King County Metro Route 38 // Photo by Oran
Within the last month, King County Metro quietly released the Fall 2010 route performance report, an annual public report that attempts to objectively distill the performance of a route down to a handful of numbers, reasonably understandable to the lay person. While the intent of this year’s report is the same as before, the metrics and categories used have changed significantly, and in this post I’m going to examine these changes in depth.
The simplest and perhaps most significant change is that the concept of “subarea” does not exist. Rather than the old East/South/West division, the performance tables are now divided according to whether a route does or does not serve the “Seattle core”, defined to include the CBD, First Hill, Capitol Hill, Uptown and the U-District. This does not imply that service is going to be allocated according to some arbitrary formula based on this criteria; rather, this division allows more reasonable comparisons and ranking among Metro’s many routes, as service that avoids the densely urbanized (and traffic-snarled) Seattle core will typically exhibit very different performance characteristics from other service, however well or poorly those routes are designed.
The old reports, such as this 2009 report, relied upon a smorgasbord of individual metrics:
- Rides/revenue hour: A “revenue hour” is an hour that a bus is in service and accepting passengers; contrast this to a “platform hour” which is any hour when the bus is outside the base. This metric heavily favors dense urban areas, where short trips and a constant churn of passengers are the norm. Because it considers revenue hours only, it fails to account for the efficiency or inefficiency of the route design, a flaw that a rides/platform hour metric would not have suffered.
- Farebox revenue/operating expense (FR/OE): Fairly self-explanatory. This metric does consider the full cost of operation, but conflates it with the amount paid by riders and expresses it in a way that makes it hard to compare directly along with other metrics based on hours.
- Passenger miles/revenue hour: Just as rides/revenue hour is skewed to urban service, this metric tells as much about the character of the route as the quality of the design; favoring long freeway-running suburban commuter routes that load up downtown, then drive far and fast before unloading at a park & ride.
- Passenger miles/platform mile: A platform mile is analogous to a platform hour: a mile traveled by a bus after leaving the base whether deadheading or in service; so this is essentially the average occupancy of the bus, including those miles when the bus is deadheading and its occupancy is presumed to be zero. This metric properly accounts for the full cost of service and efficiency of route design, as well as providing a fair comparison of routes with differing lengths and running speeds.
- Route effectiveness sum: Perhaps the most misleading of all, this number was a sum of four other numbers derived from the comparative performance of the route on each other metric. The fundamental problem here is that a bus route cannot be reduced to one “effectiveness” number: some routes, depending on the built environment and economic context they serve, will do better on some metrics than others, and that may well be perfectly okay; every route has to be evaluated with that context in mind, against an appropriate selection of peer routes.
I think it’s clear that, while the goal here is laudable, and each metric has some value, there are a number of problems with this set of metrics. They are not consistent or comparable in scope, painting a picture that is somewhat incomplete yet that also contains extraneous and distracting information. The 2010 metrics are simultaneously simpler and more complete:
- Rides/platform hour. Still a metric that favors urban service, but now includes the true cost of the route including deadhead and layover time.
- Passenger miles/platform miles, identical to prior years.
- No effectiveness score is presented, but as in prior years, individual metrics that are in the top or bottom quartile are highlighted.
I think the results these metrics yield are very sensible. Routes on the Downtown-Belltown-Uptown corridor are still shown to be perhaps the hardest-working routes in the county. Routes known to be little-used, such as the odious Route 42, consistently show up in the bottom quartile. The reader is invited to look up their neighborhood route and comment on how accurately they feel the report quantifies its ridership. Further simplifying the discussion, the report no longer breaks out turnback and shuttle variants, a distinction that is not of interest to most readers.
In closing, I note that page 11 of the report contains an intriguing typo: Route 71 is described as “Wedgwood and U-District via Latona”, an obviously incorrect description, that nonetheless indicates where Metro’s planners might be going with this route in the next big service change.
I’m trying to grok the meaning and relevance of the passenger miles/platform miles metric. How is it really a measure of efficiency?
