According to David Skok, a general partner at Matrix Partners and a leading expert on SaaS, there are three ways to estimate TAM:

- Top-down, using industry research and reports.
- Bottom-up, using data from early selling efforts.
- Value theory, using conjecture about buyer willingness to pay.

By far, the most actionable way is the second way. Again, in David Skok's words:

*"Bottom-up ... takes the form of “here’s how we price and how many units of that price we can sell.” This is a much better option than #1, because it involves tangible, relatable data on current pricing/usage of the product and imagines a larger customer base. A software startup successfully selling human capital management software at $20 per employee per month might reasonably take the number of employees in its target market and multiply that by pricing to estimate its TAM."*

From a practical standpoint, let's say you are estimating TAM for a ride sharing service business. The first thing is to estimate the unit of measure. In this case, it boils down to a few things.

- How many users are in the chosen market segment who will use this service?
- How many trips will they take on average over a given period of time?
- What is the average cost per trip?

With that, you can calculate an expected TAM in Annual Recurring Revenue (ARR) based on assumptions about the typical user and their usage pattern, as well as a typical trip that they will take.

For instance: Uber is a metered use case where it is by person by trip and each trip is by distance / time / surge pricing. Super complex, right? But you still can boil down to an expected value of trips by making an estimate of what constitutes an average user, an average number of trips, an average trip.

As you can see, the TAM calculation is heavily dependent on assumptions. It is super important to state these assumptions up front so that you know what you are dealing with.

*This article builds on content developed by the Martin Trust Center for MIT Entrepreneurship for MIT's Orbit Knowledgebase and is licensed under CC BY-NC-SA 4.0.*