How I investigated Uber surge pricing in D.C | Hacker News Reader

How I investigated Uber surge pricing in D.C

source.opennews.org

75 pointsdanso11 comments

Not to state the obvious, but surge pricing is simply the most efficient way to motivate people on the platform (read: drivers) to flow into an underserved area (read: where the customers are). The macro take-away on this is that we should celebrate the fact that Uber is trying to incentivize its driving population to serve its riding population by implementing surge pricing. The opposite would be Lyft, which I doubt has any reasonable way other than asking nicely to try to encourage its drivers to venture into these underserved areas...

And yes, I understand that this opinion (read: "interpretation of the laws of supply and demand as they affect reality as it exists") is likely to be unpopular, and am prepared to be labeled all sorts of nasty things and downvoted.

Why did the author choose Washington, D.C. rather than NYC for which there is a public data set of 1 billion yellow cab trips and partial data on 19 million Uber trips? [1]

[1] http://toddwschneider.com/posts/analyzing-1-1-billion-nyc-ta...

I think it's because the author is based out of Maryland...and the historical Uber trip data doesn't provide much insight on surge pricing, at least in the way the API can in real time and by specific vehicle.

If anyone wants to take a look at the TLC Yellow Cab dataset for December 2015 (without downloading the entire file), it can be viewed here: https://putdat.com/WgsjFsT

It's 2GB and ~11.5MM rows. That's about four trips per second!

Besides the author being local, DC has some pretty interesting demographic/geography patterns.

cool! another approach to that task: http://conferences2.sigcomm.org/imc/2015/papers/p495.pdf

> This was a practical decision made to reduce our story space, as there were some Uber products that had very different surge pricing patterns, and including these would convolute the narrative we were telling.

There was data that contradicted your narrative so you excluded it? That Is Not Science.

'convolute' != 'contradict'

Convolute means (in this context) to making something more complex to explain or difficult to follow. It does not mean that they excluded data that contradicted their analysis.

In this case, they focussed only on UberX since the surge pricing mechanism is different for the other Uber products - there is nothing wrong with that.

The thing that's wrong with it is that they have a narrative. If you take a large enough data set and keep applying filters to it until it shows the result you wanted, you can always get it to. That doesn't prove your hypothesis, it only proves that you can solve a set of linear equations.

It seems like important information is missing with regards to their conclusion.

Their conclusion is "But how effective can they be if they fail to reach 15% of a population?".

However, if they are excluding all other Uber services other than UberX from their analysis, then that's problematic since those other services could be covering that 15% of the population.

Do the totality of Uber services provide coverage in the areas that this analysis highlights? I don't know, and given the analysis as written, I don't know how anyone else can either (without doing another analysis). So it seems at least worth considering whether convoluted vs contradicted is the case since we're missing a key piece of information.

"But how effective can they be if they fail to reach 15% of a population" was in reference to the fact that almost 15% of the DC population doesn't have a bank account which the author took to mean they can't use UberX or any other Uber service. This isn't strictly true but the barrier to entry without a bank account is high enough to dissuade most of this population from using Uber services.

Reply on news.ycombinator.com