Pivot or persist

Pivot-on-mood. Two bad weeks, a competitor's launch, a podcast about someone else's market, and suddenly the grass is a thesis. The tell is timing: the "pivot" always follows the morale dip, never the data. Check your proposed pivot's date against your own morale graph. If they correlate, you've learned something, and it isn't about the market.

Persist-on-sunk-cost. The years invested become the argument. They're a receipt for money already spent, never a reason to spend more. The nastier variant is persistence inside a tarpit: ideas that attract founders precisely because thousands of teams have died on them (consumer social, marketplaces for X), where the graveyard is fully documented and your grit is just tuition.

Underneath both errors sits the variable founders refuse to price. Paul Graham's line: startups die of suicide more than homicide. Morale is a real input to survival, not a soft one. Ignoring it doesn't make you rigorous; it makes your model wrong, because the model is missing the term that kills most companies.

1. Run Caldwell's two triggers, and require both. Dalton Caldwell's test for a real pivot signal: (a) are you genuinely out of ideas on the current approach? Write the list of untried moves and look at it. And (b) are user conversations producing nothing new? If the idea list is empty AND the last twenty interviews sound like the first twenty, that's a real signal. Either one alone: keep working, you still have material. This gate exists because "out of ideas" usually means "out of energy for the ideas I have," which is a different diagnosis with a different treatment.

2. Separate evidence from energy. Two columns, written. "The data says no": churn, flat cohorts, dead interviews, with numbers attached. "I'm tired": sleep, dread, envy of other people's markets. Both columns are real. Only one is about the market. A pivot driven by depletion is premature, because the fix for depletion isn't a new market, it's rest, structure, or an honest run at the quit question. Every vehicle is hard. Switching vehicles to escape hardness just resets the clock on a different hard.

3. Run the tarpit check, both directions. The idea you're leaving: is it a known graveyard? If so, leaving isn't weakness, and your persistence was never edge; the sooner that's said, the cheaper it gets. The idea you're entering: who died there? Name three corpses and what killed them before you credit your escape route. If you can't name any, you haven't researched the destination, you've just fallen in love with not being here.

4. Check the vector. The best pivots move toward founder expertise, never just away from pain. Away-from is a flinch with a deck. Toward is a thesis: what do you know from real reps that the new market's incumbents don't? Answer in one sentence or admit you're fleeing. The strongest pivot signal in the whole procedure is when the same interviews that were dead for the current product keep surfacing a live wire pointing somewhere specific, especially somewhere one of the founders has history.

5. Price the morale, explicitly. In the health and identity currencies: what do six more months of persist cost the founders, in sleep, in relationship strain, in who you're becoming? What does starting over cost in the same units? Neither answer decides alone, but both belong on the page. Ruin check: if the new thesis needs longer to prove than your personal runway allows, it needs revenue dates attached, or it's a cliff with a vision statement.

6. Resolve with kill criteria, whichever way you go. Set the state and the date now, while you're cold. Persist: "if activation is under X by June 1, we pivot, no meeting required." Pivot: the same discipline pointed at the new thing, "three paying pilots by November 1 or it isn't a market." This converts the next version of this conversation from mood into observation, which is the entire point of deciding anything on paper.

• The adjacent shiny idea that appears two weeks after a bad month. Check its timing against your morale graph before crediting it as insight.
• Pivot as vacation. The new market feels better because it hasn't said no yet. It will.
• Keeping the old thing half-alive. A pivot with a hedge is two part-time failures. Sunset with a farewell note, or don't call it a pivot.
• Serial 90-day pivots. That's mood with a changelog.
• Crediting grit inside a tarpit. The graveyard doesn't grade effort.

Two founders, 14 months into a social app for book clubs. 4,000 weekly actives, flat for five months. Trigger check: the untried-ideas list has two entries, both reskins of features that already failed, and the interviews have produced nothing new since month six. Both triggers live. Evidence versus energy: the data column is damning on its own, and the founders are tired but sleeping, so depletion isn't doing the deciding. Tarpit check backward: consumer social around content is a documented graveyard; the 14 months were tuition, not edge, and writing that sentence hurt more than the pivot did. Vector: one founder spent six years in publishing analytics, and the dead interviews kept surfacing publishers asking about reading-behavior data. Toward expertise, with actual pull. The pivot: reading analytics for mid-size publishers. Kill criteria written the same day: three paid pilots at $2k a month by November 1, or the quit question gets run with clean hands. The old app sunsets with a farewell note. No hedge.

November 1 arrives with five pilots. The wire never fires, and nobody ever has to argue about mood again. That last clause is the whole reason the memo exists.