Indeed. That was the first thing that caught my eye in the graph since I enjoy using the MARTA system in ATL when I go there and it's pretty busy. It didn't pass the sniff test when I saw that.

I did some quick digging. The term to google to get local traffic without connecting passenger info is "origin and destination", and it looks like O&D is about a third of the overall traffic[0] since there are ~45m total enplaned annually and of that ~15m enplaned from origin. Doubling the enplaned number to get the enplaned and deplaned number would be about 82k daily average. With 10k daily metro riders, 12% ridership seems pretty good.

[0]: "ATL is anchored by a large, local traffic base with 15 million O&D enplaned passengers for fiscal year (FY) 2014." https://finance.yahoo.com/news/fitch-upgrades-atlanta-ga-pas...

I invite you to dig up _comparable_ origin/destination data for all the reported airports, for the different years the transit ridership was available. Then I shall update the graph.

But it's not going to change the fact that ridership of airport transit connectors is low, building expensive rail transit to connect 5K or 10K people is a waste of money. Building cheap rail (i.e. connect to the airport as part of another project) may make sense.

Failing that, you're just going to have to use common sense to obviously realize the Atlanta is not going to do well in this particular comparison. Just like I actually did in the article, where I was being relatively generous with my analysis.

Generous like this table from your article?

                   cost(2013$) riders(thousand)   cost/rider
  O-Train (lrt) 	27M$       14                   1901$
  

...

If you don't understand why your table is extremely flawed then you should pull this article. If you do understand then you are being disingenuous not generous.

I was generous in that I ignored the worse performing data-points according to the heuristic I picked. I wrote that the projected REM ridership doesn't look outrageous, just optimistic, and suggested that placing Montreal at the 80% instead of the 90% percentile may be more realistic. Did you even read the text?

Regarding your unrelated attack, how is it possibly flawed to ask for a transit project to bring a certain number of riders relative to its cost, no matter the mode? This is a standard way to get an idea of the effectiveness of a transit project.

These are back of the envelope calculations that are pretty fuzzy (this is not programming or math, folks), but they are incredibly useful and very easily reveal fishy projects. I wish the public would be more vigilant, even doing very simple calculations like this is helpful -- then maybe in North America we could waste less money on stupid prestige projects, and spend more money on actually building useful transit.

I find your suggestion that I should pull my article extremely unhelpful. We here in Montreal are currently being duped by a semi-private entity to waste a shit-ton of public money on a deeply flawed transit project, while useful projects sit waiting. I want to write about that because I think people should understand how its flawed. If you don't give a fuck about Montreal or Quebec, fine. But don't try to shut me up based on nitpicks that miss the point. If you want to tear down a bunch of bullshit, maybe do something useful and go write about the REM project yourself.

He was pointing out that you mixed your units in your analysis. You divided annual cost by daily ridership for the cost instead of multiplying the daily ridership for annual ridership. I think he was short in how he pointed this out to you. That being said, as the one presenting the data I believe it's in your interests to ensure that what you wrote is accurate. When someone shows inaccuracies, I think it's disingenuous to continue to use data that has been shown to be flawed. That was more tersely expressed as "you should take it down."

I would recommend changing your units in your site to either be all daily, or all annual. This means that you could meaningfully present and compare percentages of ridership, like I did for ATL above.

Getting back to the numbers at hand, YUL has 82% O&D[0], meaning that for 15M passengers there will be (15m * .82 / 365) -> 33,298 daily users. 30% daily ridership does seem like a steep target to hit. If you get the O&D numbers for all of the airports and compare daily values, you may still be able to express you idea and make a case without people being distracted by inconsistencies.

Another factor to add, eyeballing your graph, would be the correlation of metro ridership to the percentage of airport passengers opting into the metro. The numbers for the southern US would probably be explained by a poorer metro compared to SF and the northeast.

You put yourself out there, and you obviously are passionate about your subject, but please take a step back to look at the suggestions and how they will improve your case or show you new insights.

[0] http://www.admtl.com/sites/default/files/RA2015_A.pdf "The percentage of connecting or transit traffic held steady at about 18%, i.e., nearly one in five passengers, or 2.8 million people for the year, had an airport of origin or destination other than YUL."

I couldn't get the O/D data, I tried to find it for Atlanta. Didn't find it, gave up, and posted the graph with numbers that I had. Note the ridership numbers are for different years, so the only reliable matching number I could get was total passengers. It's true that will punish hubs, but I think think as a back-of-the-envelope calculation it's okay. The focus isn't on Atlanta, it's on Montreal, Toronto, Vancouver, New York, Boston - and that's where I see the comparison. Even if Atlanta was three times 'better', it would still be on the right of the graph.

The cost is not 'annual', it's the capital cost expressed in 2003 dollars, that is, adjusted for inflation. I would've never guessed crazytony's issue was with that, because it was just a flat attack without any substance. Giving cost in M, ridership in K, and cost/rider in $ seems fine to me. The numbers are meant to be compared across different projects, and are easily verified to be calculated correctly.

You can't just divide annual ridership numbers by numbers of days in a year, because ridership numbers are given for weekdays, and weekends have wildly different numbers. The ratio can be anywhere from 260 to 365, depending on the actual line... or airport.

I left the annual airport passengers compared to the daily ridership intentionally, because those are not numbers that can be directly compared for various reasons. I intentionally wanted to create an obscure ratio because I wanted to compare across different airports, not give any estimatable number for any particular airport.

In any case, this particular graph is not the overall point of the article, I'm merely trying to get an idea whether they are being crazy with their predictions.

Yep, most of the airports compared have a much higher rate of transfers than Montreal. People rarely transfer in Montreal. Toronto is a much bigger hub.

This is the second click-bait titled article from this author on the topic. It seems more like an exercise in stats, with very little actual context.

Yeah, I was surprised to see Atlanta on there too. Comparing hubs like ATL and JFK and ORD to desinations like PDX and Montreal is not a fair analysis.