Short answer
When the strategy changes with every prospect email, the company does not have a weak strategy. It has no strategy, only a queue of reactions. The tell is that nobody can say what you will not build. The fix is two artifacts and one habit, and it takes about a quarter.
Monday the direction is upmarket enterprise. Wednesday a mid-market prospect wants a feature and suddenly there is a self-serve motion on the board. Engineering has restarted the same integration three times under three different priorities, each time abandoning work that was two weeks from done.
Your best engineer asked in the last retro what the company is actually building, and the honest answer was that it depends who emailed. Nobody in that room thought they were describing a strategy problem. They were describing the most expensive one available at this stage.
Why the restart tax stays invisible
Every swerve arrives with a real prospect, real money and a real argument attached, so refusing one always looks like the unreasonable move in the moment. The damage is never local, though. It is systemic and it compounds: work started and abandoned generates no value at all, the team learns that persuading the founder beats following the plan, and nothing is pursued long enough to compound into an advantage. In our experience this is the single most expensive silent tax at $2M to $15M ARR, and it appears on no dashboard because no metric counts unfinished work.
What the team learns
Culture is what gets rewarded, and in a swerving company the rewarded behaviour is lobbying. Product managers learn that the route to shipping something is a persuasive Slack message to the founder rather than a well-evidenced case at the review. Once that is learned it is very hard to unlearn, and it survives the arrival of any new process laid on top of it.
“Founders describe this as being responsive. From inside the engineering team it reads as being unable to finish anything. Both descriptions are accurate, which is why the conversation goes nowhere until somebody counts the abandoned work out loud.”
The test: three questions, one minute each
These are deliberately blunt, and the value is in how quickly they can be answered rather than in the answers themselves. A company with a strategy answers all three without pausing. A company with a reaction queue pauses on the first one, and the pause is the diagnosis.
- What are we explicitly not building this year? No answer means no strategy, whatever the deck says.
- When did we last hold a priority for a full quarter under real pressure?
- Where does a good idea go when it arrives mid-quarter? If the answer is “into the sprint”, you have found the mechanism.
The two artifacts and the one habit
The fix is unglamorous and it works. Two written artifacts give the company somewhere to put a good idea that is not this sprint, and one habit decides when direction is allowed to change. Together they keep the learning that prospect conversations generate while removing the whiplash that currently comes with it. Notice that none of this makes the company slower to learn. It makes the company slower to reorganise around a single conversation.
- A one-page strategy carrying three commitments and five named exclusions, visible to sales rather than filed with the board pack.
- An opportunity log, reviewed monthly, so a new idea has a home that is not the current sprint. This is the artifact that makes “not now” possible without saying “never”.
- A fixed decision cadence: direction changes at the monthly review with evidence, never in a deal thread.
- One visible case where a real opportunity went into the log and the quarter held. Nothing teaches the new rule faster.
Prospect-driven learning still flows in. It simply flows through a door instead of through the walls, and the difference between those two is a full quarter of finished work per year in most of the companies where we have installed it.
Your options
Three, and the honest difficulty with the middle one deserves stating plainly rather than being buried: the founder is usually the person doing the swerving, so founder self-enforcement has a poor record. It is not an impossible route, and it is considerably cheaper than the third, but it requires the founder to be the one person in the company who is told no by a process they themselves built.
| Option | What it takes | What changes | Fails when |
|---|---|---|---|
| Do nothing | Nothing | The restart tax compounds and the people who like finishing things leave first | Immediately, invisibly |
| Founder-enforced | The exclusions page plus a protected monthly review | Direction holds if the founder holds | The founder is patient zero for the swerve |
| Outside leadership | A monthly retainer for senior product leadership | Holding the line becomes politically possible, because the no comes from evidence rather than from contradicting the founder | Nobody internal is left to execute the held priorities |
Our own version of the third row is $8,000 per month for 20–25 hours on a 3-month minimum, with the first 60 days built around 60–80 customer and prospect interviews. The interviews matter here specifically because they replace the single loud prospect with a pattern, and a pattern is the only thing that reliably beats a compelling anecdote in a founder’s mind.
Roadmap swerving with every deal?
Thirty minutes, no pitch deck. You will leave with the exclusions question answered.
Book a Product Strategy SessionFrequently asked questions
Is reacting fast to prospects not a startup advantage?
Learning fast is the advantage. Reacting fast without a filter is a different thing wearing the same clothes. The opportunity log keeps all of the speed and drops the whiplash, because a prospect signal still reaches the strategy, just at the monthly review rather than on Wednesday afternoon. Companies that make this change do not become slower; they become able to finish.
What if the prospect is genuinely large?
Then it deserves a real decision at the decision forum with real evidence, held this week if the timeline demands it. A door, not a wall, and definitely not a reflex. The rule is not that big deals cannot change the plan. It is that changing the plan is a decision somebody makes deliberately rather than something that happens to you.
How long before the exclusions page stops being ignored?
About one quarter, and the turning point is always the same: the first time a genuinely attractive opportunity goes into the log and the quarter holds anyway. Until that happens the page is a document. After it happens the page is a rule, and the team starts using it in their own arguments without being asked.
Who should own the opportunity log?
Whoever owns product decisions, and it must be a person rather than a shared inbox. The log’s whole purpose is that someone reads it monthly, groups the entries into patterns, and brings the two or three that now have weight to the review. An unread log is worse than none, because it converts a real no into an implied yes.
Strategic whiplash is not a discipline problem and it is not a founder character flaw. It is a missing artifact and a missing cadence, and both are cheap. If the deeper issue is that nobody is authorised to hold the line at all, our guide to whether you need another PM or a product leader covers which gap you are actually looking at.