I had occasion to take a 106 from Rainier Beach Station to Rainier and Henderson and then observed an operator exchange at that stop. The new operator was waiting with a county decaled car. I’ve never observed this for any other Metro route (but of course I’ve only experienced a small percentage of the route structure). Is this “normal?” How would such an exchange figure into the metrics in this report since the bus continues to be outside the base despite a change of driver.
I recall that the CTA routinely changes drivers mid-route and it seems some of them get to their exchange posts via transit. There are also rest points where some of these exchanges occur where the buses are not held on a layover. They keep moving with different drivers.
Are there operational efficiencies that could be had by reducing the need for buses to return to base at shift changes? How would that affect this metric?
You’re referring to ‘Road Reliefs’, where the bus stays out on the route during shift change, and the driver takes a ‘base car’ out the meet the bus. This is more efficient.
In Seattle, most road reliefs occur downtown in the CBD, and the drivers catch a bus into downtown to make the change. I used to enjoy the walk everyday from Atlantic Base to somewhere along 3rd Ave.
I’ve observed several such changes of driver on the westbound 3 at 3rd and James and the eastbound 2 at the library.
In answer to your first question: It’s a reasonable measure of a routes efficiency, taking into account number of riders, distance traveled and includes deadhead miles (Platform miles are from the time it leaves the base until it returns).
So double the load, the number doubles.
Double the distance, the number is cut in half.
Increase deadhead miles, the number goes down even more.
Using a fixed cost per bus hour, it’s also a good way to estimate route cost per rider.
This really puts premium express service in the spotlight, as they no longer get the benefit of a free deadhead trip back to the start point.
I was really surprised how well a lot of the peak-only routes performed on the (Passenger Miles / Platform Mile) metric — I would have thought that with all the deadheading they’d perform worse than other routes with similar (Boardings / Platform Hour). Maybe this is because the peak commuters tend to have longer (and maybe faster) trips.
This metric basically measures how full the buses are on average, including dead-head time, so it’s sort of an energy efficiency metric. The other metric, boardings per platform hour, is probably pretty close to a financial efficiency metric. Although neither takes into account the additional fixed costs needed to operate peak services.
Metro has fare recovery data, they just have chosen not to release it here.
Maybe it means Metro is being efficient in how it routes and stores the unidirectional expresses.
One further step would have been a big help to equate route performance with areas of service.
The two maps are kinda pointless without route numbers. I suppose they were trying to show how service is spread over the city and county, but the lines could have been color coded to match the top and bottom 25%. A clear picture would result in areas that perform poorly, and those that shine.
The threshold between city and county routes is different by nearly 2:1 which accounts for density. I wonder how Metro will tell how much a route is costing, as the buses generally travel more miles during a platform hour in the county than in the city. As most costs are measured by the hour (salaries, maintenance cost, admin overhead), that should equate to bus cosst per passenger is somewhat equalized for comparison between city and county. Stated differently, a bus with 30 riders on board can travel 10 miles in the city in one hour, but maybe go 20 miles in the same hour in the county. The hour costs Metro about the same regardless of speed, except for fuel and increased wear and tear.
I’ve also observed huge numbers of buses dead-heading between the eastside and Seattle. Why is this necessary? It would make sense to me to keep the buses on the side of the lake at the end of the peak period and utilize them from a base on the side they end up on or store them until the next peak period and then shuttle the drivers as necessary to their bases.
How much of this is buses running two or more peak period trips?
I’m not sure. Why wouldn’t they run revenue service in the opposite direction?
It’s cheaper. You don’t have to stop and wait for fare fumblers, and there’s no schedule to break. If somebody’s car blows up in the middle of 520 and routes are delayed 20 minutes because of it, you don’t have people waiting forever in reverse peak.
That said, you’re allowed to ride on Metro buses as they return empty to base.
Also, Metro sort of does this with some North Base routes. Operators grab the coach from North in the morning, run something downtown, park it at Central/Atlantic, and take a shuttle bus (route 661) back to North Base. In the evening, some other operator rides the 661 downtown and reverses the processes. There used to be more 661 trips than there are now.
A lot of deadheading buses on 520 run people between Montlake, Evergreen Point, and Yarrow Point (mostly reverse commuters on bikes, like me… who else would regularly make that trip along with deadheading buses?). It’s supposed to be a free service to help cyclists without many good options get across the lake, but I tap my ORCA anyway if the reader is turned on (sometimes it’s turned off because the bus was running some sort of irregular service).
