This is a hypothetical redesign of King County Metro’s transit service in the Seattle area, with smaller changes to Sound Transit service, meant to enhance riders’ ability to reach destinations throughout the city whatever the time of day. Within the city’s borders, every stop is served by at least one route with 15-minute or better headways, 24 hours per day, seven days per week. This improves riders’ ability to move through Seattle by an average of 26.9% on weekdays, with greater increases on weekends.
The resultant network and schedule is viewable in this basic viewer1.
A Measurement for the Whole City
Access measurements quantify the opportunities available to transit riders within a time budget. Opportunities are destinations with significance to riders. A time budget signifies the maximum amount of time that a person will spend riding, walking, and waiting in a single trip. Calculating an access measurement for a city involves selecting points throughout it and computing isochrones—the region that can be reached within the time budget for a single point—then counting the opportunities within each isochrone. An often-used example of an access measurement is how many jobs the average person can reach within 30 minutes. This is weighted access: the weights are the number of people living around the origin point, and the number of jobs within their isochrome.
However, the analysis in this article uses unweighted access. The access computation simply counts up the origin, destination, and starting time combinations that fit within the time budget. In this case, origins and destinations are sectors of a uniform grid overlapping Seattle. The count of combinations achievable within the time budget divided by the number of combinations, achievable or not, is presented as the Access Ratio. While a higher ratio indicates a greater ability to quickly move through a region, the value itself doesn’t map to something intuitive for riders. It is useful, though, as a comparative tool when considering network changes.
While unweighted access may seem like it’s discarding relevant information compared to the weighted variant, choosing it for this analysis is intentional. When transit funding is constrained, prioritizing the ability of transit-dependent people to reach vital destinations makes sense. The relative abundance of public transit in Seattle allows the consideration of other goals, like reducing vehicle ownership. With an 80% household car ownership rate, though, a typical person expects to be able to quickly move between arbitrary points in Seattle whatever the time of day. Unweighted access scores public transit’s ability to do this.
A Whole New Network
In order to increase the unweighted access score, you’d have to redistribute transit in several ways. Subsequent posts will describe route changes in greater detail. In this analysis and accompanying map:
- Commuter routes are truncated and connected to 1 Line stations.
- Every route within Seattle has a fixed frequency all day, typically no less than its current midday headway.
| Headway | Routes |
| 8 minutes | 1 Line |
| 10 minutes | RapidRides, 7, 36 |
| 12 minutes | 6 (portions of 7, 60), 14, 44, 48, 70 |
| 15 minutes | All other Seattle routes, except Water Taxi |
- Routes that serve stops inside and outside of Seattle are split. The portion within Seattle is rescheduled with a fixed headway. The part outside it maintains its existing schedule, but is not simply truncated at the border. The route is continued to a significant transfer point. The 1 Line, E Line, and H Line are exceptions to this rule; the entire route receives the new span and headway. Here’s a list of the new route numbers split from existing Seattle+suburban routes:
| New Seattle-Contained Route | Original Seattle+Suburban Route |
| 23 | 124 |
| 35 | 345 |
| 46 | 106 |
| 47 | 107 |
| 51 | 131 |
| 52 | 132 |
| 58 | 128 |
| 72 | 372 |
- Some streets that currently have bus service lose it.
- Routes are redesigned, sometimes extensively, to avoid most overlap and create transfer opportunities.
- Streetcars are mothballed.
- “Custom Bus” routes for schools (routes in the 980s) and the Water Taxi are unchanged;. Their stops are exceptions to the rule that all stops within Seattle have frequent 24/7 service.
Downtown Seattle is heavily affected by these changes. Buses no longer run on Third Avenue south of Pike Street, or on the Second, Fourth, and Fifth Avenue corridors used by current commuter service. Outside of walking, the 1 Line—running its current peak frequencies day and night—is the sole option for moving northwest-southeast through downtown. In the near future, the 2 Line will supplement it. Routes still cross, or stop at, Third Avenue, providing access to points in the city’s east. Routes that once approached from the south are tied into a variety of 1 Line stations. Reaching points to the north and northwest may involve a walk or very short 1 Line trip, then taking route 4, the D Line, or the E Line, before connecting to a route that has been modified to branch off of them. Eastside commuter routes using State Route 520 tie into the 1 Line at University of Washington; those on Interstate 90 connect at Chinatown/International District or Pioneer Square. One-seat rides to downtown would become less common, but with every route contained within Seattle running 15-minute or better headways all the time, transfers are less arduous. Nonetheless, this restructure asks riders to reimagine how they navigate the city.
In return, the access analysis reveals that these changes represent a significant improvement for riders. Metro’s last major service change, which expanded service rather than just redistributing it, improved 45-minute access in Seattle by only 1.3%. In contrast, this redesign improves 45-minute access by 26.9%–32.9%, and all time budgets by 16.8%–32.9%.
| Schedule | Time Budget | Current Access Ratio | Restructured Access Ratio | Improvement |
| Weekday | 20 min | 0.0300 | 0.0341 | 13.7% |
| Weekday | 30 min | 0.0865 | 0.1061 | 22.6% |
| Weekday | 45 min | 0.2433 | 0.3087 | 26.9% |
| Saturday | 20 min | 0.0288 | 0.0341 | 18.4% |
| Saturday | 30 min | 0.0821 | 0.1060 | 29.1% |
| Saturday | 45 min | 0.2321 | 0.3084 | 32.9% |
| Sunday | 20 min | 0.0292 | 0.0341 | 16.8% |
| Sunday | 30 min | 0.0839 | 0.1060 | 26.3% |
| Sunday | 45 min | 0.2375 | 0.3085 | 29.9% |
Each sector has its own access score too, revealing areas of improvement or degradation. This map shows the access change for each sector on a weekday, for three selectable time budgets2. For a 45-minute budget, improvement is widespread, with few areas showing modest losses of access. Some streets that have lost service nevertheless exhibit improved access, because nearby routes have increased frequency, span, and transfer opportunities. This even includes Third Avenue, downtown. The 20- and 30-minute time budgets show a more uneven benefit. Areas that have lost their most proximate route more often face an access decrease. Large swaths of improved access remain for these budgets, though, particularly in Metro’s Equity Priority Areas, without service investment beyond today’s level.
