Gut-Over-Data Bets (When the Numbers Say No)
Make bets the data argues against when you intrinsically understand a real, unsolved end-user problem.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 90%
A decision framework for growth bets where quantitative data points the wrong way. The permission condition is not a hunch in a vacuum — it's a deep, intrinsic understanding of what the end user wants and the specific problem to be fixed. When you have that, and the data disagrees, Kansal argues you go with your gut. Each such bet carries real career risk, which is why it must rest on genuine user insight rather than optimism.
Origin
Sachin Kansal's account of the highest-conviction, data-defying bets he drove at Uber (safety sentiment, taxis, Uber for Teens), each of which he says 'could have gotten me fired.'
Core principles
- 01Data reveals behavior and friction but can't validate an unsolved need no one has served yet
- 02The license to override data is intrinsic understanding of the user's real problem
- 03These bets carry genuine personal/career risk — treat them seriously, not casually
- 04Unsolved user problems are headroom; a 'no' from current data may just reflect the status quo
How to run it
- 1
Identify a real, unsolved user problem
Start from a problem you understand intrinsically — often one you feel personally or hear directly from users — that current products don't solve.
Pro tip Personal proximity helps: Kansal, father of a 16- and 12-year-old, knew kids' transportation was a genuine unsolved household problem.
- 2
Check what the data says — and why
Look honestly at the quantitative case. Often it will argue against the bet (fewer taxis, unreliable service, antiquated software; parents refusing; risk teams citing liability).
Pro tip Understand that data describing the status quo can't see a market you'd create with the right structure or incentives.
Watch out Don't confuse 'the data says no' with 'the opportunity isn't real' — the data may only reflect today's conditions.
- 3
Design the structure that changes the equation
Rather than accept the data's verdict, engineer the conditions that make it work — the right incentives, features, or product that resolves the objection.
Pro tip For taxis: drivers wanting demand + Uber having excess demand + the right incentives = a productive structure the raw data missed.
- 4
Commit despite the risk
Go with your gut and ship, accepting that the bet could fail visibly. Anchor the conviction in the user insight, not in optimism about recovery.
Watch out Kansal notes each of these bets could have gotten him fired — only make them when the user insight is genuine, not to chase a metric.
In the wild
Data said taxis were declining, unreliable, and antiquated, and Uber and taxis had a hostile history. Kansal pushed to launch with taxis anyway, betting that drivers wanting demand plus Uber's excess demand plus the right incentives could work.
→ Uber now operates taxis in ~10-15 US cities; in New York City every yellow cab is hailable through the Uber app, generating substantial revenue.
Every data signal said no — parents said they'd never put a teen in an Uber, risk teams cited liability. Kansal, feeling the kids-transportation problem at home, argued Uber should be the one to solve it and built a product targeting every parental safety concern.
→ Launched two years ago and growing well; when California forced it off, Kansal received daily parent emails asking why — evidence of real demand the data hadn't shown.
Common mistakes
Treating gut as a substitute for user understanding
Kansal's condition is 'as long as we understand intrinsically what the end user wants' — a gut call without deep user insight is just a guess, and these bets carry enough risk that they could get you fired.
Letting status-quo data veto a creatable market
Reading 'taxis are declining and unreliable' as a reason not to act misses that the right incentives and structure can create demand the current data can't measure.
Is it for you?
Best for
Senior product leaders with real user insight and the standing to make high-conviction, data-defying growth bets
Not ideal for
Reversible optimization decisions or anyone lacking deep, firsthand understanding of the user problem — there, follow the data
From the transcript
“I have a I have a few and each of those could have gotten me fired, Lenny.”
“the data will tell you that the number of taxis is going down, that the taxis are not very reliable”
“a lot of the growth pets that we actually make may not have the best data, you know, supporting them, but as long as we…”
“you go talk to parents, they're like, I am never putting my teen in an Uber”
From the episode
Why Uber’s CPO delivers food on weekends
Sachin Kansal