To my understanding, Metro no longer allows passengers to ride buses deadheading between a route terminal and the base.
This was relayed to me by a rider information specialist a few weeks back when I was trying to get from the Tri-Cities to Seattle, a trip that cost me 20.00 more than it used to because the last 209 of the day had been deleted.
As to the trolleybuses on Broadway, those are a special case as they are considered to be “in service” to 5th/Jackson.
If Mr. Latteman or somebody else would like to interject and correct this, please do, and be so kind as to e-mail me a letter I can present to drivers if this becomes an issue in the future.
Brian Bradford
Is the top quartile and bottom quartile for the two lists calculated together or separately?
Separately it appears. Also, there are separate break poinrs for the three time periods.
One more observation, the two maps point out the duplication of services offered along corridors. Many of the lines on both maps both go downtown and don’t go downtown.
Go figure.
The #82 and #83 night owl routes make a compelling argument for all-night service in some neighborhoods.
So naturally Metro seems to want to kill owl service, if previous reports are to be believed. Perhaps I can hope that meant only ending the separate owl network and running the same routes as during the day…
There’s a difference between wanting to kill owl service and having no choice but to kill it, as outlined in the doomsday 600k hour cut scenario which applied the new service guidelines.
I do think it’s time to overhaul owl service, which hasn’t changed much in over 50 years.
@Oran: Now that you mention it, it’s pretty obvious just from looking at the night network that it’s stale. Areas like Lake City and Northgate that draw lots of bus riders today aren’t served at all. I would guess (and I should stress that I’m not an expert or anything like that) that service to those areas, however it was provided, would perform better and serve a greater mobility need than, say, the 84.
Agreed. Owl service is horribly broken, and incomprehensible for most riders. I am amazed that anyone at all makes use of the 280. Yet, there’s obviously demand for it, as 82 and 83 (Greenlake via Queen Anne, U District via Eastlake) both manage to be in the top 25% for pasmi/platmi.
So how do we fix it? The 84′s service area seems to be the place to start, as it has the worst ridership of all, and the most confusing routing. But what do we do? Replace it with a couple run of the 48s, and just cut off Madrona & Madison Park entirely? Redraw the route to preserve those neighborhoods but eliminate the confusing loops that narrowly miss some high-volume stops?
Bleh. What a project.
We fix it by making RapidRide/Link into 24-hour service, starting today (with the predecessor bus routes). The 81 gets replaced by extra runs on the 15 (D); instead of the 82, we run the 358 (E); instead of the 83, we run the 66, possibly with a Capitol Hill detour (Link); and instead of the 85, we run the 54 (C).
That leaves the 84. But the TMP mentions that the Madison St BRT should get RapidRide branding. So for now, instead of the 84, we could run the 11.
For social equity reasons, I think we’d also want to run the 2 and/or the 3 (or maybe even a hybrid, where the 3 continues from its terminus along the 2′s route). Or maybe we could run the 48S and call it a day.
The 81/82/83/84/85 have apparently not changed since the 1950s when Seattle only went to 85th Street. I’m stunned there’s never been any extension. It all comes down to money, but still, you’d think they’d have found the money sometime in 40+ years to reach Aurora Village, Northgate, and Lake City. The only thing that has changed in the night owls since 1980 is the addition of owl runs on the 49, 120, 124, and A. The southbound 44/43 is “almost owl”, with the last run leaving Ballard at 1:54am, arriving Broadway at 2:30, and then going down lower Broadway to the base.
West Seattle is likely to lose both the 85 and one or two daytime routes, but it may get them back in night owl on the 54/C. (The 120 already has it.)
People ride the 280 because there’s no other choice. It would take some two hours to walk across the Mercer Island bridge. You can’t walk across the Evergreen bridge, and it would take two more hours to walk from Montlake to I-90. (Some poor people do walk across the Evergreen bridge, but the sidewalk is so narrow it’s extremely dangerous as well as illegal.) Plus once you arrive in downtown Bellevue you may have another hour’s walk ahead of you. The 280 can also be useful if you have a night job in Renton, although most night jobs are in Kent which is another hour’s walk.
