On Market Sizing

[How Many Drunk Students Uber in Ann Arbor?, More Numbers Sound More Right, Drowning in Stats, Avoiding Bad Questions For Health Reasons]

2025.03.16

XIC

In a lot of ways, reasoning through a problem that you don’t know the answer to can just make your guess sound better. 

My current working solution to not having a “good way” to answer such questions is to avoid them altogether.

First, some housekeeping.

[How Many Drunk Students Uber in Ann Arbor?]

My freshman year in college, I was interviewing for a professional fraternity that I ultimately would not get into*, when I had one of my first close encounters with the concept of “Market Sizing.”

I was asked how many Uber rides were taken in Ann Arbor on a Thursday evening. While this is a very hard thing to know, the task of market sizing is not to know such things with certainty–the task is to come up with a believable & reasonable estimate. 

To do this, a good starting place is to grab facts you know and transform them to the answer by leveraging assumptions or other facts.

In my case, I began strong with the total population of Ann Arbor & the total number of students. 

Then, I quickly revealed where my mind spent most of its time by beginning to estimate what percentage of students would be drunk on a given Thursday, verbally debating if a higher or lower proportion of minors would be drinking than upperclassmen. I think I was heading in the direction of intoxicated people having a higher probability of ubering, but in retrospect, this was likely too granular a detail to rapidly come up with a strong & convincing answer.

It was a moot point, though, because the exercise was derailed by the interviewer yelling at me for bringing up underaged drinking in a “professional setting.” I was no Don Draper & ended up apologizing for being inappropriate.

I share this anecdote & footnote to reveal a source of potential bias with regards to what follows.

*I would later be told by confidants that my potential admission was a “divisive” topic among the members. I was also told that my strong disagreement about the motivation of a Charles Bukowski poem with an individual at one of the recruitment events was looked at quite negatively–distance obscures, but I believe I said “No, that’s wrong” to the assertion that Bukowski was a closeted feminist. 

[More Numbers Sound More Right]

Our query last week about what percentage of Italian Americans (IA) ate cannoli in the last year is effectively a market sizing problem (from here on out, we shall refer to it as the IACC question). We want to know how much a particular group does a thing, the thing often being consuming a product or service.

Without a definitive answer being easily available, how would we get there?

My knee jerk reaction is that we would want to start with facts and make as few assumptions as possible to get to the desired output. 

Going back to the Ann Arbor Uber problem, maybe we’d try:

population * (% who I think took an uber on a Thur)

The catch is, if I’m trying to convince you I’m right, that doesn’t do much for my case. Why do I think that percentage took an Uber? 

Let’s expand that equation a bit:

population  * (% who “go out” on a Thur) *  ( 1- % who prefer driving or carpooling  - % who prefer walking - % who prefer public transit)

That looks like a bit more of a believable equation, but it does have a lot more assumptions. Anywho, why stop there?

(gen population * % gen pop who “go out” on a thurs * ( 1- % who prefer driving or carpooling  - % who prefer walking - % who prefer public transit) * (avg uber ride per person)) + (student population   % student pop who “go out” on a thurs * ( 1- % who prefer driving or carpooling  - % who prefer walking - % who prefer public transit) * (avg uber ride per student))

Here, we further broke out the general population from the student population, as they likely have vastly different behaviors. We also added average uber ride per person to account both for people who take multiple ubers & those who take one with friends.

We have to make a LOT more assumptions, but it will sound more right if we do.

The thing is, that doesn’t make it more right, and perhaps that’s the problem. Depending on the assumptions we include, we could get vastly different answers, and I’m pretty confident that I could give convincing natural language arguments for either 5% or 20% of students preferring walking.

I suppose the benefit to breaking the problem out more is to give more “explainability” to it. That way, we have more “small assumptions” that can be argued about and examined rather than one big assumption.

Still, even in the absence of a definitive answer, I’m not at all convinced this is a “good” path. Yet, it is the sort of assumption heavy deduction employed by professionals everywhere &, in varying scales, many people in day to day life.

[Drowning in Stats]

93% of people are more likely to believe a claim when a number is associated with it.

Just kidding, I made that stat up, but you probably thought I didn’t.

More broadly than the issues with market sizing, people really do employ numbers to make themselves sound more right all the time.

I take this point to be pretty self evident so I won’t belabor it, but what sounds more right to you:

“In a recent poll, 63% of Americans said that they liked cannoli.”

“Most Americans think that cannoli are good.”

It doesn’t matter if the poll was conducted by an undergrad in a college dormitory with a sample size of 50… unfortunately, it “feels” like it’s better than nothing.

[Avoiding Bad Questions For Health Reasons]

For health reasons, my solution to not knowing something is to not answer it.  

Having spent time in financial markets, I cannot begin to articulate the amount of people confidently using such “reasoning” or stats to be confidently wrong.

Perhaps this was an impediment to me in financial markets, but I have something of an unwillingness to comment on uncertain things that I have no control over. While I’d be happy to discuss the fundamentals of a business, if anybody asked me a question like “will the stock go up?” I would not answer the question. Of course, I didn’t know with certainty, and it is exceedingly difficult to communicate nuance.

Since “I don’t know” isn’t an unacceptable answer in finance, I eventually replaced it with “I bought it myself,” which communicates perhaps the only thing that really matters–skin in the game. Still, I would qualify it with, “That doesn’t mean you should buy it, I’m hedged and you are not.”

There is an urge to just give some facts and smart sounding assumptions and a conclusion and take the conclusion as “certain enough.” However, every assumption you adds reduces the certainty.

Fortunately, the product we are building now at BirdDog, perhaps consciously or unconsciously, is designed to still provide value even when their is a high level of uncertainty. If a user asks 15 questions about 100 companies, you have 1500 shots at finding something and can afford to say “I don’t know” a few times.

Live Deeply,