Stewart Butterfield on building products people love

Executive overview

Most product teams optimise for the wrong things: reducing friction, cutting clicks, shipping features. The real challenge is almost always comprehension — helping users understand what to do and why it matters.

Stewart Butterfield (Flickr, Slack) shares the mental models that shaped both companies: utility curves, the owner's delusion, hyper-realistic work-like activities, and why "don't make me think" beats every other design heuristic.

The measure of long-run success is the amount of value created for customers — there is no substitute for actually creating it.

Utility curves

• Value doesn't scale linearly with investment — it follows an S-curve: flat, then steep, then flat again.
• The decision isn't binary (have feature / don't have feature); it's whether you're on the first shallow section, the steep section, or past diminishing returns.
• Features added before crossing the threshold generate complexity without user benefit — and usually get quietly abandoned.
• Standards continuously rise (Bezos's "divine discontent") — things that were good enough two years ago may now sit below user expectations.
• Core flows (sign-up, forgot password, checkout) rarely get revisited despite the quality bar rising industry-wide.

Friction vs. comprehension

• "Remove friction" became a mantra applied indiscriminately — it's only the right lens when users have high intent and clear understanding of what they want.
• For most product surfaces, the challenge is comprehension: what is this, and what am I supposed to do next?
• Low-intent users (e.g. someone who vaguely heard about Slack and finally visited the site) are barely above the threshold to act — any confusion tips them out.
• Making users think has two costs: metabolic (actual cognitive load) and emotional (people feel stupid, not the software).
• Reducing clicks is almost always the wrong goal — eight trivially easy taps beat two confusing ones.
• The Uber "where would you like to go / other" design is the ideal: surface the one dominant action, hide everything else.

Craft and delight

• "Tilting your umbrella" — most people don't move their umbrella for oncoming pedestrians. Their failure to be considerate is your opportunity as a product builder.
• Slack's magic-link login came from asking: why do we make people type a password when email ownership already authenticates them?
• The shouty rooster ("don't be a cock"): when a user typed @channel, a rooster appeared showing how many people across how many time zones would be notified. Abuse dropped immediately — and it taught users how the product actually worked.
• Do-not-disturb rollout: admins were notified weeks in advance, defaults were set automatically (most users would never turn it on themselves), and a layered override system let organisations and individuals adjust without conflict.
• Smart defaults matter — if you don't set a default, most users will never discover the feature.

The owner's delusion

• Restaurant websites that load slowly, play music, and bury the phone number in an unclickable image — the owner visits competitors' websites all the time, knows the frustration, yet produces the same thing.
• Owner's delusion: assuming visitors share your level of familiarity, intent, and patience with your product.
• Users arrive distracted, barely above the action threshold, ready to bounce in milliseconds — they are not a seated audience waiting for a curtain to rise.
• Fix: name the pattern, discuss it explicitly, then take a breath and look at the product as a first-time user with no context.

Hyper-realistic work-like activities

• At the start of any company, there is abundant known valuable work — things you know how to do and know will matter.
• As headcount grows faster than clear priorities, the supply of obvious valuable work shrinks relative to the demand for it.
• People fill the gap with work that looks identical to real work from the outside: previewing a deck to get feedback before the meeting where you present the deck.
• This isn't stupidity or malice — people are under career pressure to demonstrate output, and the activity genuinely resembles work.
• The Slack threads example: thousands of person-hours spent A/B testing whether pre-populating a reply box with @previousposter increased average thread length by 0.03 messages. The analysis cost more than any conceivable gain.
• Responsibility sits with the leader: create clarity around priorities, say no to things upfront, and never just chide the team for wasting time.

Parkinson's Law and org bloat

• Work expands to fill the time available for its completion — and org charts expand to fill headcount budgets.
• Nearly every person hired wants to hire someone below them, not from malice but because direct reports correlate with pay, authority, and career progression.
• Product managers at Slack would immediately want to hire a junior PM: "they'd do the product management, I'd do the strategy."
• Navy ships declined; navy administrators kept growing — the same pattern appears in universities, governments, and tech companies.
• Complexity compounds: if a problem seems simple, you probably don't understand it yet.

Don't make me think

• Don't make me think (Steve Krug's maxim) was as central a principle at Slack as utility curves.
• If software stops a user and asks them to make a decision they don't understand, they feel stupid — they blame themselves, not the software.
• People who aren't technically confident almost always internalise confusion as personal failure.
• The worst outcome: a user associates a feeling of inadequacy with your product and never returns.
• Apple's "Sleep" feature in the Clock app is a standing example — years after launch, most users still don't know what it does or what turning it on costs them.

The "we don't sell saddles here" memo

• Written before Slack launched publicly, when the team was ~8 people, to align thinking early enough that it could survive scaling.
• The core argument: you are not just building a product, you are creating a market.
• For anything that differs meaningfully from existing alternatives, the job isn't just to make it good — it's to help people understand why they'd want it at all.
• Positioning (the book) makes the same point: it's nearly impossible to create a new idea in someone's head from scratch; anchor to two existing ideas and combine them.
• Harley-Davidson sells the open road, not engine specs. Lululemon sells aspiration, not yoga pants.

On pivoting

• The default advice ("persevere") is almost always what people want to hear — and it's often wrong.
• The real question: have you exhausted all the realistic possibilities, or just the easy ones?
• Glitch (the game that became Slack) still had $9M in the bank and a team that liked the work — but Butterfield had run out of non-ridiculous ideas to make it commercially viable.
• The barrier to pivoting is humiliation: you convinced investors, employees, and users to believe in something, and admitting it isn't working is painful.
• The answer is to create intellectual distance — make the decision rationally, not emotionally.
• Contrast with Melanie Perkins at Canva: 100 investor rejections, but she was right about the vision and just needed the right articulation. Not all persistence is irrational.

Long-run value creation

• Slack's SLA offered 100× money back for downtime — fine when revenue was small, costly when a multi-hour outage hit at public-company scale.
• Proactive, automatic credit (no claim required) was the operating principle: create value, demonstrate it, don't make customers fight for it.
• Fair billing (no charge for inactive seats) and COVID credits followed the same logic: generosity that builds genuine advocacy.
• The prisoner's dilemma framing: acts of generosity signal cooperation; cooperation compounds over iterated games.
• Being ethical attracts better employees, creates a better internal culture, and reduces internal friction — the benefit isn't just moral, it's structural.