I would expect people who work very late hours and don’t drive to either carpool or move close enough to work to bike, rather than depend on transit.
People who work late hours often are low-income folk don’t have a choice – consider restaurant or retail employees working the closing shift. If they have access to a car at all it’s usually driven illegally, without insurance. Moving to walking distance is typically not a viable option for people in this income range, because they’re either in a shared living arrangement with a group of friends/family, or locked into one of a few low-rent neighborhoods. Even if moving is feasible, getting 1st/last/deposit together is no easy task for someone spending well over half their income on rent.
Bicycling is totally a legitimate option, and some low income workers try it. But the reality is that unless you have an indoor place to store it, you cannot keep a bicycle in this city for any long period of time before it is stolen, and that is the end of most wage-earners flirtation with bicycle commuting.
So people either just deal with the shitty late night transit service, or give the lucrative late shifts to an employee with an uninsured car.
I’ve been in the industry for upwards of 10 years now, and it’s always the same.
I suspect the choice of colors (Black for Bad), is somehow related to being on the black list for service elimination down the road.
Not necessarily. While I haven’t done any digging, there are some routes that don’t deadhead. For example, the old 915–Auburn Station to Enumclaw–was always connected to some other route; often the 152 during peak (off peak it’d connect to itself). Even though it’s a ways from South Base, it never deadheaded. It’s a route that would perform poorly based on rides per any hour, but if you had to deadhead it, it’d be even worse. I don’t think that’s the case any more; it has been renumbered to 184 if you want to do some digging on block trips on OBA.
While definitely the minority case, there are also a sprinkling of South Base routes that are connected to route 600 trips. I don’t think the deadheads would count there. But I could be wrong.
If a route doesn’t deadhead by virtue of its route design, then revenue hours == platform hours and the riders and hours that would otherwise be ignored are charged to the second route. I don’t see how this undermines my point that platform hours are a more accurate measure of the cost of a route.
I massaged the “Seattle Core” routes into a spreadsheet, which is here. I’ll do the non-Seattle Core ones in a minute. Hopefully you can “Save As” it and sort by whatever you like?
Non-core is here
So… the “top 5″ in various categories:
Peak Boardings per Platform Hour (Queen Anne folk sure do board a lot of buses per hour!)
1. 4N, 74.8
2. 1, 72.5
3. 3N, 70.0
4. 13, 69.3
5. 48S, 67.2
Peak Passenger Miles per Platform Mile
1. 358X, 18.8
2. 18X, 18
3. 301, 17.5 (ironically, this is more of an express than the 358X or 18X)
4T. 120, 17
4T. 150, 17
Off-peak Boardings per PH (a Queen Anne sweep!)
1. 2N, 74.5
2. 13, 69.5
3. 3N, 69.1
4. 4N, 65.8
5. 1, 64.7
Off-peak Passenger Miles per Platform Mile
1. 358X, 20.5
2. 120, 20
3. 101, 18.8
4. 124, 18.7
5. 9X, 17.3
Night Boardings per Platform Hour (UW students ride the late buses!)
1. 67, 43.9
2. 49, 41.4
3. 73, 37.9
4. 72, 35.3
5. 2N, 34
Night Passenger Miles per Platform Mile
1. 358X, 15.4 (The 358 wins the “most full bus” category at all hours of the day!)
2. 120, 14
3T. 73, 12.9
4T. 101, 12.9
5. 150, 12.7
No non-core routes came particularly close to cracking the top 5 in any category.
The 67 is an interesting showing in that list. UW-Northgate via Roosevelt is more popular than any of the downtown buses. Of course, the downtown-UW riders are split between 3 different routes, so that number gets diluted somewhat.
One of my only complaints about this document is that it doesn’t list any data for shared corridors or frequent turnbacks. The frequency chart lists a number of corridors where multiple buses combine for shorter headways, such as the 2/13, 15/18, 26/28, and more. It also lists a number of corridors where a turnback routing is so common that one segment of the corridor has much more frequent service than the rest, including the 12, the 41, and a bunch of non-Seattle routes.
I’d love to see the performance numbers for these segments. I’m guessing that almost every single one would be at the top of the list, but it would be interesting to see by how much.
