Pivoting — what to change, what to keep

Persevere when:

• Cohort retention is flat or improving (the customers who stay, keep staying)
• A specific customer segment is consistently delighted, even if it's smaller than you wanted
• Customer interview language is consistently energetic for the problem (not the solution)
• You're learning something new every month — the iteration loop is fast

Pivot when:

• Retention curves all decay to zero — no segment has PMF
• Customer interview language is polite but flat — no one cares enough
• You've iterated 5+ times on positioning and conversion is still grim
• A specific subset of users does something unexpected — that's the wedge

The most common pivot mistake: pivoting too late. Founders defend the original thesis for 6-9 months past the point the signal is clear. The second most common mistake: pivoting too fast — chasing whichever customer requested the loudest feature.

Make the pivot call in a structured way:

1. Write the original thesis on paper
2. List what you'd expected by month 12 vs what's true
3. List the strongest unexpected signal
4. Sleep on it
5. Decide explicitly: persevere, pivot, or kill

Zoom-in pivot: a feature of your current product becomes the whole product. Slack started as a gaming company; the team-chat feature they built for themselves became the product. Keep: technology, team. Throw out: original positioning and most of the codebase.

Customer-segment pivot: same product, different customer. You built for marketers but it's the developers who love it. Keep: product, technology. Throw out: positioning, marketing, sales motion.

Problem pivot: same customer, different problem. You built CRM for agencies and discovered they actually need invoicing more. Keep: customer relationships, brand. Throw out: product, positioning.

Technology pivot: same customer, same problem, fundamentally different approach. You built rule-based moderation; it doesn't scale; you rebuild with ML. Keep: customer relationships, problem definition. Throw out: codebase.

Most "pivots" you read about are actually customer-segment or zoom-in pivots. Pure problem pivots are rare and often look like new companies. Technology pivots are common but rarely make headlines.

Parallel bets vs hard pivot:

• Hard pivot — abandon the original; commit fully to the new thesis. Required when team morale or runway demands a decisive call.
• Parallel bet — keep the old generating revenue while one or two people validate the new thesis. Required when revenue from the old keeps the lights on and the new thesis isn't yet validated.

Parallel bets sound nice but rarely work past 60 days. Founders end up doing both halfway. If you find yourself in a parallel-bet that hasn't resolved in 8 weeks, force a decision.

What to keep across a pivot:

• The team's domain knowledge (almost always)
• Customer relationships that pre-date the product (often)
• Brand and domain (sometimes)
• Specific tech infrastructure that's actually reusable (rarely — most reuse is a sunk-cost rationalisation)

Communication is the unglamorous critical part:

• Existing customers: tell them what changes for them. Many will be fine; some will churn. Don't hide the pivot.
• Investors: write the pivot memo before the term sheet conversation. Show the data, the new thesis, the new milestones. Investors generally respect a decisive call.
• Team: tell them in person, in a room, in one conversation. Give every team member a clear answer to "what do I do Monday?" before they leave the room.
• Public: low-key. Don't blog about the pivot until you've shipped something for the new direction. Pivots-as-content rarely age well.