The Outliers
Terminal performance and runtime variance challenges are structural features of 77 service—they impact all trips, all day. Yet like most such phenomena, their impact is uneven. The 77’s gap rates readily display the lumpiness of service degradation: ten trips in the line’s schedule (approximately 5 percent of the day’s scheduled service) are responsible for 13 percent of all service gaps. While these trips’ impacts are extreme, their challenges are generally pedestrian; understanding them can help draw out how operational challenges compound global issues with variability and terminal performance to produce inconsistency.
It does not take long to notice that the 77’s outliers are patterned. Though they are spread across the day, the structure of outlier trips on the 77 reflect the route’s tendency towards ricochet, with outlier trips often related to previous outliers. This effect is much more clearly viewed on a stringline chart (a time-distance view of performance) than a heatmap, and is even more legible if one expands the list of outliers to include the top 15 and not just the top 10. The 77’s outlier trips not only pair over a round trip, but also display tight sequential linkages throughout the day. In the Winter 2025 timetable, eight of the fifteen worst performing trips on the 77 were part of the same vehicle block, T77-121.
Representative of block 121’s troubles is the 3:17 PM departure from Harvard to Arlington Heights, the second worst-performing trip in the 77’s Winter 2025 timetable. 56 percent of its arrivals at measurement followed service gaps, or more than twice as many as the 77’s average gap rate.
As with all outliers, the trip’s poor performance was related to lateness variance. Where the preceding outbound trip from Harvard (the 3:04 departure) generally left Harvard on time, the 3:17 tended to leave 4 minutes late, effectively guaranteeing a persistent gap in service.
The 3:17’s poor departing performance is, in turn, a study in compounded impacts. The trip is scheduled to turn from an inbound trip that arrives at 3:11, but unfortunately, that trip tends to arrive at Harvard about 3 minutes late. Under favorable conditions, it takes about 4.5 minutes to turn an inbound bus back for outbound service at Harvard; with 3 of the 15:17’s 6 minutes of turn time generally lost to lateness, the bus’s schedule ends up forcing the trip into a late departure from the Harvard terminal. To put some numbers to this pattern: in the Winter 2024 timetable, two thirds of all 15:17 trips leaving Harvard more than 90 seconds behind schedule arrived at the terminal sufficiently late to make a late outbound departure inevitable.
If outbound performance is driven by unrecoverable inbound lateness, inbound performance appears to be a product of some murkier dispatching problems in Arlington. In keeping with the 77’s general theme of terminal performance challenges, the inbound trip that feeds the 15:17 (the 14:36 from Arlington) develops a significant portion of its lateness at its origin terminal. Interestingly, however, the trip’s late departure rate does not appear to be a (direct) function of inbound lateness. While the 14:36’s feeder trip is also a poor performer, the trip has 17 minutes of turn time in Arlington; theoretically, a turn of that length should provide adequate margin to absorb the 6 minutes of lateness that generally attends the arrival of the previous trip.
This pattern is, unfortunately, representative of a broader challenge in the outlier block. As its day goes on, this block’s timeliness leaving its terminals falls. The first inbound trip of the day in this block leaves Arlington late 8 percent of the time; the final one does so 62 percent of the time. While some of this deterioration is a function of accumulating delays throughout the day, it is difficult to tell exactly what is driving the underperformance of this block’s 14:36 and 16:07 trips at Arlington. Perhaps some local ridership generator tends to let a wave of people onto this bus, causing boarding delays; perhaps the operator of this block changes over around these two trips; perhaps these trip are often held. Whatever its cause, the pattern and its effects are clear: the 14:36 leaves late, gets later, and arrives at Harvard with insufficient time to turn around before its next trip.
A Day of Gaps
Up until this point, we have been discussing the 77’s challenges in aggregate terms. Months-spanning analyses of variability, trip performance, gap rates and more are indeed essential to understanding how routes break down—but making these observations actionable requires an understanding of how they ramify real buses, rather than an imaginary average vehicle.