I’m sure you know this, but the official definition of “express” is “skips any stops”. The 358 qualifies because it misses a few stops on Aurora south of the ship canal. The 301 has its own routing, and so does not.
This does lead to some perverse labeling. But from the perspective of a rider boarding a bus, it definitely seems like the right distinction. If you board a local bus, you know it will make all the stops. If you board an express, don’t be surprised if your stop is skipped.
One thing about night routes… I’m not sure how they define “night” for these purposes. The 67′s last outbound trip of the day leaves campus at 9:46, so it doesn’t have super-late trips dragging down its numbers. The 72 and 73 have trips leaving downtown at 1:10 and 11:59, respectively; the 49 has a trip leaving downtown at 2:15; the last outbound 2N trip leaves downtown at 1:05.
The “real” night buses (8x routes and similar) have outbound runs leaving downtown at 2:15 and 3:30, and nothing else. So they can’t really be compared to the 67, or even the 49, which has a lot of other trips included in its “night” category.
Bruce mentioned earlier that much of the ridership on the Queen Anne routes doesn’t go north of Uptown.
This does show the problem of rating by route rather than by street-segment. The 71/72/73X would overwhelm all other routes if they were a single route. The 48 gets its distinction of highest ridership only because it’s so long. On the other hand, the charts do show which routes are out of balance and need asjustments, but they don’t give any indication of what the adjustments should be.
The first thing I noticed, when looking at this data, was that off-peak was consistently higher performing than peak.
That seemed odd. So I looked a bit deeper.
It turns out that almost every individual route is higher-performing during peak than off-peak. The main exceptions are routes that have peak express variants, like the 2N, 5, and 48N, which cannibalize a good portion of the peak ridership. So why is off-peak ridership higher, then?
The reason is actually very simple: there are a number of peak-only routes with no full-time counterpart! Almost all of these routes have terrible ridership. They bring down the numbers for peak, but have no impact on off-peak.
I would love to see Metro release a new version of this document with the following changes:
- For both categories (serving Seattle and not), split them again into all-day destinations and peak-only destinations, so that there are four tables total. (Obviously, two of the tables would not need columns for off-peak and night.)
- All buses which are scheduled together should be listed as a single entry, and the combined number should be used for computing the top/bottom numbers. The routing variants can then be listed underneath. For example:
48S … …
- local …
- express …
Or:
71/72/73
- 71
- 72
- 73
- 71X
- 72X
- 73X
I think that would paint a more accurate performance picture.
For the most part, I agree. However, there are a few turnbacks which are so frequent that they’re listed separately on Metro’s frequency table. I would have loved to see the performance of the turnback variants of the 12 and the 41, for example. That seems every bit as meaningful as data on the 2X.
To avoid cluttering up the data (both visually and numerically), any important variants — and I’m including the expresses here — could have been listed as subcategories of their main route. Thus, we’d have one performance number for the whole route (every variant), and separate numbers for each variant (which would not get averaged into the percentile brackets).
Also, on that note, I think it’s interesting how badly some express variants performed. There seemed to be a direct correlation between how much the express route diverged from the regular routing, and how well it performed. This doesn’t surprise me at all; it’s worth waiting an extra 5-10 minutes for a 15X instead of a 15, but that’s probably not true for the 7X. So if we need to save money, getting rid of these same-routing-fewer-stops variants seems like a great place to start.
My favorite term for all of Metro’s perverse system of variants is “Route Mutation”
Next in line are “Fare Mutations”
We usually use revenue hours (which usually includes layover as well as on-street time) instead of overall vehicle hours because the deadhead time tells us nothing about the productivity of the average local route (express routes would be different). Excessive pull-in and pull-out mileage says more about the poor location of bus garages than anything about the route itself. Marking down a route because it has excessive deadhead mileage (again, this is only for local routes) ignores the possibility that the deadhead time was less than the layover time would have been if there was no deadhead; in fact, this aspect of Hastus, known as MinBus, has resulted in great savings to transit agencies.
Interesting that the 8 was counted in the downtown core group.
It’s not “downtown core”, but “Seattle core”, which is defined as any route that touches Downtown, Uptown, First Hill, Cap Hill, and U District.