An Unprecedented Shift
Seattle Transit Blog advises contributors that “proposals should be somewhat realistic.” While this proposal makes sweeping changes to Metro’s network, it, critically, requires less in-service time to operate than the present one, on a per-week basis. In-service time tracks with the labor costs that dominate agencies’ operating budgets, and is calculable from schedules. It excludes layovers and deadheading, but—by virtue of reducing peak-only service—this network also necessitates fewer deadheads. This property also reduces the maximum number of simultaneously-operating vehicles, permitting a smaller fleet and workforce, or improving reliability with existing resources.
Yet aspects of this network may raise questions about being “somewhat realistic.” This proposal increases higher-paid overnight shifts. The schedule is also not fully specified: trips have not been collated into blocks and parceled into operator assignments. It may be less conducive to efficient scheduling than the current network. Finally, operators of King County Metro buses, Metro-operated Sound Transit buses, Sound Transit light rail vehicles, and the Seattle Streetcar are considered fungible. All are King County employees, but funding may not be as transferable as this proposal requires.
Reducing peak service also invites concern about crowding. Metro’s last two System Evaluations, however, have indicated that its crowding threshold has not been met network-wide, and Metro’s crowding definition is conservative. According to New Flyer, a 60-foot Xcelsior coach can hold 61 seated passengers and 62 standees, but Metro considers the same bus crowded if there are on average 83 passengers, or there are more passengers than seats for 20 consecutive minutes. I previously examined Metro’s ridership and found that few stop pairs exceeded the manufacturer’s capacity of one 60-foot bus running every 15 minutes. The proposed network overlaps routes or increases frequency in these areas. Introducing bi-articulated buses—deployed in European in South American cities—would provide additional relief. There is a greater, and difficult to model, danger of overloading the Link light rail system. This restructure leans heavily on the 1 Line for trips to and through Downtown Seattle. Even with the 2 Line’s future capacity infusion, ensuring that future Link vehicle procurements and refurbishments wring as much standing capacity as possible from the vehicle’s footprint would help allay the crowding concern.
Perhaps the most unrealistic aspect of this proposal is that it’s unprecedented. It doesn’t comport with Metro’s service guidelines or long-term plans. Planning doctrine positions frequency as a demand-management tool primarily; that it makes transit use more convenient and viable is a consequence, not a motivation. Running service where and when there isn’t commensurate present demand is heterodox. There are no examples of a transit agency redistributing resources this extremely.
These unknowns could render this network unrealistic, but whether they are insurmountable is difficult to prove in advance. The proposal’s measurable reward does require assuming risk.
Challenging the Status Quo
Transit network redesigns are risky by nature. They never yield universal improvement; this one is no exception. Its tradeoffs are of unusual magnitude, though. The access improvements could be life-changing, particularly for those enduring at-best hourly service where and when they need to travel. Conversely, it subjects riders of some popular routes to longer walks and more-crowded buses. With such polarized outcomes, full implementation of this proposal is unlikely.
As a hypothetical, it is nevertheless useful: it exposes the cost of the status quo. The access increase from completely redesigning Metro’s network could have been minimal. This analysis shows that’s not the case. While Metro’s current choices do benefit riders in ways that unweighted access does not measure, the volume of unrealized access should press Metro to contemplate whether its policies strike the right balance.
A community of transit advocates continually making specific, quantitative appeals to improve access would aid this contemplation. The software used to create and evaluate this proposal is open source and free to deploy.
Footnotes
1 Due to a data error, the West Seattle Water Taxi appears in the map of 24/7 service. Its schedule is unchanged.
You first lost me at mothballing streetcars then really lost me at no buses on 3rd Avenue.
It’s a big change, so I get it. If you’re interested in the steps that got me to point of mothballing the streetcars, I’ve written about it elsewhere in the past https://busgraphs.com/post/2025-02-16-streetcars/. As for 3rd Ave, I hope to go into this more in an upcoming post.
Mothballing the streetcars is not a radical idea. Lots of people (including planners from Metro) have proposed it. It is easy to see why. The South Lake Union Streetcar is largely redundant now. Send the H up to South Lake Union and you have better service for no extra money. The Capitol Hill Streetcar is a different beast. It makes a unique connection. But it also wouldn’t be hard to replace with buses and in the long run it would be better. For example you could send the 49 down Broadway until Jackson and then downtown. Also straighten out the 60 so that it runs on Broadway from Jackson as well. Now run both every twelve minutes. Bingo, you’ve got six minute headways (all day long) on all of Broadway. Since there is no streetcar we can remake Jackson. Add BAT lanes and push the cars to the middle of the street. Better yet rearrange the center platforms and have the buses run in the middle of the street but use a weave technique (https://i0.wp.com/seattletransitblog.com/wp-content/uploads/2024/03/segment1_busway-1.png). That means every bus on Jackson would use center lanes and center running buses. The 7, 14, 36, 106 would all be dramatically faster.
So yeah, mothballing the streetcars is quite sensible and I think most transit experts would just get rid of them. On the other hand getting rid of the Third Avenue transit mall/spine is radical. I don’t know anyone who proposes that. The closest I’ve heard to it is one of the ideas put forth by the Downtown Seattle Association (https://cdn.downtownseattle.org/files/advocacy/dsa-third-avenue-vision-booklet.pdf). They would have the buses stop at the edge of town and then run shuttles. The model is Denver (which is like saying the health insurance model is USA — not exactly a city you want to emulate when it comes to transit). It is pretty clear that even with the shuttle it would suck. It would require massive bus stations on each and end. Then, after all that work is done it would be very unpopular. The DSA is mainly concerned that the street appear more lively and it is just one of the many proposals they are throwing out there. As “The Blade” moves around they will focus on other streets and other projects, leaving the issues surrounding the buses on Third to the people who focus on the buses.
Getting rid of the streetcars is quite sensible. Getting rid of buses on Third (with no alternative) is not.
You’re thinking like a transit planner. Thinking about it as a transit user, replacing the First Hill streetcar with a bus would make it strictly worse, partly because it would be a crappier vehicle, and partly because I couldn’t trust you not to change the route anymore.
I am thinking about it like a transit user. I don’t want to wait ten minutes for a vehicle stuck in traffic. I don’t care whether the wheels are rubber or steel — and neither do most people.
replacing the First Hill streetcar with a bus would make it strictly worse, partly because it would be a crappier vehicle, and partly because I couldn’t trust you not to change the route anymore.
