Pivots
What counts as a pivot, what triggers a Pivot review, and how a Startup changes course without losing a record.
A pivot changes a Startup's founding Hypothesis at the root: its customer, its Problem, or its approach. It is a Decision and a new Version, never a fresh start.
What counts as a pivot
The founding Hypothesis has five slots. Changing the customer, the Problem or the approach is a pivot. Adding an Offer, an Interface, a human step or a second side of a marketplace is growth on the same root.
Take a Startup that sells an API of restaurant health-inspection scores to insurers.
| Change | A pivot? |
|---|---|
| Sell to restaurant groups instead of insurers | Yes: a new customer |
| Solve kitchen staffing gaps instead of inspection risk | Yes: a new Problem |
| An Agent that files corrective plans instead of an API | Yes: a new approach |
| Add an MCP server beside the API | No: a new Interface |
| Add a $49 self-serve plan | No: a new Offer |
The line matters because a pivot invalidates Evidence. Interviews with insurers say nothing about restaurant groups, so the customer slot drops back to the grade the new customer has earned.
When to review
Pivot review runs whenever the Founder asks, and it runs on its own on four signals:
- The first outside paid Order has not arrived by the Startup's first-revenue date, 30 days after launch by default.
- Weeks flat reaches the Startup's line: eight weeks of outside revenue growth below target.
- The riskiest slot is still graded opinion after four weekly tests.
- Office hours name the founding Hypothesis as the week's biggest risk.
The review compares the Evidence with what each Goal promised and ends in one Decision: persevere, iterate, or pivot. Iterating changes an Offer, a Price or a Play on the same root; pivoting changes the root.
How a pivot is recorded
A pivot is a Change to the founding Hypothesis, merged as a new Version. The Decision behind it records the options, the Evidence, the choice, and what would reverse it.
Site copy, Offers, Prices and Plays are compiled from the Hypothesis, and each records the slot and Version it came from. After a pivot, everything compiled from a changed slot is rebuilt, and everything else stays as it was.
Nothing is lost. Interviews, Evidence, Customers and Orders stay on the Startup, and the old Version stays one revert away.
What happens to the state
A pivot does not reset the Lifecycle. The Startup keeps its state, and every entry check that reads a changed slot runs again against the new Version.
The inspection Startup pivots to restaurant groups while Live. Its Site, API and payouts still pass, so it stays Live.
The rebuilt pricing Page and Offer reach Customers at the next deploy. A check that fails opens a Task for the Role that owns it.
Pivot or start a sibling
A Startup with paying Customers has a second option: keep running and test the new direction beside it. clone copies the definition into a sibling Startup with the changed slot, and both run until the Evidence decides.
The Studio works this way by default. One Idea runs as several sibling Startups that differ in one slot. The winner's choices flow back into the Startup Template every new Startup inherits.
Next
Research and define covers grading the Evidence a pivot rests on, and the Lifecycle overview shows where each state's number comes from.