Don't Make Me Think: Stewart Butterfield on Utility Curves, Taste and the Case Against Removing Friction

Butterfield says he has been mostly relaxing since leaving Salesforce two and a half years ago, spending time with his young daughter and on philanthropic and art projects. He and Slack CTO Cal Henderson, who have worked together for 23 years, keep discussing what to build next. His worry is that phones are "destroying the world," and what they are good at is making software, so any idea would need to help people use their phones less.

The first concept Lenny asks about is the utility curve. Butterfield describes an S-curve with effort or cost on the horizontal axis and value or convenience on the vertical one. The first bit of effort yields little, then a threshold produces a big jump, then continued investment stops paying off. His example is a hammer whose handle breaks on impact: it is junk until it is strong enough, and after that more strength doesn't matter.

He says this helped teams stop treating features as binary. A half-built feature adds complexity without value, and people then conclude the idea was bad. So the team would ask whether they were on the first flat part of the curve, the final flat part, or just approaching the steep section. Lenny summarizes it as a lens for finding the "aha" moment.

Butterfield adds that the curve itself moves, a Bezos-style "divine discontent" effect: as users get familiar with products, their standards rise. His example is the iOS Google Calendar time zone picker, which lists every time zone in the world alphabetically, even when he searches "east". Nobody will switch email providers over a time zone picker, he concedes, but small frustrations build emotional connections, or resentment, that shape which tools people advocate for later.

Asked about taste, Butterfield points out the word comes from experiencing food, and chefs clearly improve with training. Some people have a natural edge, but he thinks taste can be practiced. He also warns that obsessing over one small thing beyond the knee of its utility curve is a failure mode of its own.

The second point is that leaning into taste is a real advantage because most people don't invest in it, echoing Bezos's line that your margin is my opportunity. He tells a story of walking in the rain in Vancouver with a colleague, Brandon Velestuk, and noticing how few people tilted their umbrellas out of the way on narrow sidewalks. They guessed that about a third did.

The explanations they came up with were that some people enjoy exercising a bit of power, some see the problem and can't think of a fix, and most simply don't notice their effect on others. Butterfield argues that this lack of empathy is exactly the gap a company can exploit. "Tilt your umbrella" became a Slack saying, even appearing on company swag.

He says Slack had tailwinds, but its growth also relied on little conveniences that made people attached to it. Much of the spread came from someone moving from one startup to another and insisting their new team try it. That kind of advocacy, he says, comes from emotional connection.

Butterfield gives several concrete examples of early craft at Slack. One teammate asked why users should enter an email address and a password when owning the email address was already the proof needed. Typing a strong password on a phone is a poor experience, so the mobile app asked only for an email and sent a link that opened the app and signed the user in. He credits others with the idea but says Slack scaled it.

On notifications, people expect every message to ping them, as in older messaging apps, but that makes less sense in Slack. Still, new users who invited a colleague and got no ping would complain. So the team reluctantly notified for every message by default, then after about ten messages offered a switch to recommended settings where only DMs and mentions notify.

To stop the abuse of @everyone, which he compares to the overused priority flag in Outlook, Slack added a confirmation with a rooster. It warned that the message would notify, for example, 147 people in eight time zones. Internally it was called the shouty rooster. Lenny notes the feature survives, and Butterfield says it was easy to build, made a big difference and also taught people how the product works.

For Do Not Disturb in 2017, Slack faced huge customer bases, including on-call alerts for major systems. The team told every administrator weeks ahead that an org-wide default was coming, that admins could override it, that users could override that, and that a later admin change would override again. Slack set defaults because most people would never turn the feature on themselves. He says this convoluted rollout avoided conflict.

Butterfield says removing friction became an unquestioned mantra, but it only helps in some cases. Slack was hard to market because nobody needs to be told why they want beer or a car, while everyone has to be told why they want Slack. His example is buying Taylor Swift tickets: because your intent and its specificity are maxed out, you'll fight through a slow site. Registration, authentication and checkout are cases where friction reduction does matter.

