Foundational Insight Over Customer Research
For consumer and SMB products, customer feedback is directionally wrong — build from your own insight instead.
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 90%
Rabois refuses to let himself or colleagues talk to customers for consumer and SMB products, because customers don't know what they want and are especially bad at explaining subconscious purchase decisions — so the input is often directionally wrong and, worse, sticks in your brain and contaminates every subsequent decision. The best companies rest on a foundational insight you can pressure-test with logic, not survey. The sharp exception: hardcore enterprise, where a finite set of real decision-makers makes utilitarian choices you genuinely can extract.
Origin
Rabois's own contrarian take, illustrated with the Porsche/Lamborghini buyer example and his origin stories for DoorDash and Airbnb; he frames enterprise as the exception via a company with '30 must-win accounts.'
Core principles
- 01Consumer purchase decisions are largely subconscious, so conscious answers are misleading even when honest
- 02Customer feedback is not just unhelpful but actively harmful — it locks into the team's brain and corrupts later meetings
- 03A fake, isolated experiment isn't a proxy for the real world where products must cut through the clutter
- 04The best companies rest on a foundational insight you can acid-test with logic
- 05Enterprise is the exception: a finite, real set of decision-makers making utilitarian choices can be interviewed usefully
How to run it
- 1
Classify the buyer
Determine whether you're selling to consumers/SMB/micro-merchants or to hardcore enterprise. This single split decides whether customer development helps or hurts.
Pro tip Anything sub-mid-market Rabois treats as directionally dangerous; Square and DoorDash's SMB side included.
Watch out Don't assume 'talking to customers' is universally good advice — for consumer/SMB it's an 'unmitigated disaster.'
- 2
For consumer/SMB, refuse the feedback
Don't build product direction from customer interviews. Recognise buyers will give you every reason except the real one, because the decision is subconscious.
Pro tip The tell: ask any supercar owner why they bought it — 99% of the time you get every reason except the real one.
Watch out Even caveated 'I know this isn't statistically representative' anecdotes contaminate the team's brain and every subsequent meeting.
- 3
Build from a foundational insight and pressure-test the logic
Rely on a foundational insight you have about the world, then acid-test its underlying logic rather than validate it with a survey. Look for hard evidence that reveals latent demand.
Pro tip Airbnb's Chesky cited ~30 Bay Area Craigslist posts from people wanting to rent a room — evidence that required people to type it in themselves, signalling real latent demand.
Watch out If you sampled 10 random people on a contrarian idea like Airbnb, 90%+ would say no — random sampling would have killed it.
- 4
Validate feasibility, not desire
Instead of asking consumers if they want it, test whether the business can be made to work — e.g. run supply-side experiments and unit-economics checks.
Pro tip For DoorDash, the move wasn't asking diners if they wanted a delivery button; it was the stat that 93% of US restaurants don't deliver, plus testing placards, deliveries-per-hour, and density to break even.
- 5
For enterprise, do the opposite
When the target is a finite set of must-win enterprise accounts, do talk to customers — meet every decision-maker, influence the CEO, and extract their utilitarian requirements.
Pro tip With ~30 must-win accounts over two years you can literally meet the decision-maker at all 30; that exercise is genuinely useful.
Watch out This only works because the buyer set is finite and real — it does not transfer to a billion-person consumer market.
In the wild
Rabois says no customer told them they wanted a delivery button on their phone. The epiphany was a stat — 93% of US restaurants don't deliver — plus Andrew Mason's 'I'm hungry' button idea. The validation was supply-side: would restaurants place a placard, and could enough deliveries-per-hour and density make the economics work.
→ DoorDash was built from a foundational insight and feasibility testing rather than consumer surveys.
When Chesky pitched Airbnb, the number that stuck was ~30 Bay Area Craigslist listings from people wanting to rent someone's bedroom — demand people had to type in themselves. Rabois was leaning in before the 3-minute pitch ended, even though random sampling of 10 people would likely have produced 90%+ 'no.'
→ Rabois invested; the founder's own foundational insight, backed by hard latent-demand evidence, beat what customer surveys would have said.
Common mistakes
Running consumer customer-development
For consumer/SMB products, interviews yield directionally wrong input because the purchase is subconscious, and the anecdotes then lodge in the team's brain and poison every later decision.
Applying the no-customers rule to enterprise
Refusing to talk to a finite set of must-win enterprise accounts throws away genuinely extractable, utilitarian decision-maker requirements — the one context where customer development works.
Is it for you?
Best for
Consumer, SMB, and micro-merchant founders deciding product direction who are tempted to over-index on customer interviews
Not ideal for
Hardcore enterprise products with a finite set of identifiable decision-makers, where customer development is valuable
From the transcript
“Yeah, I hate talking to customers. I refuse to allow colleagues of mine to talk to customers.”
“99% of the time they will tell you every reason except the real reason.”
“if it's hardcore enterprise, customer development does work”
“93% of restaurants in the United States don't deliver”
From the episode
Hard truths about building in the AI era
Keith Rabois (Khosla Ventures)