April 2nd of this year happens to have exemplified the 77’s issues. There was no immense breakdown of service that day, nor any rash of cancellations—and yet, putatively frequent 77 service that afternoon contained gaps close to half an hour in length. Behind those sizeable service gaps lies the sort of ricocheting service deterioration on which this post has focused.
The sequence of events on the 2nd appears to have been as follows (see the animated stringline below):
- Due to some traffic or loading-related variability, block 112’s 13:39 trip from Arlington to Harvard arrived at Harvard about 4 minutes late (1)
- With only 4 minutes of scheduled turn time at Harvard, that late inbound late trip produced an even later outbound trip. The 14:17 trip from Harvard back to Arlington left Harvard 8 minutes late, and (predictably) lost time on its journey back north. By the time it reached Arlington, the bus was 13 minutes late. (2)
- The 13 minutes of inbound lateness on block 112’s trip does was not absorbed by the block’s 8 minutes of turn time at Arlington. The outbound trip left Arlington 11 minutes late, and ended up behind a pair of buses the 77’s timetable inserts at Appleton St, seemingly to handle dismissal crowds from Arlington’s high school. The resulting gap in 77 service ahead of the school trips ensured that they would perform poorly: the first of them (part of block 113) ended up arriving at Harvard about 8 minutes late (despite a punctual departure from Arlington), while block 112’s inbound trip from Arlington struggled along with almost 15 minutes of lateness arriving at Harvard. (3)
- Predictably, these late inbound trips produced lateness outbound. Block 113’s 15:53 outbound trip left 6 minutes late, while 112’s 15:42 trip left 19 late. Both trips maintained their lateness to Arlington, where they arrived 9 and 15 minutes late, respectively. [The missing stop records for Arlington Heights appear to be a data error] (4)
- Upon these trips’ arrival at Arlington, they finally caught up to their schedules. Both trips made roughly on-time departures back to Harvard, ending the propagation of the gap. (5)
The relevance of this day of service is similar to that of outliers. Each day’s traffic, loading, and dispatching conditions will vary, but the basic weaknesses of a route will generally stay the same. That the trips involved in April 2nd’s gaps performed well on April 3rd tells us little; that gaps on a different set of blocks ricocheted through Harvard that evening tells us lots.
A Better Bus
On April 6 of this year, the 77’s Winter 2025 timetable ran for the last time. When it started the next day, the 77’s Spring timetable brought a significant service increase, improving travel for the thousands who use this route each day. Unfortunately, the improvement in service did not accompany an improvement in reliability. In this new timetable, the 77’s problems appear to be fundamentally the same: the route still suffers from terminal-driven gaps and extensive variability.
This point bears emphasis as incitement. 77’s problems are complex and persistent, but fundamentally tractable. There is no entropic malaise cast across the route, nor is there a lack of information about its failures. Improving the route requires some tinkering with street and bus schedule design, and ambition with service management.
The easiest way to fix the 77 is to not break it in the first place. To succeed, the route needs a street environment that minimizes traffic interference to limit service variability. Cambridge and Arlington’s existing bus lanes may not be perfect but constitute commendable steps in the right direction; they produced significant reductions in travel time when installed. Upcoming work to redesign the remainder of Massachusetts Avenue in Cambridge promises to bring similar improvements. These projects will significantly expand the scope of bus priority on the 77, and will come with an accompanying build-out of floating bus stops that should reduce delays 77 trips experience when pulling out of and into stops.
Despite these measures, challenges with traffic conflicts will persist. Cambridge’s transit priority plans are laudable, but are neither comprehensive nor continuous: they preserve most right turns across bus lanes, maintain complex traffic flows at intersections, and (prospectively) involve part-time bus priority measures. If the controversy over the existing spate of improvements is any guide, a more ambitious traffic circulation redesign for Massachusetts Avenue may have been politically infeasible. Yet maximizing the reliability of surface transit inevitably means confronting these sorts of tradeoffs. Even if on a limited scale, our understanding of bus priority must evolve to include a greater emphasis on controlling the conflicts inherent to complex urban traffic patterns.
