What founders learn by genuinely caring about their customers

Executive overview

Early-stage founders assume they understand the problem, the customer, and the solution. That assumption costs them the insights they most need. The fastest way to learn is to care about customers enough to spend real, unscripted time with them.

Three case studies — Airbnb, Brex, and Twitch — show the same pattern: founders who personally engaged with individual customers discovered things no survey, PM report, or data science team would have surfaced.

Caring isn't a soft virtue — it's the fastest learning mechanism available to an early-stage founder.

The Airbnb photography story

• Airbnb noticed hosts had poor photos and decided to fix it directly — flying in, rearranging lights, reshooting properties
• They knew the hosts would use those photos on competing platforms and that the exercise lost money
• One host offered coffee after the shoot; Joe accepted instead of moving on
• That conversation revealed the host had rented his apartment for 10 years and kept detailed notes in a notebook
• The host handed over that notebook — a decade of hosting knowledge no casual interview would have uncovered
• The learning only happened because Joe stayed, cared, and was present
• Airbnb founders still remember the names and life stories of their first hosts

The Brex pivot

• Brex started with a VR hardware idea they had no real expertise in; it was essentially dead on arrival
• They pivoted to fintech because they genuinely understood other YC founders — the people around them in the same building
• They talked directly to batch-mates: do you have a credit card? No? Want to use ours?
• Their customers were founders under 25, non-Americans, or otherwise ineligible for standard credit — people the incumbents simply ignored
• The product didn't need to be great to start; competing with nothing is a low bar
• Deep familiarity with customer edge cases (multi-country employees, non-US structures) drove product decisions that would be very hard to fake

The Twitch inflection point

• Justin.tv had a complicated relationship with users — rights violations, chat abuse, legal exposure
• The shift came when Emmett and Kevin started personally calling StarCraft 2 streamers, one by one
• Streamers were surprised: a CEO and COO calling about their channel five years into the company
• The first requests were embarrassingly simple — stream in higher resolution
• Higher bitrate was a config file change; it had been deprioritised because it increased costs for rights-infringing content
• For legitimate streamers, there was no reason not to do it; it was shipped within two days
• Streamers then asked: can we make any money? Even $20/month felt significant to them
• Kevin handwrote and mailed the first checks
• Core insight from those calls: if streamers could earn enough to quit their jobs, they'd produce content millions wanted to watch
• That learning — which became Twitch's entire model — only happened because two founders picked up the phone

What kills this kind of learning

• Hiring a PM to write a report, or asking a data science team for a survey, creates distance that kills the signal
• Adding layers between founder and customer feels more professional but produces worse outcomes
• Money doesn't unlock these insights — more money often adds the very layers that prevent them
• Founders who are "flailing" and getting no results despite high effort are often just not doing this

The pattern across all three

• All three stories start with founders who remembered individual customers by name
• None of them hid behind keyboards, landing pages, or mass emails
• The customers felt special; feeling special made them open up
• The insights they shared were things no structured process would have surfaced
• Caring is not fakeable — and it compounds