Uberdata: The Hidden Cost of Cabs

What up humans?! Bradley Voytek here bringing the #uberdata back! The last post was a nice warm-up.

This time I’m not pulling any punches. It’s about to get real. This is the post I’ve wanted to do since I first started here.

You see, most of the time we here in the uberoffices are pretty happy. We’re happy because we’re building something that makes all of you happy. But every now and then there’s a darkness that comes upon us when someone on the interwebs complains about our cost.

Which got me to thinking. We do cost more than a cab. This is because we are offering more than a cab. In fact, we offer a lot more: reliability, customer support, and, of course, style and comfort. Importantly, we also reduce frustration.

But how does Uber really compare to cabs?

Seriously, when was the last time someone from a cab company helped you out after you tweeted about a bad cab ride? How many times have you called a cab company to request a car, only to have it never show up?

So a few weeks ago I started digging a bit. We all know that Uber is quick and reliable, but we didn’t have a quantitative comparison between our performance and that of cabs.

What I found was pretty amazing.

In this post I explore the hidden cost of cabs and weigh those costs against our performance data.

To start, let’s look at some numbers for cab dispatch times. According to the City and County of San Francisco Office of the Controller 2005 Taxi Commission Survey Report, for dispatched cabs:

• 27% of cabs show within 15 minutes
• 63% of cabs show within 30 minutes

Now it shouldn’t come as much surprise that only 34% of the people surveyed in the report said that they could “almost always” get a cab in San Francisco “in a reasonable amount of time”. (43% said they “sometimes” could, and 23% said “usually not”.)

How do those cab pick up times compare against Uber’s? Let’s take a look:

What’s that?! 94.62% of our cars show up within 15 minutes? And 99.98% show up within 30? Cabs are only at 27% and 63%, respectively? I’m sorry, but that’s just sad.

For the flip-side comparison, 27% of our cars show up within 5.4 minutes and 63% within 8.5 minutes.

Those are our data based upon tens of thousands of rides. But the nice thing about being a data-oriented company is that we can get a lot more detailed.

Here’s our data broken down by minutes until car arrival:

But it gets even better. The data I’m showing you from Uber are for all rides since our founding. But like Travis said in the recent Wired article about us:

“At the end of each day, Uber creates charts to analyze how accurately it was able to predict demand for rides throughout the city versus how high demand actually turned out to be. Using these charts, the company refines the prediction algorithm, so Uber gets better and better at estimating how many cars will be needed in the city on certain days and at certain hours.”

Watch me back that statement up with math.

Here’s a chart showing the percentage of our cars that show up in under 10 minutes, broken down by month:

Every month we get better at making sure you get a car faster.

“But Uber,” you say, “what about hailing down a cab? Sure your cars are faster than dispatched cabs, but it’s so easy when I can just go out to the corner and flag a cab down!”

Sure, that sounds reasonable. But is it really that easy?

According to the report, as one would expect, the numbers for flag downs are better than dispatch:

• 59% of cabs show within 15 minutes
• 88% of cabs show within 30 minutes

But those numbers still don’t compare favorably to ours!

Good job guys!

Apparently, however, only 35% of all cab rides taken in SF were flag downs, so most of you don’t bother trying to flag down a cab anyway!

Also, keep in mind that the numbers in that report don’t take into account the fact that you can book an Uber from work, home, the bar, or wherever before you even have to step outside and look for a cab to hail.

The data from that report are somewhat in contrast to the San Francisco Municipal Transportation Agency 2006 Taxi Availability Study (PDF). They found that:

• 64% of cabs show within 10 minutes
• 82% of cabs show within 15 minutes
• 99% of cabs show within 30 minutes

But those numbers are deceptive, because they’re only for cabs that actually showed up after dispatch promised a car. The SFMTA report found that 35% of dispatched cabs just never arrive!

35%.

Every third time you call, the cab they say is coming to pick you up just straight up won’t.