Once variability sets in, it becomes the MBTA’s responsibility to manage it. While the T has made strides to improve bus dispatching and routing in recent years, the 77’s performance highlights the some of the agency’s potential avenues for further enhancement. On the 77 and elsewhere, the T has an opportunity to develop a contextual service management playbook that aligns schedule design, ridership patterns, and disruption response strategies to ensure that riders receive the best service possible.
Many contemporary bus service management discussions revolve around the tradeoff between two different models of service regulation:
- Schedule regulation. Many transit agencies provide bus operators with detailed schedules that tell them what time they should leave various stops along their trips (timepoints). If they arrive at any of these timepoints early, they are generally asked to hold until their scheduled departure time. On low-frequency services where riders are reading a timetable before catching their bus, this practice ensures riders do not miss an early bus; on medium- or high-frequency services where riders look for even headways, holding reduces total variance from schedule by controlling earliness. Schedule regulation has the advantage of being simple and largely passive (as in, implementing it requires little more than a paper card with a timetable)—but it can both be a blunt and costly instrument. If traffic happens to be light one day, all buses will end up holding for earliness; if it is heavy, the schedule provides little spacing regulation because all buses will be late. Schedule regulation’s efficacy as a gap management mechanism also hinges on there being sufficient slack in the schedule for poor-performing buses to recover delays, which can lead to excessively-padded timetables and therefore higher costs.
- Headway regulation. Contra schedule regulation, headway mangement proposes a more selective approach to service regulation: dispatchers make every effort to depart buses from terminals at a set headway, and to re-space buses for evenness at mid-route points irrespective of their timetable deviation. As the theory goes, this more selective (and active) approach helps align service management incentives with riders’ interests on high-frequency routes where people are not likely to be waiting for a specific scheduled bus, and also helps cut the amount of padding in timetables. Operations that have switched to headway management have generally (though not universally) seen improved reliability on high-frequency routes. However, the practice can impose significant demands on agencies’ dispatching staff and IT budget due to the need for more active interventions in service and (often) more extensive technological supports for service management.
The MBTA’s network is currently managed on a blended schedule and headway basis, and with a light touch. As discussed previously, the agency’s metrics are headway-based on frequent routes. In practice, however, it appears that the agency’s service management strategy is built around schedule management. Unlike other schedule-driven agencies, the T makes limited use of passive schedule regulation; at least as of 2018, operators did not get instructions to hold at mid-route timepoints. The absence of regulatory holds means that service is only managed by dispatchers and by the capacity of terminals to absorb delays, which helps explain why variability on the 77 increases monotonically between terminals.
Ironically, the 77’s ridership patterns underscore the import of how one operationalizes either schedule or headway management, more so than they provide a resounding endorsement of either model. The 77’s largely end-to-end ridership profile means that it is important to ensure evenly spaced service at origin terminals—and, conversely, that any service regulation actions which slow buses down mid-route may inconvenience more customers than they help. Because of its high frequency and considerable variability, it likely makes sense to emphasize headways over lateness on the 77. But in practice, the literature that offers the greatest actionable input for the 77 service is that which describes ways to optimize the placement of “control points,” the stops in the schedule or a headway regulation scheme where buses will be more frequently held for alignment. On the 77, these key points are the line’s termini. While the T certainly should monitor mid-route service quality to help build intra-line ridership, improving 77 service in the short run probably means focusing on improving terminals, reducing operations-side sources of en-route variability, and making surgical applications of active service management.
The obvious place to start in any such effort is with scheduling. Many of the problems discussed in this post hinge on the difficulty of recovering from delays at the 77’s terminals—especially Harvard. While longer turn times have a cost in crews and equipment, the extent of their impact on the 77’s performance suggests that this is a cost that should be borne to ensure better alignment between promised and delivered service. In line with accepted practice elsewhere in the industry, the T might evaluate terminal turn times long enough to absorb normal levels of variability, with a margin for operator breaks. Explicitly tying turn times to variability could both help reliability and demonstrate the benefits of bus priority measures: every increment of reduced traffic interference from bus lanes can produce faster bus trips and less slack time at terminals.
