Definition (canonical)
A pivot is a structured change of one core hypothesis of the business — the customer, the problem, the product or the model — while keeping what has been validly learned. It is neither a tweak (that's iteration) nor a restart (that's a new company): one foot stays planted, the other turns.
The classic pivot types
The canon (Ries) names the recurring moves: customer-segment pivot — same product, different buyer (the SMB tool enterprises kept sneaking into); problem pivot — same customer, different pain (discovered while solving the wrong one); zoom-in — one feature becomes the whole product (the group-photo app whose filters were the business); zoom-out — the product becomes a feature of something bigger; business-model pivot — licensing→SaaS, product→marketplace; channel pivot — direct sales→product-led. Labels matter less than the discipline they encode: change ONE major variable at a time, or you can't attribute what worked.
Signals it's time — data, not despair
Legitimate triggers: months of flat cohort retention despite shipping; customers consistently misusing the product for something else (follow them — that's demand talking); a segment you didn't target pulling harder than the one you did; unit economics that no realistic improvement fixes. Illegitimate triggers: one lost deal, a competitor's launch, founder boredom. The honest test before any pivot: have we actually FALSIFIED the current hypothesis with a real experiment, or are we fleeing discomfort? Serial pivoting every quarter is not agility — it's a company that never ran a clean test (MVP discipline applies to pivots too).
Executing without dying mid-turn
A pivot consumes cash and morale, so treat it as a project with a budget: check runway first — a turn needs 6+ months of fuel to reach new evidence; define the new hypothesis and its kill metrics BEFORE moving; decide what carries over (team, tech, customer trust) and what gets sunset — half-maintained legacy products are pivot poison; communicate the story straight to team and investors (the learning that led here IS the pitch); and time-box the validation. Companies rarely die from pivoting — they die from pivoting too late, too vaguely, or while dragging the old business behind them.
FAQ
How many pivots is too many? Count clean experiments, not turns — two well-run pivots beat five panicky ones; investors read the difference instantly.
Pivot with revenue coming in? Hardest case: milk the legacy to fund the new, but assign separate owners — one team cannot serve two masters.
Do you tell customers? The ones you keep — yes, early and honestly; sunset plans with migration paths protect the reputation the next product will need.
Related: Startup · Product-Market Fit · MVP · Runway Reviewed: 2026-07-16 · Sources: lean startup canon