Four signals that mean stop, four that only feel like it, and why the criteria have to be written before you need them.
Knowing when to kill a SaaS idea is harder than knowing when to start one, because by the time the evidence is clear you have already invested enough to want it to be wrong.
Two of the four is a serious warning. Three is a kill. One is normal and means keep going.
Set these criteria before you need them. Deciding what would change your mind while you are emotionally neutral is the only reliable defence against deciding while you are not.
Before starting, write one sentence: what would have to be true in ninety days for me to stop?
Make it specific and observable. "Fewer than three people agree to a paid pilot" works. "It does not feel like it is working" does not, because that sentence can be reinterpreted indefinitely.
Founders who skip this step do not fail to kill ideas — they kill them eventually, after nine months instead of three, having learned the same thing at six times the cost.
Nobody describes the problem unprompted. In ten conversations with people who should have it, count how many raise it before you do. If the number is zero, you are describing a problem people recognise when prompted but do not experience as a problem. Recognition is not demand.
Nobody pays anything today. Not to a competitor, not to a freelancer, not in staff hours. Zero existing spend means you would be creating a budget line rather than redirecting one, which is slow work for one person.
Enthusiasm without commitment. The most misleading signal in early validation. People are generous with encouragement and precise with money. Ask for a card, a pilot date, or an introduction to their boss — a yes that evaporates at any of those three was never a yes.
Two failed relaunches. One quiet launch is a distribution problem. Two, to genuinely different audiences, with the positioning fixed in between, is information about the idea rather than the channel. Work through the flopped launch diagnosis before counting either one.
Four things that feel terminal and are not.
| Signal | What it usually means |
|---|---|
| A competitor exists | Demand is proven. This is good news. |
| Slow growth in month two | Normal. Three to six months to $1k MRR is typical. |
| One customer churned | With eleven customers this is noise, not a trend. |
| You are bored of it | A stamina problem, not an idea problem. |
The competitor one is worth dwelling on. Founders abandon ideas on discovering someone else is doing it, when a funded competitor is the clearest demand signal available — it means budget exists and buyers are already educated. The positioning framework covers how to enter a market that already has players.
Most ideas that look dead are actually pointed at the wrong people. Before killing, check whether anyone stayed.
If even two or three customers use the product regularly, the idea is not dead — the segment is wrong. Find what those people have in common and narrow everything toward them. That is a positioning change, not a rebuild, and it is far cheaper than starting again.
If nobody stayed and nobody activated, the idea itself is the problem. Reading fit at small scale covers how to tell the two apart with a handful of users.
Ask one question before killing anything: has this idea ever had ten users who reached first value? If not, you have tested your distribution and your onboarding, not your idea. Those are cheaper to fix than a new product is to build.
Three things, and the third is the one founders skip.
Tell your customers. Thirty days' notice, an export of their data, and no excuses. Founders who wind down well get remembered well, and the next thing you build launches to people who trust you.
Keep the assets. The audience, the mailing list, the domain knowledge, the code. Most of it transfers. The distribution you built is worth more than the product it pointed at.
Write down what you learned. Two paragraphs, the same week. What you believed, what turned out to be true, what you would test differently. Without this the next idea repeats the same error with different branding.
Do not start the next thing immediately. The instinct to replace a dead idea within a week produces a choice made on relief rather than evidence.
Use what you learned instead. If distribution killed the last idea, weight distribution heaviest — choosing between two ideas covers the five tie-breakers in order. If demand killed it, validate more cheaply next time before building.
Founders coming off a fast AI-assisted build often find the product was never the constraint — what comes after vibe coding covers that specific pattern.
When two or more of four signals hold: nobody describes the problem unprompted after ten conversations, nobody currently pays anything to solve it, people who said yes will not commit money or a date, and you have relaunched twice to different audiences with the same result.
No. A funded competitor is one of the clearest demand signals available — it proves budget exists and that buyers are already educated about the problem. Entering an existing market is a positioning question, not a reason to stop.
Write kill criteria before you start and set a ninety-day horizon. Make them specific and observable, such as 'fewer than three people agree to a paid pilot'. Vague criteria can be reinterpreted indefinitely, which is how nine-month failures happen.
Rarely. Three to six months to reach $1k MRR is normal for a solo founder with no audience. Slow growth in month two is a miscalibrated expectation rather than evidence about the idea.
Check whether anyone stayed. If even two or three customers use the product regularly, the segment is wrong rather than the idea, which is a positioning change rather than a rebuild. If nobody activated at all, the idea itself is the problem.
Tell Marcus what you have tested and what came back. You get a specific read on whether the idea, the segment or the channel is the problem.
Try GhostCoach free →14-day free trial · cancel anytime · 30-day money-back on Lifetime