Even as timetables change, the nature of distributions means that a few trips in every schedule will cause outsize performance problems. To the extent that these trips can be quickly identified and then managed with operations changes (e.g. different instructions to operators, or regularly allocated dispatcher time to a period with a concentration of outliers) they should be. Not all of these challenges will be solvable, but investigating outliers equips agency teams with valuable information on which scheduling, operations, crew performance/training, and traffic factors tend to create them. In the complex information environment of a control center, knowing what to watch and where to focus efforts can help leverage resources to great effect.
Lastly, the T likely needs to evaluate the resourcing of its bus dispatch function. The T relies on control center staff and strategically located field supervisors to control bus service. Recent data are not available, but as of 2018 the T’s control center had one of the higher buses-per-bus-dispatcher ratios in the American transit industry. This statistic is significant because most bus gaps are recoverable with active service management. Active service management contrasts with passive management (e.g. holds to schedule) in that active strategies are implemented on an as-needed basis, and that they can change a bus’s movement plan by (for example) having a trip skip some stops or make an unplanned short-turn. Given the 77’s tendency to acquire snowballing lateness and its ridership’s sensitivity to mid-route holds, providing more active supervision may be especially beneficial for the route. Recalling the example from April 2nd, running one of the impacted buses express for a few stops could have helped prevent further delays—but having the bus do so would have required a dispatcher’s intervention. If this staffing discrepancy still exists, increasing the number of control center dispatchers (or the level of automated support available to them) may yield a strongly positive return on investment.
Conclusion: Putting it Back Together
The 77’s story is not unique. Those who have marinated in bus dispatching literature will probably find my observations a tad unoriginal—and that is because they are. For decades, practitioners have identified street design, terminal operations and schedule structure as key drivers of bus performance, and for decades, these issues have persisted. Yet the universality of these issues is both a challenge and a strength, because it means that operational remedies for the 77’s problems may be transferable. One need look no further than the rest of the MBTA’s network for evidence of the same. In 2022, the busiest bus route in the T’s network was the 28, a workhorse route through Mattapan, Dorchester and Roxbury. Plotting its service gap charts reveals ailments similar to the 77’s—mid-route variability, and a high volume of service gaps originating at the line’s terminals.
At their core, then, the 77’s challenges both reflect the unending tyranny of operational details across bus networks, and these details’ entanglement with much larger strategic and policy choices around scheduling, street space, and more. It is both true that the incremental work of addressing the tangle of operational, schedule, and traffic performance problems that produce outliers like the 15:17 can have outsize impacts on any bus route’s performance–and that such work is essentially Sisyphean absent a policy decision to invest in realistic schedules and build an extensive network of good bus lanes. Conversely, addressing the fact that the T (pre-COVID, at least) had fewer dispatchers per active bus than most peer agencies would achieve less than one might hope without a finely tuned sense of where and how different bus routes tend to break down. To generalize: producing good outcomes for the users of complex systems requires clear and well-crafted policy goals, intelligently designed network strategies, and strong operational execution. Critically, information must flow in every direction across this stack. Strategies and policies developed independent of tactical knowledge will fail; operational plans produced without any strategic or political framework will not cohere.
The importance of this feedback loop is perhaps my main point in this post. Transit faces innumerable political challenges in this country—funding crises, a political refusal to confront road space allocation tradeoffs, fickle support for network expansion, and more. There is no way to operate one’s way out of these challenges, as they are endemic to this country’s political economy of urbanism and transport, and thus require political advocacy. And yet, within this challenging landscape, and in the face of ever more attractive alternatives to bus service, leveraging the controllability of details to ameliorate service provides an accessible (and often cheap) lever for agencies to achieve ridership and revenue improvement. Many more riders might avoid a walk through the rain if agencies choose to pull it.