So you also get the 30, 31, 44, 46, 48, 67, 271, and surely many more, that don’t go downtown (some of which perform really well, some of which are not so hot).
After looking at the data more, I think that their use of “Seattle core” isn’t quite optimal.
First, I think they made a mistake in not counting Northgate as “Seattle core”. Look at the 345-348; those would be good routes by Seattle core standards, and by “non-core”, they’re almost the top performers.
Second, their definition leaves out routes like the 128, 140, RRA, and the 230/253. These are some of the top-performing routes in the city.
And finally, Seattle Core includes all-day routes like the 25, which perform poorly by any standard.
I’ve said this elsewhere, but I think the only useful distinctions to make are frequency/span. I’d like to see one table with all of the peak-only routes, and the other with the all-day routes (i.e. everything else). It could also be useful to have a third redundant table, with the data for all of Metro’s frequent all-day routes and/or segments. That way, you could see at a glance how much better Metro’s frequent network performs compared to the other routes.
Honestly, if I lived on the Eastside or in South King, I’d be upset at Metro for the organization they used. RapidRides A and B are two of the most important routes in Metro’s network, so why are they being relegated to the B-list? Does Metro expect these routes to never perform as well as their Seattle counterparts? (I sure hope not, because if that’s true, then RapidRide A/B were a waste of money.)
I’m surprised the Seattle-core routes and suburban-downtown expresses are on the same chart. A Seattle-core route goes to North Bend? Does it stop in SLU and First Hill as well as downtown, to give some semblance of being able to ride it from one part of Center City to another?
The in-city routes and suburban-expresses are such different markets that I’m not sure it’s fair to put them in the same chart and combine their “top quartile” and “bottom quartile”. On the other hand, I’m not sure which kind of route benefits by listing them together. Certainly, it’s surprising that some of the suburban-expresses get into the top quartile, including the 102 which is a favorite target of many STBers.
As I mentioned above, it’s not just the suburban expresses. All the peak-only routes belong on a separate chart.
I disagree. The suburban-expresses are different from the 45 and 46 because they go a much longer distance.
The reason I want to separate peak-only from all-day routes is that they serve completely different markets. To a first approximation, the people who take the 45 are commuters; they head to UW in the morning for school or work, and head back to QA in the afternoon. That’s the same as the suburban expresses. And it has nothing in common with a route like the 48, for which commuters are only a part of the ridership.
More importantly, including the 45 with the 212 does not pollute any numbers. But including the 45 with the 48 does, because it artificially deflates the peak productivity brackets for those routes.
By any measure, the 45 is a bad route. But comparing it to the suburban expresses will show that in a useful way. Comparing it to frequent arterial services will not.
Oh, and FWIW, the 46 is not a peak-only express. It’s just a route in Seattle with the frequency you’d expect for a route in North Bend.
Most peak-express riders are commuters, but at the same time, most peak-hour riders on every route are commuters. For instance, I’m going to take the 273 today to the Shoreline open house, and a CT-Lynnwood route on Thursday, and once I rode Sounder to Tacoma in the morning. These aren’t commute-to-work trips but they happen to be when the peak-expresses are running. If I were travelling another time I’d have to go a roundabout way that would take two or three times as long.
The 46 has added some midday runs, but the 46 reverse-commute (an almost separate route from Ballard to Shilshole northbound in the mornings and southbound in the afternoon) is peak only.
It’s of course true that most peak-hour riders are commuters. But in my mind, the fundamental difference is that some destination (or destination pairs) *only* have service during peak, while others simply have *more* service during peak (or just as much, or even less, in some bizarre cases).
If you’re not a commuter, then the 45 is useless to you. (And let’s stop talking about the 46, since it’s useless to everyone. :D) That’s the same as the suburban peak expresses. And it’s very different from frequent arterial service.
That’s the distinction I really want to make. There’s the all-day network, and then there’s the peak-only network. That’s a fundamental difference. It’s useful to compare commuter routes to other commuter routes, and if the 45 doesn’t perform well (because it’s such a short distance), then so be it. It’s not useful to compare commuter routes to arterial routes, even if the latter’s peak ridership is largely commuters.