How Stripe's CEO Patrick Collison thinks about running a startup

Executive overview

Most startup CEOs treat their growth curve as something to nudge. Collison argues it is something you build — the market is finite, the percentage you're serving is under your control, and the organisation you build determines how fast you reach it.

Pre-product market fit, speed of iteration is the only thing that matters. Post-PMF, the job switches to building the organisation ahead of where you are today.

The growth curve is not a cosmic trajectory — it is something you construct.

Getting to product market fit

• PMF arrived at public launch: demand immediately exceeded supply, which inverted the problem from "generate users" to "serve users"
• At launch Stripe had ~10 people and had been iterating for nearly two years
• Early signal: every API request triggered an email; unusual behaviour was investigated immediately
• Proactively reached out to users who had sent test requests but never gone live
• Embedded feedback prompts on every page framed as negatives ("the worst thing about this page is…") to surface real problems
• The Collison installation — going in person to help users integrate — served two purposes: UX research and creating a "why now" moment that removed the friction of postponement

Iteration speed is the pre-PMF metric that matters

• Pre-PMF, qualitative feedback beats quantitative metrics — you have too few users for numbers to be meaningful
• The right question: will this additional action shorten the cycle from "observed user problem" to "shipped fix"?
• Team size 2–10 is typically the optimally responsive range; beyond that, coordination costs can outweigh the throughput gain
• Each hire adds: time to hire, time to onboard, ongoing coordination overhead — worth it only if it shortens the feedback loop
• Reference frame: Boyd's OODA loop — optimise for responsiveness to a rapidly evolving situation, not raw capability

The mental model shift after PMF

• Wrong model: you are on a growth curve and your job is to marginally inflect it upward
• Right model: there is a finite market; you are currently serving a fraction of it; the rest is addressable by building the right go-to-market organisation
• Map the concentric circles of your market immediately after PMF: early-stage startups → all technical startups → all businesses handling online payments, etc.
• Estimate size of each ring, current penetration, and what it would take to serve the next ring
• Repeat founders do this instinctively — they build the organisation ahead of where they are today because they know the growth curve is under their control
• B2B is more mappable than consumer: businesses know what they want, making market sizing rational rather than speculative

Decision-making and co-founder dynamics

• Stripe is unusual: Patrick and John genuinely co-run the company; title (CEO vs President) is less determinative than domain bias
• Dispute resolution is based on who cares more, not who holds the title — passion correlates with execution quality
• The key structural requirement: have an efficient mechanism for reaching a decision — consensus loops and quasi-democratic voting are both failure modes
• Dispassion in disagreement — treating divergence as non-threatening — is a hallmark of successful co-founding relationships

Scaling decision-making and communication

• Between 5–50 people, Stripe was too consensus-oriented — a common mistake as teams grow from small flat structures
• Around 10 people, switch deliberately to an explicit communication model: decisions are named, communicated, not just understood by osmosis
• Being non-consensus-based is not the same as being autocratic — the goal is to preserve autonomy and ownership while breaking ties efficiently
• Strong ownership mentality requires people to feel they can identify problems and inject ideas outside their direct area
• The failure mode to avoid: Brownian motion — everyone moving independently, no coordinated direction

Hiring and what to look for

• Three qualities: intellectually honest, cares deeply, loves getting things done
• Hard to fake being smart; also hard to fake intellectual honesty — genuine ability to hold multiple sides of a question signals it
• Warmth and pleasantness matter and are also hard to sustain as performance indefinitely
• First non-engineer hire (Billy Alvarado, partnerships) was transformative — previously banks were unreachable; immediately unlocked trajectory
• Lesson: founders who are all engineers often can't conceive of what a non-engineer does — push past that

Preserving a culture of new ideas at scale

• Default outcome as companies grow: ossification, orthodoxy, resistance to contradiction
• Counter-measure: yes-and culture — enjoy contemplating ideas even when you know you won't pursue most of them
• Annual "crazy ideas" process: a shared document open to the whole company for ideas that are probably bad but could be great if they worked
• Ideas must be probably bad — if they're probably good, just do them; the document is for things people would self-censor out of fear of looking foolish
• Stripe Atlas and support for the Stellar cryptocurrency both originated from this process
• Benchmark examples of large organisations that successfully repeated this loop: Amazon and Google

On happiness, fulfilment, and knowing when to quit

• Happiness during the early years was not the operative state — stress and deficiency awareness were constant
• More useful utility function: fulfilment — the sense, looking back, that the work was meaningful
• Working on something truly miserable is a signal worth heeding; determination at all costs is the wrong heuristic
• Knowing when to quit is underrated in startup culture — time has high opportunity cost