← Back to Blog

When to Pivot a Startup: The Signal Was Already in the Record

The Compass Team

November 12, 2026

The question of when to pivot a startup usually gets asked about six months too late.

Peter Reinhardt and his cofounders at Segment sat at the back of a university classroom and counted laptop screens. At the start of class about 60% of the students were on Facebook. By the end, 80% were.

That is a clean measurement of a product losing. It is also the kind of measurement most teams never take, which is why the pivot conversation arrives late and lands badly.

Dalton Caldwell, who runs admissions at Y Combinator, says in his Startup School lecture that most founders take too long to pivot. His explanation is loss aversion. Once you have invested in something, letting go feels like accepting a loss you could still avoid.

Time makes it worse. Lenny Rachitsky assembled 40 startup pivot stories and found a rough clock. Ideation pivots, where the idea changes before a real product exists, tend to happen within 3 months of launch. Hard pivots, where a live product with users changes direction, tend to happen within 2 years, and mostly around the 1-year mark.

His advice is a hard conversation with your cofounder at 3 months, then a yearly check-in.

Founders usually know the clock. Reading it is the hard part.

Eric Ries built the loop around a comparison: build, measure, learn. Measure against what? The hypothesis you wrote before you built. When that hypothesis lives only in your head, and your head is nine months into a fight, the measurement has nothing to sit against. "Lukewarm" becomes a feeling that moves with your mood. The pivot turns into an argument with yourself at 2am instead of a delta you can look at.

There is a boring version of this that works. At the moment you pick a direction, write 3 lines: the assumption, the number you expect to see, and the date you will check.

Signups at 200 a week by June 1.

Reply rate above 12% on the first 50 cold emails.

Two of five design partners actually using it by week 6.

When the date arrives, read the diff. The number either matched the one you agreed to or it didn't, and you settled that months earlier when you were still calm.

The painful pivots are the ones where you were right about the problem and wrong about the wedge. Segment built a lecture tool for classrooms where, by their own count, 80% of students were on Facebook by the end of class.

A number with no target next to it is a story you can retell either way. That is the gap: the measurement existed, and the expectation it should have been compared against did not.

CB Insights looked at more than 400 public startup post-mortems and concluded that failures rarely come down to one cause. They come in patterns. A pattern only shows up against a history, and a history only exists if someone wrote it down at the time.

A pivot is the most expensive decision most founders make. It is also the one they make with the least instrumentation. Three lines at the start of each direction is enough to change that. Then you can kill your own idea on schedule, instead of on your worst night.

Share this article