It gets even worse. That report breaks data down by specific days and times. Weekend nights? Horrible times to try and get a cab:

• 12% of cabs show within 10 minutes
• 22% of cabs show within 15 minutes
• 27% of cabs show within 30 minutes
• 72% (!!!) of cabs never show up

Holy. Crap. The vast majority of the time you call dispatch (if you can even get someone to answer the phone), your cab’s just not ever going to come. And if it does, it’s going to take well over 30 minutes.

Gonna say that again.

(Although according to a report from the Goldman School of Public Policy at Berkeley in 2007, titled “San Francisco’s Taxi Dispatch Service: Improving Reliability and Response”, the numbers reported in the SFMTA report are probably not statistically sound… But they do give us a good place to start. Nevertheless, please note that we’re very grateful for the hard work done by the SFMTA to collect these data. Without their diligent work we would not be able to revel in the irony of using their data like this.)

What’s wild is that Uber doesn’t even really have a concept of “no-shows”. If you open our app and we don’t have a car available in your immediate area, then we tell you that. We don’t make promises we can’t keep!

But to be fair… sometimes taxis get waylaid by crazy assassins. So you know, they’ve got that to deal with.

Okay, okay. Enough rubbing it in. What does all of this mean for you?

First of all, with the numbers from these reports and with those from our ride database I can run some simulations. That is, I can simulate millions of imaginary Uber vs. cab rides using the statistics at hand.

The idea is simple: given our knowledge about how long it takes cabs to show up (and how often they don’t), and given our knowledge of the statistical distribution of our own pick up times, how much time would you have saved, on average, if you’d used an Uber instead of a cab?

The median time saved by taking an Uber instead of a cab in San Francisco is:

• ~4.8 minutes compared to hailing a cab
• ~16.2 minutes compared to calling dispatch
• ~54 (FIFTY. FOUR.) minutes compared to calling dispatch on a Friday or Saturday between 6pm and midnight

Unreal. Seriously I’m not making these numbers up. It’s the actual data collected by the SFMTA compared to our actual pick-up times. Even on Friday and Saturday nights we still manage to pick up 90% of our riders in 13.1 minutes. And hell, the median length of Uber ride during weekend nights is only 10.2 minutes. That means:

Here are some fun things you can do with 54 minutes:

• Fly from San Francisco to LA
• Walk across 1/3 of San Francisco
• Run 6-8 miles
• Watch an episode of Lost
• Read about 13,500 words (a short story)
• Listen to The Dark Side of the Moon
• Build a cake
• Get a massage
• NOT WAIT AROUND FOR A DAMN CAB

Even on normal days, you still have a 38% chance of getting an Uber and getting to your destination before a cab would have showed up. The cab numbers are pretty bad, but my guess is that most of us have had a “good enough” view of them for a long time. It’s not until you experience how much better the process can be with Uber that you realize how bad it can be with cabs.

So yeah, Uber costs twice as much as a cab ride. But the median Uber ride only costs $24. That means you’re paying $12 more than you would for a cab.

How much free time is $12 worth to you? 30 minutes? An hour?

How much frustration are you willing to accept to save $12? What if you’ve got an important meeting, but there’s only a 65% chance that the cab you call will even show up?

What if you’ve got a date on Saturday night, but there’s only a 28% chance a cab will show up?

Forget about the amenities and style that we bring. Those are great. But we’re also giving you some time back, freeing up your day just a little more. Getting rid of just a little bit of hassle and frustration. How much is your time and stress worth to you?

If you do a straight “time = money” calculation, if you earn the median per capita income for San Francisco ($44,373), you would have earned $19.20 in those 54 minutes you wasted waiting for a cab.

$12 doesn’t look so bad now, does it?

Questions or comments on this? Leave them below. For anything else, hit me up on the twitters, my blag (Oscillatory Thoughts), the interwebs, or the emails.