But the route is crappy! That is the point! The streetcar takes a stupid route and is stuck in traffic. It doesn’t get crappier than that. As I wrote you could have six minute service along *all* of Broadway. Right now the streetcar just ends south of Denny. Why? Because it is streetcar. Extending it just a few blocks becomes a major undertaking. It takes me twenty minutes (at noon) to go about a mile from stop to bus stop along Broadway. From stop to stop! Here is another one. Basically a half hour just to go a little bit over a mile. At noon! Never mind traffic (which is a major problem made worse by the fact that one of the vehicles is a streetcar) the combination of bus and streetcar is terrible. It isn’t money either. The 60 and streetcar run every 12 minutes. They should combine for 6 minute headways and run in BAT if not bus lanes the length of Broadway.
And that isn’t even the worst part of the streetcar! The one unique routing of the streetcar is Broadway to Jackson. At least it does that well, right? Right? Wrong! Instead of running down Broadway and turning on Jackson it makes this ridiculous button-hook. It is almost as if it is trying to get stuck in traffic — and it does! But again, it is terrible at noon as well. This trip takes a half hour. Stop to stop. Leave at noon, arrive at 12:32. Walking is faster. 1.3 miles and walking is faster — stop to stop! Hell, walking and catching the 60 is just as fast despite the convoluted routing of the 60 (that you definitely can’t trust me not to change).
It needs to change. The 60 needs to change and doing that isn’t that difficult. But things would be much easier and we could build a much better transit system if we didn’t have the streetcar(s).
While shifting to a grid-based network is a noble idea, I would not be happy about service every 15 minutes in such a network. The grid would need to operate every 8 minutes for me to feel that the transfers were not overly burdensome. One alternative is to have trunk RR/Link lines that run every 6-8 min to important destinations, with bus lanes and priority for reliable on-time performance, and decrease the frequency of some feeder lines. Suburban residents would thus be asked to consult a schedule for one leg of their trip, but otherwise also have much improved access.
Appreciate your perspective. What I’m curious about is if there’s something about the present network that makes its largely 15 minute headways more tolerable, when they wouldn’t be in a more grid-based one.
The current network is optimized for commutes to downtown and U District – nearly everyone has a fast one seat ride at least one of the two. They are also the two highest concentrations of jobs. 15 minutes is an acceptable frequency for me when I don’t have to transfer (in fact even 30 min is fine! I will look at a schedule and leave home/work at times compatible with the schedule). When I have to transfer for my commute, this adds a layer of unpredictability (what if my bus that’s running 5 min late makes me miss a transfer and I end up 20 min late). Plus, an ideal commute to me is 30 minutes. If I have to transfer with both of my legs being 15 min lines, about half of that 30 min will just be waiting for the bus (7.5 x 2)!
I don’t think this proposal is especially grid like (unlike this proposal — https://seattletransitblog.com/2013/08/19/your-bus-much-more-often-no-more-money-really/). There are too many places where people have to make same direction transfers just to keep going straight. That is not a grid. Remember one of the key aspects of a grid is that you can get anywhere with *only one* transfer. This fails that test repeatedly.
Consider what Jarrett Walker called “An almost perfect grid”. To quote from that essay:
So look at the network map again. Downtown is the peninsula sticking up on the north side of the city. If you look closely at the system map, you’ll see that most of the north-south lines extend to the north edge of the grid and then bend east or west as needed to flow into downtown. There’s a bit of inefficiency in the resulting duplication, but it’s not too bad. It means downtown’s major attractions and connections anchor the north end of most of the north-south lines.
See? There’s a bit of inefficiency in the resulting duplication. That isn’t the end of the world. In this case there are trips that involve two transfers despite someone going the same direction. I get why you want to avoid duplication, especially when the area is not especially dense. Duplication is like a route with a branch. Just as you want the trunk to have a lot more demand than each branch you want a section that is duplicated to be the same way. But forcing multiple, same-direction transfers is not a good solution (especially when the overlapped section really is higher demand).
Interesting idea. A few questions:
– Have you considered what service would look like if overnight service was reduced, or had a different network entirely? It seems like a waste to run at 15 minute frequency overnight
– Do you have a map that compares against weighted access? Is that data readily available?
– How are you calculating transfers? It seems a lot of trips will become 3-leg here (most north-south trips), and some of the forced transfers are a bit painful (Mt Baker for example)
I also was wondering about cumulative wait times for multi-leg journeys, where the worst-case scenario is just missing a connection for each leg. I’ve made a journe from West Seattle (Westwood) to Lake City (bus-train-bus) and ended up spending a significant amount of time just waiting.
I guess along those lines, it still seems even in this new and interesting proposal to be hard-ish to go E-W (?)
I don’t usually look at it that way, thought it’s interesting way to frame it. It’s true that trips are likely to take more legs, an upside to this proposal is there are no Seattle-contained routes that will add more than 15 minutes for a single leg, regardless of when the trip is happening. Today, missing route 79 midday or any bus after 2AM can be brutal, even if it’s the only leg of the trip.
On the E-W routes, I hear you! So many times during this project I was screaming “WHY DOES THE ARTERIAL END HERE?!” at a city map. I think the revised 45 and 60 and new 58 help fill in some gaps, but I wish I could do more.
I’m glad you found this interesting!
1. This is a reaction I get a lot, and it’s understandable. I have strong feelings about consistent headways.
a. I tend to focus on fixed route transit as a means to reduce car ownership, so I want it replicate the abilities that a person with a car expects to have, at least within the city. Trips at night might be less common, but I suspect they’re typically more urgent. I don’t want transit to fail people when they may need it most; that seems likely to make car ownership seem like a high priority.
b. I think there’s an equity need filled by having high quality transit at night. Agencies will mix an equity component with their ridership measurements when evaluating routes, but it’s strictly based on where disadvantaged people live as opposed to when they’re traveling.
c. There’s a comprehensibility benefit. There’s no need for a rider to know when buses get less frequent. There is a consistent expectation about how easily one can get around.
d. Scheduling is easier because you can always do a 1-for-1 replacement of a bus or driver.
That being said, if you’re looking at it from a trip occupancy level for overnight trips, yes it’s wasteful. I think looking at ridership at the trip level is kind of barking up the wrong tree. I want people to feel comfortable with the system, and I think part of that is running trips that they might never use, but could find useful in an unexpected scenario. Maybe fixed-route service overnight isn’t the best way to do this, but I when I see agencies pilot demand-response (like Metro Flex) they also don’t tend to focus on the overnight. I want there to be something overnight, and fixed-route is what I can model.
