The pivot that was really a retreat

A team pivoted three times in eighteen months. Each was justified. Together they were a company that never found out whether anything worked.

This is an anonymised composite, not a report about a named company. The situation and numbers are typical rather than reported.

Where they were

Two technical founders, $600k pre-seed, building developer tooling. No revenue, eighteen months of runway at the start.

The problem

Their first product — a CI optimisation tool — had 200 signups in four months and eleven weekly active users. Engagement was flat.

The decision

Whether flat engagement meant the idea was wrong, or the execution was.

What was on the table

• Talk to the eleven active users
  For: Eleven people using something with no prompting is a real signal worth understanding.
  Against: Eleven is a small number to build a company on, and talking is slower than building.
• Fix onboarding and re-measure
  For: Signup-to-active is a known common failure, and cheap to test.
  Against: Might just be polishing something nobody wants.
• Pivot to an adjacent problem
  For: The market was clearly not pulling. Move to one that does.
  Against: You are discarding four months of learning on the basis of one flat metric.

What they chose

They pivoted, to a test-flakiness dashboard. Four months later it had similar numbers, so they pivoted again, to an incident tool. Each decision was individually reasonable and each was made at roughly the same point: about four months in, with weak engagement and no conversations with the users they did have.

What happened

At month eighteen they had three abandoned products, four months of runway, and no more understanding of their market than at the start. They raised a bridge on the strength of the team, then spent six weeks interviewing the users of all three products.

The flakiness dashboard turned out to have eight teams using it daily and quietly hating one specific part of it. That became the company, and it worked. But they had discarded it a year earlier without asking a single one of those eight teams anything.

They did not fail. It cost them a year, and the thing they eventually built had been in their hands the whole time.

What transfers

• A pivot decided by a metric with no conversation behind it is a guess replacing a guess.
• Eleven unprompted active users is not nothing. Small numbers of genuinely engaged people are the usual shape of an early real signal.
• Decide in advance what evidence would make you change direction. Deciding in the moment means deciding while discouraged.
• Serial pivoting can look like responsiveness and be avoidance — each restart resets the clock on ever learning anything conclusive.

More case studies