Unfreeze the TAM Ceiling
Never cap penetration at the historical norm — a genuinely better product grows the pie.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 92%
Gurley gives three instances of the same error. In 1999 he recruited a public-company CFO into OpenTable; one day the CFO said he was quitting because his model said the business would never work. Gurley asked to see the model and found penetration frozen at 17 percent in every city. Asked why, the CFO said no one gets more than 17 percent market share in any of the businesses he had worked with — a benchmark imported from retail. Gurley's reply was that because he believed the network effects, they were going to 99, not stopping at 17. The CFO quit; OpenTable went on to file an S-1 while throwing off cash. The second case is Aswath Damodaran's public valuation of Uber at two or three billion dollars, which took the taxi market as the upper limit. Gurley wrote a reply called How to Miss By a Mile, having already observed that Uber in San Francisco was 20x bigger than the San Francisco taxi market, because Travis and the team had built something 10x better and more available than a taxi. The third is McKinsey being hired to size the global mobile phone market and returning 900,000 units as the upper limit. Gurley's conclusion: TAM conservatism hurts you more than it helps you as an investor.
Origin
The framework crystallised from Gurley's 1999 argument with OpenTable's incoming CFO over a model that froze city penetration at 17 percent, and was reinforced by his later public rebuttal of Aswath Damodaran's taxi-market-capped valuation of Uber.
Core principles
- 01Taking the incumbent market as the upper bound is the single most common market-sizing error.
- 02A product that is 10x better than the incumbent changes usage behaviour, so the existing market is a floor and not a ceiling.
- 03Analysts import penetration ceilings from their previous industry without noticing the structural difference.
- 04In network-effect businesses, penetration can approach total rather than settling at a competitive equilibrium share.
- 05TAM conservatism hurts an investor more than TAM optimism does, because the misses are the outliers.
- 06When something is genuinely disruptive, the optionality to build on top of it is not in the model at all.
How to run it
- 1
Interrogate every hardcoded ceiling
Open the model and find the assumptions that are frozen rather than derived — penetration rates, share caps, terminal values. Ask the author directly where each number came from.
Pro tip Ask why did you freeze it at that number, rather than debating whether the number is right.
- 2
Trace the assumption back to its native industry
Establish which industry the benchmark was learned in and whether that industry had the same structure. OpenTable's CFO brought a retail benchmark into a network-effect business.
Pro tip Experienced hires import their priors invisibly; the more senior the hire, the harder the priors are to see.
Watch out This is not a reason to dismiss the person — it is a reason to surface and test the specific assumption.
- 3
Measure how much better the product actually is
Quantify the improvement over the incumbent experience in multiples. Uber was, in Gurley's framing, 10x better than a taxi on convenience and availability.
Pro tip Convenience improvements expand usage frequency, which is what makes the market grow rather than transfer.
Watch out A 20 percent better product does transfer share rather than growing the pie; the argument only holds at large multiples.
- 4
Model demand expansion, not just share capture
Assume the pie can grow. In the Uber case the number of black cars in San Francisco went from a few hundred to a thousand plus because rider demand pulled supply in.
Watch out Do not use this to justify an arbitrarily large number; it should be grounded in an observed elasticity from a live market.
- 5
Check the ceiling against your most mature market
Use real data from your densest market as the empirical rebuttal. Gurley already knew Uber in San Francisco was 20x the San Francisco taxi market when Damodaran published his cap.
Pro tip One saturated market is worth more than any top-down estimate.
- 6
Add an explicit optionality line
When something is genuinely disruptive and unlocking behaviour, note that the ability to build on top of it is real value your model cannot yet quantify.
Pro tip State it qualitatively rather than inventing a number, so it informs judgement without corrupting the model.
Watch out Optionality is the most abused line in any deck; only invoke it where the platform behaviour is already visible.
In the wild
In 1999 OpenTable's newly recruited CFO told Gurley he was quitting because his model said the business would never work. The model froze penetration in each city at 17 percent, a number he had carried over from retail, where no business he had worked with got more than 17 percent share.
→ Gurley told him they were going to 99 because of network effects. The CFO left; OpenTable kept scaling and was throwing off substantial cash flow per quarter by the time it filed its S-1.
An NYU valuation professor published a piece arguing Uber could not be worth more than two or three billion dollars, having taken the taxi market as the upper limit. Gurley wrote a long rebuttal, calling him beforehand to warn him it was coming.
→ Gurley already knew Uber in San Francisco was 20x the size of the San Francisco taxi market, so the ceiling had been empirically blown through before the analysis was even published.
Common mistakes
Taking the incumbent market as the upper limit
This is the classic error in Gurley's telling — Damodaran on taxis, McKinsey returning 900,000 units for global mobile phones. If the new product is dramatically better, the incumbent market measures the old behaviour, not the new one.
Importing penetration norms across industry structures
A 17 percent share ceiling is real in fragmented retail and meaningless in a winner-take-most network business. The number is not wrong, the transplant is.
Treating conservatism as safety
Gurley says explicitly that TAM conservatism hurts you more than it helps you as an investor, because the conservative model is exactly what makes you pass on the outliers that pay for the fund.
Is it for you?
Best for
Founders, investors and analysts sizing markets for products that are dramatically more convenient or cheaper than the incumbent.
Not ideal for
Mature, supply-constrained categories where demand genuinely is fixed and the incumbent market really is the ceiling.
From the transcript
“so we look at the model and I dive in and he has frozen penetration in each City at 17”
“no one gets more than 17 market share all the businesses I've worked with because I believe the network effects I was like we're going…”
“he basically took the taxi market and said that's the upper limit yep and that's just the wrong math like Travis and the team made…”
“the upper limit and I have found you get into more trouble with this kind of Tam conservatism then like it hurts you more than…”
From the episode
#651: Legendary Investor Bill Gurley on Investing Rules, Finding Outliers, Insights from Jeff Bezos and Howard Marks, Must-Read Books, Creating True Competitive Advantages, Open-Source Strategies, Adapting Mental Models to New Realities, and More