EPA estimates

[pfibiger]

Include EPA estimates alongside cars in your garage.

It'd be interesting to compare your economy to everyone's economy for a particular model/year, and also to compare that to EPA data. Looks like you can grab all that data for free, not in the best format but probably not too hard to match up a large percentage of it:

http://fueleconomy.gov/feg/download.shtml

[mathowie]

Yeah, that's definitely on our punchlist before launch. Before we get loads of data from other drivers, we always wanted to show if you were +/- from the EPA estimates.

I had no idea they had a downloadable format, that's going to save us loads of dev time incorporating it, thanks for finding it (we were talking about scraping the entire dataset to get our own).

[pb]

Thanks, that's a great idea. Yep, the challenge will be matching up the EPA data with the car data.

[pb]

Their data structure changes from year to year, of course. And sometimes it's a binary excel file instead of a plain text CSV file. And, and, and... ugh.

[pb]

It turns out this data is very hard to match up with the car data we have. For example, a 2005 Mini Cooper has 12 entries in the EPA data. There's the Mini Cooper, Mini Cooper S, Mini Cooper Convertible, and the Mini Cooper S Convertible. All of these come in Manual or Auto transmission—each with separate mileage estimates. This is true across most makes/models. In some cases the differences are types of fuels, 4-wheel-drive vs. 2-wheel-drive, and/or the number of cylinders.

Our car data has some of this engine and body data, but we chose to keep the data entry form very simple. When you add a 2005 Mini Cooper with our form, you have the option to choose convertible vs. coupe, but not the transmission type, and not the S class. (Plus, I'm guessing a large swath of potential users don't know whether they have a V6 or a V8 under the hood, let alone some of the more obscure differences.) Making users choose this level of detail just to get EPA estimates seems like a bad trade off.

So with these 12 records for a 2005 Mini Cooper, should I just average them all to give a "best-guess" EPA estimate? Should we provide an "advanced" car form so folks who know an S from a non-S can fill in that info?

[pb]

I ended up adding a message to car profile pages that says, "There are several EPA mileage estimates available for [this model]. Please identify your model details to get estimates, or skip EPA estimates for your car."

And if you click "your model details", you get a page that lets you choose a more specific model that matches the EPA data. Here's a screenshot. If you click "skip EPA estimates", any estimate features will be removed for that car.

It's not ideal, but I think it's an ok compromise.

By the way we have model/epa estimate data for model years 1998-2008. The data gets insanely sketchy before then. I'm not giving up on using it yet, but it's going to require lots of massage-time.

[mathowie]

huh, pb, I don't see any EPA info on my 2003 Jetta TDI Wagon, and no link to correct it. Is there any info on that model on the EPA site?

[mathowie]

Yep, here it is. Maybe the EPA thing is only shown when you add a car?

[pb]

oops, code typo. If you go back to your VW page you should see the note under Recent Fuel-ups.

My Prius profile has a working EPA Estimates section. I like seeing the red or green numbers to let me know how far off I am from the estimates at a glance. (Even though I could just do the math in my head.) And 60mpg?! I don't think I've ever had a tank at 60, though I bet commuting in stop & go traffic would help.