Pivot

When a signal arrives that the hypothesis was wrong, a team has three options: push on, adjust incrementally, or change direction. The trouble with the third is that without a name for it, the decision sounds like declaring failure.

Pagedeck's first product ran into this. What the team originally built was a personal notes app. Six months produced 8,000 signups and a one-month retention of 4 percent. Looking at that 4 percent, though, they had something in common: they weren't people working alone, they were people sending note links to colleagues and editing together. When Grace proposed changing direction, the first response in the room was whether that meant throwing away six months.

What got thrown away was one hypothesis, not six months. The editor, the storage design, and the fact those 4 percent had revealed all stayed. The decision has its own name because this conversation only becomes possible once what's kept and what's changed are stated separately.

The word comes from keeping one foot planted and turning on it — which is what separates this from starting over. Five things are usually available to turn.

1. Change the customer. Keep what was built and go find the people for whom it's urgent. Pagedeck moving from individuals to teams is this one, and it's the turn that changes the least code and the most of everything else.
2. Change the problem. Move to a different difficulty the same customer has. Deciding that the problem was editing together rather than organising notes overlaps here, since the two turns frequently arrive in the same conversation.
3. Change the weight of a feature. Promote something that was a side feature into the whole product, or fold what was a product into a single feature of something larger.
4. Change how money is collected. Individual subscriptions become team contracts, with almost nothing built differently and the average revenue per account moving by an order of magnitude.
5. Change the route to the customer. Waiting to be found in search becomes sales meeting people directly, which changes the price the product has to carry.

What the decision requires is learning rather than opinion. Learning confirmed against real data is validated learning, and that's where the grounds come from. At Pagedeck there were three: one-month retention at 4 percent, the record that 80 percent of those who stayed had shared a link, and the gap that retention among sharers whose colleague edited at least once was 31 percent.

Once the judgement is made, the next step is running the build-measure-learn loop again on the new hypothesis. Turning doesn't produce an outcome by itself. The new hypothesis can be wrong too, and companies on their second and third turn are not unusual — what distinguishes them is whether each turn was cheaper than the last.

Setting the trigger in advance is what makes the decision possible at all. Write down that if one-month retention hasn't passed 20 percent in three months the direction gets revisited, and at that point the conversation opens on schedule. Without a threshold, teams drift toward adjusting slightly every month, and a year of that leaves no money with which to turn.

Three situations bring the word out.

1. The review of an early product. Having missed the threshold that was set, the team decides what to keep and what to change. The most common failure here is deferring the decision, usually on the expectation that a little more data will make things clear. It rarely does, because the data has already said the same thing for two months.
2. Funding conversations. Raising the next round means explaining how you read the failure. A history of turning reads either as evidence of learning or as evidence of not holding a direction, so say what was learned and what was kept in the same breath.
3. Editing the product strategy. That document is the record of what you will and won't do, so a change of direction starts with rewriting the "won't" list rather than the vision at the top.

Turning with existing customers in hand means settling what was promised to them. That none of the 8,000 individual users had paid anything made Pagedeck's move straightforward; with 200 paying companies, the same decision would have arrived attached to a migration plan and a refund policy.

"That's not a pivot, that's retiring one feature."

Keep the term for decisions that actually earn it.

"We wrote down that if retention is under 20 percent in three months, we revisit the customer."

The threshold and the date were set in advance.

"Start with why the remaining 4 percent stayed."

Look for the grounds in the users who didn't leave.

The meeting where Grace and Mia made the call ran two hours, and ninety minutes of it went to reading the usage records of the 40 remaining users one at a time. What changed afterwards was the first screen, the pricing, and the signup flow; what stayed was the editor, the storage design and search. The new hypothesis was that the people who stay are the ones with a colleague to edit alongside, and three months later one-month retention was 26 percent. Of the original 8,000, seven thousand never came back, and none of them had ever paid anything.

The most common reading is that changing direction is an admission of failure. It means the original plan was wrong, so it sounds that way. But what was wrong is a hypothesis, and hypotheses exist to be tested. On teams that call this failure, the turn arrives late, and lateness costs exactly the money and time that were left. What deserves to be called failure is pushing on for another year after the signal arrived, with the same numbers on the same dashboard.

The other reading is that it means starting over. "Change direction" sounds like "from scratch." In practice, keeping one leg planted is the defining condition. If there's nothing to keep, what you have is a new company, and then the first question is why it would use this team and this funding. What Pagedeck kept was the editor and the storage design, and having both is why the new hypothesis could be tested inside three months.