From complaint to gold
Five stages that turn a customer complaint into a business worth keeping, or kill the idea cheaply. Each stage says when you're in it, when you're done, and when to turn back.
New marks a step added in this revision. Open any step to see when it counts as done.
Always
- Every answer you wait on gets a deadline. No reply from an adviser, lawyer or partner in 14 days: chase or switch the same day.
- The entity question never pauses selling. Conversations and commitments don't need a company. Only the money route does.
- A timebox that runs out forces a verdict the same day: reframe, kill, or one 7-day extension with a written reason. Only one.
- Every candidate carries its stage and the date it entered. Without that, no timebox can fire.
- Substeps inside a stage can run in parallel. Only the stage exits happen in order.
- A killed idea comes back only with new evidence that changes the reason it was killed. Name the evidence.
1Find
Scoop & sieve. Automated (OIS).
Turn complaints from the internet into a short list worth a human look.
Enter when
The watchers are running and healthy.Leave to Pick when
A candidate reaches the review shortlist.Set aside when
It fails the gate. It goes to the bank, can be re-panned, and is never deleted.1.1Capture. Watchers pull complaints from Reddit and 1–2★ Shopify reviews.›
Done when the row exists with a link back to the original post.
1.2Gate. Drop anything that isn't a real problem. Set it aside; never delete it.›
Done when the verdict is recorded: passed, or set aside with a reason.
1.3Score. Weighted ranking of what passed the gate.›
Done when the row has a total score.
1.4Shortlist. Send the best to human review.›
Done when the row is in the review queue, including rows that score above the review window.
2Pick
Wash & choose. Manual.
Choose one bet worth testing, and write down what would kill it.
Enter when
The shortlist has candidates and nothing else is in Assay.Leave to Assay when
The bet and kill criteria are written, and you can name where 10 buyers are.Kill or turn back when
The premise is dead (kill), or nothing survives the lenses (back to the shortlist).2.1Premise check. A dead market, channel or listing kills the idea, and so does a wedge incumbents already give away free. The number of competitors never does.New›
Done when the market, channel and listing are confirmed alive with a source link, the price for this wedge is confirmed above zero, or the idea is killed.
2.2Reframe with four operators.New A shift altitudeB shift buyerC shift artifact typeD shift market (language or region)›
Done when each operator has produced one or two framings, or an explicit "none".
2.3Collapse to root. Merge posts that share one pain.›
Done when the pain is written as one root cause.
2.4Profile with the 7 lenses, plus lens 8: am I inside the buyer's community?New›
Done when all 8 lenses are marked strong, weak or unknown. It's a shape, not a sum.
2.5Pick one, based on current need.›
Done when one framing is chosen and the rest are banked.
2.6Write the bet.New <buyer> will pay <price> for <smallest thing>, because <break>Add the expected metric move and a guardrail.›
Done when all four blanks are filled and the metric and guardrail are named.
2.7Write kill criteria before starting stage 3.›
Done when at least three observable kill criteria are written down.
3Assay
Prove it's gold.
Prove real buyers want it, that you can reach them, and that you can get paid, before building anything.
Enter when
A bet (2.6) and kill criteria (2.7) exist, and you can name where the 10 people are.Leave to Mint when
10 conversations are logged, at least one commitment is made, a channel reached them, and the get-paid check is clear.Kill or turn back when
A kill criterion hits, or one channel can't reach 10 people. Back to 2.2 to reframe, or kill.3.110 pain conversations about the pain and how people work around it today. Never a pitch.New›
Done when 10 conversations are logged, each with the pain and the current workaround.
3.2Make something concrete: a demo, listing, price, free slice or free tool.New›
Done when the artifact exists and can get a clear no.
3.3Channel test. Reach buyers where you would sell. Test one channel at a time and drop a dead one fast. If the channel needs a working product to enter (app stores, marketplaces), prove the buyers are reachable another way now, and run the real test inside Mint.New›
Done when one channel has reached the buyers, or has been declared dead. For gated channels: reachability shown by proxy, such as competitor review volume or buyers contacted directly.
3.4Ask for commitment: a preorder, "send me the link", or money.›
Done when every person has been asked, and each answer is recorded.