2. I don’t pull in the data to do this currently. Conveyal is the gold standard for that sort of analysis.
3. I use Metro’s schedules and compute walking distances using data from OpenStreetMap. I don’t do any approximations like waiting for half the headway; I simulate every possible destination reachable from every origin point at every minute of the day. I adjust walking speed for elevation too (https://busgraphs.com/post/2024-12-20-elevation/). This isn’t perfect (sometimes a stop will get associated with the wrong road), but it compares favorably to most analyses I’ve seen agencies run or in academic literature. This restructure definitely involves more transfers, but it largely seems to work out in its favor, especially for longer time budgets.
1. That makes sense, though I’d argue that as a car replacement people generally care much more about headways during the day than at night
I think comprehensibility and consistency are important but there are other ways to accomplish that. The way you’ve laid out route headways here is nice and consistent. I think something similar could be done by adding additional tiers of routes. I think Metro’s definition of “frequent” is too broad. Routes 5 and 7 are both “frequent” but have very different level of service, for example. It would be nice to be able to think “oh this is a bronze route; it’s 15m minimum during the day and 30m starting at 9pm every day” for instance.
I suppose even headways makes it easy to do the math; halving the frequency overnight here (50% service from 10PM to 6 AM) would equate to 25% more service during the day (15m headways become 12, etc)
3. I see. I was wondering whether things like crossing the street or taking escalators down to the station platform were taken into account
This exercise demonstrates an extreme example of prioritizing coverage over ridership, and I applaud Matt for putting in the work to demonstrate the concept at the scale of Seattle. There are obvious shortcomings to this approach – mainly that the network as proposed would vastly underserve the densest parts of the city while generally overserving others. But I think the point is well made in quantifying the implicit trade-offs between servicing ridership versus coverage.
Thank you Nathan; I appreciate the opportunity to share my work in this venue—the venue ignited my interest in transit planning years ago.
It’s interesting that you’ve brought up the coverage-ridership tradeoff because I’ve been thinking of it a lot recently. I’ve always interpreted the classic Jarrett Walker coverage definition as corresponding to streets with bus service. By that definition, this restructure is a coverage _reduction_ because some streets are losing service. It could also be called a ridership reduction because it in places puts less frequency in the places/times that have historically generated the most transit ridership.
So I find myself straining against the tradeoff. I don’t intend to reduce ridership. When I do reduce frequency, I want to keep it at level that will still be convenient to induce riders to use the service (while frequency and ridership are related, I think the relationship gets frayed at the extreme ends). I just don’t want frequency and network design being used as the first lever to manage crowding. I also think the use of “coverage” in Walker’s work should be “spatial coverage,” and that true coverage is across space (which roads have transit) and time (what the frequencies of these routes). Why I focus on maximizing unweighted access is because it’s encoding those two important things.
I hope to write about this in more detail, but it’s definitely the sort of abstract crap that belongs on my personal site rather than STB…
I didn’t intend to imply that this is ignoring ridership intentionally, but I do think Walker’s philosophy of treating ridership and coverage as a spectrum of priority is a useful construct. To me, the approach of creating a network that maximizes geospatial access agnostic of density of destinations is, implicitly, creating a network that does not maximize throughput, which will impact ridership in dense parts of town.
Again, I think this is a valuable baseline analysis treating mass transit as a system of getting people where they might want to go. The problems that arise when many people want to go to the same place are naturally handled by increasing frequency on those corridors.
To me, the approach of creating a network that maximizes geospatial access agnostic of density of destinations is, implicitly, creating a network that does not maximize throughput, which will impact ridership in dense parts of town.
Agreed. I think the same is true for being agnostic about the time of day. Or transfers. Running the same at 3:00 am as noon (let alone 8:00 am) will definitely lose riders. But the riders at 3:00 am will be happy. They will have something. Transferring twice just to go a couple miles in the busiest part of the city will lose a bunch of riders. But the savings mean other riders have 15 minute service in their low-density part of town (even at midnight). This is not a classic coverage versus ridership trade-off although it is very similar. The word “ridership” is still appropriate but it would be nice to have a word other than “coverage” to describe this more encompassing idea. As I wrote below it is basically making the worst-case-scenario not that bad.
The purpose of transit planning is to move people–not vehicles.
That is a good way to put it Nathan. In that context it makes sense. You could even extend this concept to be focused on the worse-case-scenario rider. This is what coverage service is all about. Instead of walking a mile to a busy corridor with a frequent bus they have service within a couple blocks. The same idea applies to late night service. It is much, much better for the handful who are interested in taking transit at that hour.
But for the vast majority of transit riders it is worse. There is no massive improvement in midday frequency for the buses that run often. Things are actually worse for the RapidRide E most of the day. A lot of riders have to transfer just to keep going the same direction. Multiple times. It is damn inconvenient and there are plenty of studies showing this hurts ridership (both the waiting *and* the transfer itself). But the worse-case-scenario is not that bad.
I’d like to increase more routes to 10 minutes day/evening by reducing night owls to 30 minutes, and eliminating some night service like the 28. Would this be enough to get all routes to 10 minutes or better? If not, how many additional hours would we need?
I also like Sunny’s suggestion of increasing all Seattle RapidRides to 5-6 minutes. Those corridors were chosen to serve their quarter strategically and to mitigate not having rail in that quarter. So the least we could do is make them ultra-frequent.
The best night owl service in other countries I’ve seen is 30 minutes, with parallel routes a mile apart. That’s what Chicago and San Francisco have, and I think London and Vancouver. Even the New York City subway goes down to 20 minutes after midnight in the 8th Avenue corridor, as I found one night when I went to a club. So I don’t think we need to exceed what the best cities have at night.
It’s more important to increase day/evening routes to 5-10 minute service so that there’s at least there’s some time in the day when people can go on errands and activities without excessive waiting. That’s how to make transit competitive with driving and maximize ridership. Frequent night owls everywhere can wait until after we achieve that.
I’m hoping my response to John D’s comment further up will be somewhat elucidating. I admit that this is the aspect of the proposal where I’m most governed by ideology over pragmatism.
There are a lot of variations to try and analyze, so I’m hoping people do use the tooling I’ve built to experiment. I’m also working on optimizing the computations of the analyses to make them less costly, in hopes that analyzing a bunch of networks instead of just one becomes less of a burden.