Most visitors, however, don't have 100% intent. Someone arriving at Slack.com after hearing about it for months is just barely over the threshold, and has only a vague idea of what they need it for. The challenge is making clear what the thing is and what to do next, through explanation, visual hierarchy and obvious next steps. Butterfield thinks 70 or 80 percent of product design lives in this comprehension step.

He says people rarely discover capabilities in preference screens and cites Shopify reports and Apple's Clock app. The Clock app's alarm tab says sleep is off and that you need to turn sleep on to edit your schedule, which he finds incomprehensible. He expects the great majority of users just set an alarm and never learn what the sleep feature does.

Butterfield says "Don't Make Me Think" sat alongside utility curves as a Slack mantra, for two reasons. Decisions are metabolically costly and lead to fatigue. And a confusing prompt makes people feel stupid; he notes that people talking with their parents about software see the parents blaming themselves. A bad experience can then become permanently associated with the product.

He also argues that reducing clicks or taps is almost always the wrong goal. The easiest way to make every action one tap is to expose every option on one endless screen. He praises the old Uber app, which asked where you wanted to go and put everything else under "other," and says eight trivially easy taps beat two taps full of difficult choices. Comparing many options against each other gets geometrically more expensive.

To illustrate, he tells a story about Patrick Collison raising a Gmail annoyance with Google's CEO at a conference: moving recipients between To and CC takes too much dexterity. He also complains that Gmail splits actions across two long menus and an unlabeled icon. Then he describes watching a teenager on a plane tap Snapchat four times a second for six minutes, a fluid experience that would be harmed if the goal were fewer taps.

Lenny notes that successful product leaders tend to be constantly dissatisfied with how things work. Butterfield's version of a cynical exercise is to tell people to close their eyes, breathe, pretend to be a real human being, and then check whether the screen makes sense.

Butterfield returns to the quote he gave MIT Technology Review in 2014, the year Slack launched. He said Slack was a giant piece of junk that the team should be humiliated to offer the public. He says that not everyone finds that motivating, and the next day staff had printed the quote on about forty sheets of paper and pasted it on the wall.

His point is that you should always see almost limitless room to improve, even while taking pride in individual pieces of work. If you can't, he says, you shouldn't be designing the product or running the company. He connects this to Kaizen and Toyota's continuous improvement.

Butterfield cites Ray Dalio's story about Michael Jordan learning to ski, wanting the instructor to point out every mistake because each was a gem to collect. Getting a team there requires a lot of trust. But when it works, he says, direct criticism, from peers or from users, becomes motivating, and people are grateful for it.

Lenny prompts a rant about how much effort it takes to make anything work. Butterfield recalls a thread about how long it took to get a stop sign installed on a residential street, involving engineering, traffic planners and an HOA. He contrasts that reality with politicians who suggest that things would be good if some villain weren't blocking them. He recommends the book Why Nothing Works, which describes a growing number of vetoes on any course of action.

He thinks people struggle to understand complex systems except by assigning blame to someone, and says the real world is too multivariate to make a satisfying political message. The same dynamic shows up inside organizations. Many people assume there is a single decision-maker to blame, when usually it's complexity.

On Parkinson's Law, Butterfield says work expands to fill the time available. He draws on Parkinson's chart of Royal Navy ships, sailors and administrators, where administrators keep growing. Most people you hire want to hire others who report to them, he says, because headcount correlates with career trajectory and pay, and it is not because they are evil or stupid. His own variation is that everything is simple if you don't know what you're talking about.

Butterfield gives an example from Slack's threads feature. He had removed a pre-filled @mention of the previous poster in the reply box because it taught people to use threads wrongly. Six months later it came back, backed by an analysis showing threads averaging 2.17 messages instead of 2.14. He questions whether longer threads are even better, and suspects the difference was within the margin of uncertainty.

His larger point is the cost: feature flags, A/B tests, instrumentation, database tables, queries, dashboards, slide decks and meetings add up to thousands of person-hours. The most that feature could change was tiny, so the analysis was a guaranteed loser. The story leads into the concept Lenny's guests suggested he explain: hyper-realistic work-like activities.