3.5Can-I-get-paid check. Can the first payment go through what you already have? Legal entity, payment rails, compliance. Send any question to your adviser on day 1 of Assay.›
Done when you know the exact route: which entity invoices, how the money arrives, and whether this product needs anything special. Nothing undecided.
3.6Verdict against the 2.7 kill criteria: go, reframe (back to 2.2), or kill.›
Done when the verdict is written with its evidence.
4Mint
Build the smallest thing you can sell.
Build only what delivers the promise, and get it into real hands until someone pays.
Enter when
Assay says go: at least one commitment, and the get-paid check is clear.Leave to Weigh when
The first payment lands (4.7). The 90-day clock already started at 4.6.Kill or turn back when
Office hours doesn't survive (back to 3, or kill), or day 45 passes without a payment. Cut scope or go back to Assay; don't extend.4.1
/office-hours(gate G3).›Done when the verdict says it survives.
4.2Promote it to a project. Only you can do this (gate G5).›
Done when you have created the project row, within the WIP cap.
4.3Define the activation event before building.New›
Done when it's one measurable sentence: the moment the user got value.
4.4Build only the top-line promise from 3.2.›
Done when the promise works end to end for one real user. For gated channels, the listing goes live here and the real channel test starts.
4.5Minimal tracking: one event per onboarding screen, plus activation. No dashboards.New›
Done when those events fire in production.
4.6Milestone: first external user.›
Done when someone other than you reaches the activation event. The 90-day clock starts.
4.7Milestone: first dollar.›
Done when a payment lands, and your adviser has one question: at what revenue, customer geography or contract type does your current setup stop working?
5Weigh
Keep, kill, learn.
Grow by hand, decide keep or kill on evidence, and make every kill teach stages 1 and 2.
Enter when
The first payment has landed (4.7). The clock counts from the first external user (4.6).Keep when
At day 90 you are at rung 30 or above, with most users coming from one channel. At 10–29 with a channel: one more 90-day cycle, once.Kill or reframe when
At day 90, below rung 10: kill. 10–29 with no channel: reframe (back to 2.2) or kill. 30+ all from your own network: back to 3.3.5.1Climb the ladder by hand. Each rung answers one question; don't climb until it's answered. Users means activated users.New›
Done when you reach rung 100, or the day-90 verdict arrives first. Note the paying count at every rung.
- 1Works for a strangerYou bring them. Watch them use it.
- $1Someone paysYou bring them. Ask what nearly stopped them.
- 10The sale repeatsYou, one by one. Talk to every one. Stuck under 10 for 30 days: back to 3.3.
- 30A channel existsMost from one channel, not your network. Metrics start here.
- 100The channel repeatsMostly the channel. If growth only comes when you push, it's you, not the channel.
Priced at €200/mo or more? Count paying accounts instead: 1 · $1 · 3 · 10 · 30.
5.2Track activation, W1/D30 retention and ARPU, starting at rung 30.New›
Done when all three can be read every week without building anything new. Below rung 30, conversations replace numbers.
5.3Numbers show where users drop off; conversations show why.New›
Done when every drop-off you act on has at least one conversation behind it.
5.4Day-90 verdict on leading signals, not revenue: which rung are you on, and does one channel bring most users?New›
Done when the verdict is written on day 90. 30+ with one channel: keep. 10–29 with a channel: one more cycle, once. 10–29 without: reframe or kill. 30+ all from your network: back to 3.3. Under 10: kill.
5.5Double down on the channel that works. Save SEO, comparison pages and free tools for this step; they are scale levers.New›
Done when most of your acquisition effort goes to that one channel.
5.6Incorporate when a trigger fires: revenue nears your adviser's threshold, a customer or platform requires an entity, real liability appears, or the day-90 verdict says keep. Never before the first dollar.New›
Done when the entity exists and new payments flow through it.
5.7Kill the rest, and keep the lesson.›
Done when the killed thing is shut down with no running costs.
5.8Feed each kill reason back into stages 1–2.›
Done when the reason exists as a criterion in stage 1 or 2.
Parked until there's traction
- Spanish and LATAM sources for the watchers.
- More channels and viral loops (watermarks, referrals).
- A/B testing.
- Public in-thread replies (OIS-10). Parked, not rejected.