There are opportunities for combining routes here. For example, The 5 and 40 both end around 45th and aurora (which is a very car centric area) They could easily be combined.
It also seems a bit odd to have a bus terminate just outside of downtown. This network could work well as “all day” service but would almost surely require a mentation during peak hours. (Imagine getting almost downtown then waiting a long time to transfer as multiple full busses or light rails pass by without room)
The geography of Seattle makes a grid a lot more challenging than a place like Chicago.
A combined 5/40 would amount to moving the current 5 to Westlake, which isn’t that far off from a proposal that was floated at some point to move it to Dexter.
The geography of Seattle makes a grid a lot more challenging than a place like Chicago.
Or Vancouver (https://humantransit.org/2010/02/vancouver-the-almost-perfect-grid.html). But it clear that the author of this proposal is not trying to make a grid (as I wrote up above). A grid can involve overlap (as is the case of Vancouver) when the street grid overlaps. While less efficient it retains one of the key advantages of a grid — the ability to get anywhere to anywhere with one transfer.
This proposal is instead a hyper-efficient system designed to minimize overlap. The savings then go into providing more service late at night and in areas with relatively few people. It isn’t a grid. It isn’t ridership focused either. I struggle with a term to describe it.
Sometime around 15 years ago this blog saw a couple of proposals for restructuring service around Link that both involved sending the 7 somewhere other than downtown. One swapped it out for the old 9 up Broadway; the other sent it to First Hill and SLU. Both times the proposals drew howls of protest from Rainier Valley residents; among the complaints was that Rainier was a single unified corridor for transit ridership, even though those proposals didn’t break it up south of Jackson St. Yours does, so I can only imagine what sort of screaming would result if this were seriously proposed.
In the last 15 years, Route 106 was created to duplicate Route 7 north of Mt Baker. Having Link stations opened on Rainier at both Mt Baker and Judkins Park is also a big systems change that offer faster ways to and through Downtown. Finally, many land uses have changed between Walden and Dearborn along Rainier. I suspect that these aspects really change what kinds of opposition would come out compared to 15 years ago.
Yeah I don’t doubt that it would elicit some screaming. One benefit of having the access map alongside a proposal is that it’s possible to click on of a grid sector twice and see how access from that location compares between the existing network and the restructured one. So in theory, someone could choose some points significant to them in the Rainier Valley and see how access to points along the present 7 corridor changes. My (naive?) hope is if the ability to access those points in the same amount of time is shown to be largely unchanged, even though the mechanism of reaching them has changed, then that would at least be convincing to some.
This might make a good “skeleton” grid that you then want to add on to to serve the more dense areas with highest ridership at better than 15 minute frequency. Keep in mind many King County routes start from outside Seattle, and could be used to supplement critical corridors, such as 3rd Avenue, Aurora, and Rainier. Nevertheless, I’m not sure Magnolia should have three routes with 24 hour, 15 minute frequency!
What they have today in Magnolia probably makes more sense than this.
If there is any additional service hour for it, adding more trips to 33 and move it to 24th Ave to be closer to that huge multi-family complex by Magnolia Manor Park is a good improvement before Ballard Link Extension opens.
I am not familiar with every part of Seattle, but that’s largest multi-family complex I’ve run into without frequent transit service.
I’m having trouble with the map so I’m not quite sure what you are proposing. For example the 322, 372 and 522 all run down Bothell Way. They all have an error between Bothell Way and 5th NE. Putting that aside, I don’t know why there are three routes as they appear to be just subsets of each other. Why?
The maps are machine generated and admittedly ugly. When I’m connecting points that don’t currently have service between them, the software can’t use one of the nice paths that Metro provides in its schedules, so it just draws a straight line between the stops (this also means that I have to approximate a time between stops rather than interpolating the time from other trips in Metro’s schedule).
If I were doing a full King County restructure, I’d do something about all those routes being subsets. For the purpose of this analysis, I didn’t focus on anything outside of Seattle beyond just connecting commuter routes to Link stations.
I’m not sure if this is just a glitch in the map when reading this. Have you verified that each ending is at an existing (or at least reasonably theoretical) layover? For example the 45 just ends in Crown Hill. I definitely see the value there but there is no existing layover there. My guess is if there was the 61 would go that far. Speaking of which, the 61 seems to end at Aurora & 85th. That seems even more difficult from both a layover and turnaround standpoint. The 35 also ends in the middle of nowhere.
Yet aspects of this network may raise questions about being “somewhat realistic.”
Indeed. Like whether the layover locations were properly vetted. I’m not sure how you created the map. Was it created using a program (i. e. “AI”) or was it done by hand?
I was pretty relaxed about layovers. I tried to make sure there was some possible way to do a live-loop, and or that there was some street parking that could be grabbed, but it wasn’t a priority in where I ended routes. The idea wasn’t to figure out every operational facet—I’d be way out of my depth there—but basically to provide a semblance what an access-increasing, expenditure-neutral network would look like with resources that were available to me. I’m sure there are some routes that won’t work as represented here, but I’ve seen this Metro’s own proposals (https://seattletransitblog.com/2015/12/01/metro-cancels-capitol-hill-restructure-reconfigures-downtown-trolley-service/).
I went into some aspects of the map in my reply to your last comment, but answer what you’ve asked here: Machine generated, but nothing that I would consider to be “AI,” the source that software that I’ve built to generate them is available at link in the end of the post.
OK, that makes sense. One thing I’ve learned as an amateur planner (working with a professional planner) is that layovers and turnaround locations are critical. There are a lot of great ideas that simply aren’t possible because of those. For example it would make sense to just extend the 61 a little bit to 145th (and Lake City Way) once the 522 is sent to 148th Station. That is a very short distance (the extra service cost would be minimal) but it would make a huge difference. It would connect Lake City Way/Northgate Way (and 5th NE) to bus service on Bothell Way. But you would need to add a new layover there. Worse yet, turning around would be difficult if not impossible. The only thing that could maybe work is go up 30th and south on Lake City Way. But that changes the nature of that route. It wouldn’t be able to take advantage of the northbound BAT lanes on Lake City Way. So maybe you move the BAT lanes to the other side of the street and …
You get the idea. What would otherwise be a very simple, little extension is made very difficult if you can’t turn around or layover. In general I use existing layovers for that reason. When I do suggest something different (like a live-loop) I acknowledge that it hasn’t been done and it is possible it wouldn’t work. But I also dig into the details to see if there is anything that looks like it would be an issue. That is trickier if it is machine generated.
hi appreciate the article and the analysis that went into it.
