Many of my students are heavy Uber users.  So I asked them one day, “What do you guess should Uber’s monthly ride number be in the US?”

It turned out to be a fascinating exercise, especially in light of logical thinking.

A good logical-thinking starts from the bottom, and has the following three steps:

Step 1- Establish a base.  We can start with the US total population, which is about 327 million.

Then you might say: Wait a minute, Uber is not serving all over the US.  It is a taxi service, so it mainly operates in urban areas.  Good point there.  We can then do some more Google search, and find that 80% of the US population are considered living in urban areas.

Step 2- Now we need to make some assumptions on Uber’s business.

(1) In those urban areas, how many are currently served by Uber? 10%, or 90%? It is a tough call.  And since we have little idea, let us take the middle road, and call it 50%.

(2) In those areas where Uber is operating, how many are actual Uber users?  After some deliberation, we tentatively put down 10%. Remember the basis is the whole US urban population.

(3) Among the users, how frequently do they use Uber?  There should be large variation here, as some people -like many of my students – are heavy users, but many others might be occasional users. But we are guessing the average here.  But some lively debate, we think it is somewhere between 2 to 4.

Step 3- Put everything together

Now it is just a matter of multiplying the base population in Step 1 with all the assumptions, with the following formula:

Monthly rides volume = US urban population * percentage of population where Uber is present * percentage of Uber users in those areas * average frequency of usage per user per month.

The model is that straightforward, as is shown below:


With the model, we have a number of about 39.2 million.

How close is our guess here? Time to do some more Google search.

According to a Wikipedia article, it is reported that “In October 2016, 40 million riders used the service in a single month and that riders spent an average of approximately $50 per month on the service.”

We got lucky, I have to say, to get an almost spot-on estimate.  But the key is not asking how close we get with the model, but instead asking “Does the model produce a result that is on the right magnitude?”

To that question we can probably say the answer is yes.

What do you think this approach?