He pairs it with the idea of known valuable work, work whose nature and value are both clear. At a startup's start everything is like this, such as opening a bank account. As more people are hired, supply of obvious work falls behind demand for work, and people start to call meetings to preview decks or pursue questions like building a separate FedRAMP High version of Slack. These activities look like work, and executives and board members are just as prone to them.

The fix, he says, is not to scold people. It is the leader's responsibility to maintain a sufficient supply of known valuable work, by creating clarity about priorities and explicitly saying no to things up front. Whose fault is it when people waste time? In his view it is the manager's, the VP's or the CEO's. He admits this is harder than it sounds.

Butterfield says the famous "We Don't Sell Saddles Here" piece was really an internal memo, and he posted it verbatim from what he told the company when it was about eight people. It was written during private beta before Slack launched, to start instilling ideas early so they would survive as the company scaled. He notes that an irritating thing about Slack is that after you leave a company, you lose access to your history.

Lenny summarizes the gist as: it isn't enough to build a great product, you have to communicate the outcome it achieves. Butterfield adds that unlike beer or cars, a new AI recruiting tool or calendar app is responsible not only for creating the product but, to a degree, for creating the market. He recommends the book Positioning, whose idea is that it is hard to put a new idea in someone's head, but easy to combine existing ones.

Lenny describes the saddle idea as selling the experience of horseback riding so people decide they need a saddle. Butterfield says marketers have long done this: Harley-Davidson sells the open road, and Lululemon sells health and aspiration. Lenny adds the line about instilling a yearning for the sea instead of telling people to build a ship.

Asked for pivot advice, Butterfield says the question is whether you have exhausted the possibilities. With Glitch, the game that became Slack through its internal IRC-based communication tool, the team had spent several years doing only the minimum guaranteed-to-succeed features of the chat tool. They still had $9 million left and liked the game, but he had run out of non-ridiculous long shots for making it commercially successful.

The default advice is to persevere, like a kitten on a branch with a "Hang in there" poster. He argues you must be coldly rational, citing Annie Duke's *Thinking in Bets*, which uses Glitch and Slack as an example of a smart fold. It is rational to abandon an idea when its expected value has shrunk.

The reason he stresses cold rationality is that stopping is humiliating after convincing investors, employees, the press and users. Many founders prefer to continue until the company dies for lack of capital. When Glitch shut down, a community vanished and people lost jobs, some after moving cities. He also notes a three-person team switching apps six months in doesn't count as a real pivot, and that survivorship bias applies.

Lenny reads examples from people who worked with Butterfield: handing an employee $500 before Christmas, crying while laying people off at Glitch and then helping them find jobs, paying 100% of health insurance, and an employee-friendly direct listing and acquisition. For customers, there were free credits during COVID and fair billing that stopped charging for unused seats. Lenny asks how much was strategy and how much simply who Butterfield is.

Butterfield says a lot is just how he is, thanks to his parents, but there is also a game-theory angle: generosity signals that you will cooperate in a repeated prisoner's dilemma, which invites others to cooperate. He made employees chant at all-hands that in the long run success is measured by the value created for customers. Anything shady, he says, shouldn't be done, and there is no substitute for actually creating value.

He also describes a policy that backfired: Slack's SLA gave customers 100 times their money back for downtime, automatically. After the company went public, a many-hour outage meant roughly $8 million in credits, which cost no cash but removed expected revenue from the next quarter. They changed the terms afterward, though he says the other generous decisions helped the company succeed.

Butterfield closes with a concept called the owner's delusion, named by a reply to one of his tweets. Restaurant websites, he says, were once full of slow-loading photos, music and unclickable phone numbers, while visitors only wanted the address, hours, menu, phone number or how to reserve. The people who built the site had surely visited other restaurants' sites for exactly those things.

He ties this to everything earlier: visitors arrive barely over the threshold of intent, and owners forget that they are busy, late or distracted humans. They are not an audience waiting for a curtain to rise, and will bounce in a fraction of a second. Butterfield says Apple's Sleep feature is another example of an owner who understands their own labels too well.

Asked for the solution, he says first you recognize and name it, discuss it, and train yourself to take a breath, pretend to be a regular person and look again. Without naming it, he says, you are stuck. He compares it to Murphy's Law, which doesn't prevent mistakes but helps you watch for them.