I think the consolidation of the smaller 30 min lines into 15 min lines is fine. while they can cover more, they are also much harder to use. aka the better magnolia buses and west seattle buses
That being said, i’m not sure about the forced transfers to go from chinatown to westlake. if im in denny traingle heading to swedish hospital would i reallly want to take the 40, go down to westlake station, take one train stop, and then take the new 2 bus?
right now it is a 22 minute bus ride. with the new situation it’d take like 5 min bus ride + 2 minute walking downstairs + 3 min wait + 3 min train ride + 3 min walking back upstairs + 8 min wait +12 min bus ride =28~36 minutes.
And that’s a lot of effort to travel 1.5 miles
There’s just a lot of friction to force 3 transit trips with 2 transfers. Even if it was hypothetically slightly faster travel time on average I don’t think many would want to transfer that many times for short trips
I said the same thing up above (but I think you put it better). It is really the core that struggles. It is one thing to catch a more frequent bus from South Seattle College but then take a frequent express into downtown (connecting to Link and a lot of other buses). It is another to transfer twice just to go a mile in a half in the heart of the city.
It’s a good exercise — but there there are some bus routing fundamentals that are not given the priority that they need.
1. Housing location choices for targeted demographic groups. Many people choose residences because of connectivity to other things. It may be a disability like blindness, or a cultural aspect like businesses that speak a foreign language or other social service reasons. Things like access to a store owner that can converse in a specific foreign language is very important. Splitting a direct route that has operated for 20 or sometimes even 60 years may really have cultural consequences – and the impacted l, disadvantaged demographic may create a politically untenable, visible protest.
2. Seattle’s mostly north-south commercial districts. With very few exceptions, our commercial districts run north-south. People don’t like to transfer if they’re only riding a mile or so. A route that only goes through residential areas has less appeal and has a history of weak ridership and would quickly be on the list for service reduction or elimination.
3. Lack of shared transfers. A good route layout can mean that routes feature transfer points where a rider does not have to cross any streets to transfer. It’s usually complicated to site bus stops near major intersections — as many have backed up traffic or driveways or major turning movements that hinder operations.
4. Driver assignments. Some routes get extended or cut to maximize driver assignments. Union work rules have to be considered. It’s made more complicated when estimated schedules in congested travel conditions. Conversely, if a run time is too short, it may be that an agency can extend a route a few more blocks without a significant cost. Agencies have to plan around this.
5. Driver breaks. Drivers have break needs like the rest of us. It is why many routes end at transit centers.
And then there are concerns about productivity. It’s a balance between too much and too little.
Taken together, bus route changes are not as easy as it looks.
Asking people to transfer is quite reasonable and can result in a much more efficient network. But asking people to transfer multiple times to go a relatively short distance the same direction just won’t work. From what I can tell that is what is required. For example consider a trip from the top of Queen Anne to the south end of downtown. First you take the 28 to Uptown. Then the RapidRide D south. That ends at Pike. Then Link? That is two transfers to go the same direction and not very far. None of the vehicles are running every couple minutes. The 28 runs every fifteen minutes (same as the 2/13 combination today). The D Line runs every ten (same as today) and Link runs every five minutes (after East Link). That is a basically the same as if you caught one bus running every half hour. The transfer to Link is awkward and involves additional walking to get to the entrance as well as walking back and forth to the platform (this isn’t New York where the stations are close to the surface). More to the point that is considerably worse than today. Right now someone takes the 2/13 and then rides the spine to the other side of downtown. Or if they do get on Link it is one bus before they get there.
I am reminded of this post from over ten years ago: https://seattletransitblog.com/2013/08/19/your-bus-much-more-often-no-more-money-really/. This was also a radical rethinking of the network. But all the plans were plausible. It was the type of thing that Jarrett Walker’s firm (and similar organizations) make all the time. This is different. There were transfers but not like this. Obviously some would object to various aspects of David Lawson’s plan but it was easy to see the advantage of such routing for so many trips. Various buses were just a lot more frequent and many were extremely frequent. Core corridors combined for very good frequency. I’m probably missing something but this just doesn’t offer that. Most of the buses *in Seattle* are still not especially frequent.
I’d love to see a weighted analysis. Downtown is just way more important than the low density fringes. Such an analysis might increase justification for route elimination or truncation at link stations too.
There are some aspects of this plan that I find thought provoking and other parts that are mind boggling. I don’t think I’ll ever be convinced that service levels should be equivalent at both 3:45am and 3:45pm. But I won’t dwell on that point.
What I find most interesting is the idea of clearing buses off 3rd Avenue. When I’m in Salt Lake City, I notice that the light rail lines all run in one direction and the buses intersect the light rail lines but don’t run parallel to the light rail on the same street. I’ve frequently wondered if that idea would work here in Seattle. maybe a plan could be developed here given the political and financial realities of digging a 2nd bus tunnel. Would it be feasible to scrap the notion of a 2nd tunnel and accept street level operation of West Seattle to Ballard trains n 3rd Avenue? I doubt any of us are ever going to see a 2nd bus tunnel in our lifetimes.
I looked for your answer to how to improve the 8 and it appears that the 8 is simply deleted. Will riders simply have to find other ways to commute?
Similarly, people in Rainier Valley will be forced to make a transfer at Mt. Baker Station if they want to go downtown. Will they be forced to walk a considerable distance if their destination is north Rainier or Jackson Street?
I’ve seen double-artic buses in Bogota, but I don’t think they operate regularly in mixed traffic lanes. They are usually running in the grade-separated Transmileno BRT lanes. I’d check the operating specs carefully about how well they would work in Seattle’s tight, winding and steep street before proposing them as a solution.
Matt Laquidara,
I’m not a transit expert, but thank you for your post.
One of the things that’s often brought up in this blog, time and time again, is zoning. Current zoning, (the urban village plan) and the current transit system go hand in hand. But if we blow up 50 years of zoning, shouldn’t we blow up 50 years of transit planning? I’m not the expert, but transit needs to change. Thanks for thinking about how that might work, even if the edges are a little rough.
One systemic change of the last 15 years ago in the Seattle area is the growing evolution of a day-long hierarchical system of longer-distance limited-stop routes, and shorter-distance local-stop routes. The longer route systems include Link, Stride (upcoming), RapidRide and Swift and arguably some ST Express. At some point, a single route hierarchy is just not expansive enough to cover our geographic area. It’s great for Olympia but not Seattle.
This hierarchy is important in a large urban area and is how many major metro area transit systems are designed. Taking local routes and transferring multiple times to other local routes is more impractical and suppresses transit use.
Metro route structures have operated for decades before the hierarchy began to be evolved beyond peak direction express routes starting about 15 years ago. Getting the riders to think differently about their trip-making involves unplugging the “way things should be” in favor of the new hierarchy and it still haunts our local transit network planning.
Sadly, the evolving long-distance hierarchy is still skeletal. In addition, RapidRide and Swift are increasingly designed as a more local service rather than a longer-distance one. I have to wonder if what the region needs to do is to abandon the expensive light rail extensions that add little, and instead spend the capital resources to supercharge a more comprehensive (and more limited stop) Swift, RapidRide and Swift system. By “supercharge”, I mean more busways, lane separations and higher-capacity vehicles — along with fewer stops and signal priority that isn’t compromised by short stop spacing..
I suggest thinking of restructuring to have both a solid longer-distance route structure and a healthy shorter-distance route structure as a more useful typology for the Seattle area. A single layer isn’t sufficient for a region our size.
I even wonder if we should be looking at a three-tiered typology. That’s what I see behind ideas like keeping STX 594 or making Sounder or Cascades more frequent. I see a third tier as a growing element even though it’s not explicitly recognized as the third tier.
One systemic change of the last 15 years ago in the Seattle area is the growing evolution of a day-long hierarchical system of longer-distance limited-stop routes, and shorter-distance local-stop routes.
That is largely because of Sound Transit. They are a regional transit agency focused on longer distance regional transit. Sound Transit is also spending a huge amount of money on this while various agencies struggle. It helps explain our relatively low transit ridership despite spending massive amounts of money on transit in the area. The money isn’t really focused on ridership — it is focused on the relatively small number of riders who go long distances.
This again is similar to the ridership/coverage trade-off although it isn’t really “coverage”. This goes back to the comments made above: https://seattletransitblog.com/2025/07/01/a-bus-network-for-access-above-demand/#comment-960141. A ridership focus would put most of the money into more local service (since that is where most of the trips come from). Relatively few people are taking long trips. But ridership isn’t the only consideration. There are other options like making it easier to get to a bus (coverage); good round-the clock service (a key aspect of this proposal); and travel time for a long distance trip. Balancing those is tricky. In my opinion we actually have a fairly good regional transit system. ST Express is quite good. There will be a few minor additions with Stride to make it faster and in some ways better. There are still some key regional service that could be improved (e. g. I would like to see an all-day express bus mimicking Sounder (Tacoma/Puyallup/Sumner/Auburn/Kent/Seattle)).
But from a ridership standpoint we have grossly underfunded the core of potential ridership (Seattle and to a lesser extent the East Side). Meanwhile Pierce County is terribly underfunded and thus every aspect of transit (ridership, coverage, etc.) suffers.
> By “supercharge”, I mean more busways, lane separations and higher-capacity vehicles — along with fewer stops and signal priority that isn’t compromised by short stop spacing..
we could build center running brt for some corridors
“That is largely because of Sound Transit. ”
While ST is notable, both Metro (RapidRide) and CT (Swift) have also been investing in new services that can cover longer distances faster. It’s not just Sound Transit.
RapidRide is not regional-express transit. Every route has standard stop spacing. It works very well for both short trips and long trips. In contrast the old 301 (from Aurora Village to Downtown Seattle) was a regional-express bus. Metro has reduced the number of such buses in part because of Link but also because of the pandemic. Thus you have it backwards. Overall, Metro is investing *less* in service designed for longer trips. Partly because ST is focused so much on that but also because ridership for such trips have shrunk (disproportionately).
As for Community Transit, they have been busy with regional transit. But they always have been. A huge portion of their buses used to go into Seattle (with minimal stops along the way). Now only the bus from Snohomish does. There are still plenty of long distance express routes (201/202, 905, etc.) but my guess is the amount of money spent on express service has gone down (just like it has with Metro). Swift is more limited-stop than RapidRide but not all the Swift lines are geared towards long-distance trips (the Orange Line is not very linear and it makes a bunch of detours along the way).
“RapidRide is not regional-express transit. Every route has standard stop spacing.”
I agree that it isn’t regional express transit. I only said that it was designed to serve longer trips.
And many RapidRide projects have consolidated stops. They may still be spaced in a “standard” way, but they certainly reduced the number of stops in many corridor segments.
The more important point I’m making is simply that our metro area is better served with a hierarchy of transit services. The Metro history of the last several decades has always included them. It’s how we ended up with limited stop routes, a transit center at Northgate and stops off of HOV ramps. Let’s not get sidetracked by semantics.
I agree that it isn’t regional express transit. I only said that it was designed to serve longer trips.
No its not. Hell, the shortest route in our entire system is RapidRide (the G). RapidRide E is not designed to serve longer trips — the 301 is. RapidRide D is not designed to serve longer trips — the 15 is. These serve the same area but the RapidRide is designed to serve shorter trips. Metro still sends a lot of buses onto the freeway but not a single RapidRide. They aren’t designed for longer trips.
The RapidRide routes are just normal routes except for the fact that they have off-board payment, normal stop spacing (instead of the ridiculous “stop every block” that is common in the U. S.) and are sometimes more frequent than you would expect. A large percentage of the riders take these buses a fairly short distance and that is the way they were designed.
Swift is definitely designed for longer trips. It is basically an express overlay and thus requires a regular bus to serve the stops it skips. But Community Transit has plenty of other buses that are designed to serve long trips. The 201/202 are clearly designed for very long trips — longer than makes sense for Swift routes. It can be considered an express-overlay for half the routes in the system. It starts way up at Smokey Point and ends at Lynnwood. If I’m trying to get from Everett to Seattle I would ignore Swift Blue — the 201/202 (and 512) are designed for those longer trips. But the 201/202 is nothing new.
The more important point I’m making is simply that our metro area is better served with a hierarchy of transit services. The Metro history of the last several decades has always included them. It’s how we ended up with limited stop routes, a transit center at Northgate and stops off of HOV ramps.
Fine, and all I’m saying is that we have spent way too much on the long-distance aspect of our system and way too little on the core. The core transit always gets more riders. Yet is has been short-changed. It isn’t because of Metro — it is because of Sound Transit. They are geared towards longer trips (that are taken less often) and they are spending the most money.
It feels like you have found one metric you can use and have convinced yourself that you are brave for being the first to use only this metric rather than asking yourself why others use more than just one. You’ve done what a lot of Silicon Valley tech bros do.
I think you need to think more about a city as a complicated system and so placing a grid on a map doesn’t really serve that system very well. Context is very important when route planning. This is like how some cities in the US and Australia had grids platted by some dude in England who hadn’t heard of a topographical map which is why cities like Pittsburgh have unusably steep streets.
Context is vital for transit systems because cities aren’t just perfectly flat grids with perfect density distribution, ESPECIALLY not Seattle. I think this is one way to approach the problem of system design and I don’t even think it would be wrong to use this as ONE input among many when deciding how you will design a transit system.
As I’ve pointed out before, this proposal is not a grid. To call it that is a disservice to those who argue for grids. It is system focused mainly on reducing overlap. Your main point remains: While reducing overlap is generally a good thing it can be taken to an obvious extreme (as with this proposal).
This is a very ambitious effort with the very earnest goal of making transit a viable alternative to cars in all places at all times. Unfortunately it misunderstands the nature of how cars and transit works. By trying for transit to replace cars for all trips, this plan would actually REDUCE ridership because it gives up some of transit’s strengths in trying to make it competitive where it cannot be.
The key is to understand that in the lingo of transportation, cars offer zero-headway, infinite span, express service. Zero-headway: there is no waiting for your car to start, you can drive off in it as soon as you arrive (note this is different from access, you might have to walk to a garage for your car, but you also might have to walk to a bus stop). Infinite span: you can drive your car 24 hrs a day, whenever you feel like, with no cost difference. Express: the car goes from you origin to destination without stopping or connections, always a one seat ride.
Transit cannot offer these traits for geometric or cost issues. By nature there is space and time between vehicles; a few major cities like NY offer 24 hr service (but not on all routes); in a few rare cases there is express service but otherwise transit has to stop to let people get on and off. The key to making transit successful is to focus where structural issues make transit faster than cars despite cars’ advantages. Usually this is where transit can be faster/cheaper than cars because the cars are stuck in congestion or where the access time to get to parking is much higher than the time to get to a stop. This in turn is usually in high density areas where there are more people than the road can accommodate and intensity of development means parking is limited, far away and/or expensive – i.e. downtowns (it can also be a choke points like bridges where is less space than people wanting to cross. Transit can do this naturally, but it is enhanced with infrastructure (tunnels or bus lanes) to speed through congestion.
Another fundamental is that people perceive wait time as ‘longer’ than travel time. A 10 minute wait then a 10 minute ride will see fewer passengers than a 5 minute wait and 15 minute ride even though each has a 20 minute trip time. A very good thing of this plan is trying to get every line to be high frequency. However, wait time comes into play every time there is a connection where people must get off, wait, and get on again. Connections are ESSENTIAL for a successful transit network (again, express service everywhere like cars is not possible) but people really do like 1 seat rides.
With these considerations, there are several serious things to consider with this plan. As others have noted, identical service all night is a non-starter. The largest busiest cities on earth don’t offer this; most people are asleep at night. Since there is no traffic at night and parking is all empty, there is no way for even the most-frequent bus to beat a car that goes straight to where you want. It costs too much in driver premium pay and wear and tear on vehicles that are empty (if you could even get 1/3 of drivers to work night shift all the time).
Second, the lack of routes to or through downtown. Downtown is both the largest destination and the place where transit has the highest advantage. This is lost if you don’t send most routes through downtown and force everyone onto Link – by creating a connection you increase the wait time penalty. Because ridership is highest, losing 5% of downtown riders would be worse than gaining 50% of outer neighborhood riders. This particularly acute for near downtown trips – making Central Dist. to Belltown a three seat ride via Link or the 14 would be death to ridership. The idea of truncating busses at a high speed/capacity system like Link is a good one, but is always done by using perpendicular routes (i.e running E-W) to feed a rail line (running N-S); or by consolidating busses at the end of a line (which is why Flushing Meadows is one of the busiest NYC subway stations gathering travelers from outer Queens). NY, Paris and London have densest metro networks in their core – and yet all have busses there too.
Third, using only unweighted access is not the best idea. This analysis assumes that the world is isotropic (the same everywhere), but it is not. As noted downtown exists, but there are also many other areas that see higher use/travel demand (urban villages, shopping districts, etc.). This plan generally puts the same level of service everywhere. However, there are a lot of areas where two 10 minute bus routes combining on a key stretch for 5 minute headway is better than two stand alone 8 minute routes. If you are trying to get from your neighborhood to a commercial strip, a 10 minute headway (5 min average wait) isn’t so bad if the total trip is 20 minutes; if you are on the strip and need to make multiple stops, then 5 minute headway (2.5 min wait) is very attractive when the whole trip might be 4-5 min. In contrast, the 8 min headway/4 min wait is a problem if you have to do it twice because the routes don’t overlap and you need a connection to get to the destination. The same 4-8 min wait is also a problem for the very short trips from store to store when it is longer than the whole time spent riding. Ross has pointed out all the places where this network forces connections to keep going straight or reach a destination.
There a number of things with network design than can be done to maximize transit advantages, combining two infrequent routes a few blocks apart into one frequent route is the obvious case – adding 3 minutes walk but reducing wait by 5 minutes will increase ridership no matter how many people complain that their new bus stop is ‘farther away’. Same with stop consolidation so stops are a little farther apart but busses move faster and come more often from not pulling over as much. The general idea of this design – fewer more regularly spaced routes with higher frequency – is strong, but it is too dogmatic in its approach. Cut back night service, pour the service hours saved into strong all-day frequency (with a slogan, maybe “6am to 6pm” covering both commute time and the workday, or “5 to 9” to serve early shifts and after work entertainment) and routes that make efforts to reach and serve high demand areas. The more widely spaced grid is good for outer areas where demand really is isotropic, but an idealized grid is just that, an ideal that only works if you can guarantee every trip can be served with a single right angle connection. Seattle’s geography is not like that.