An investor asks how marketing is going. The answer you have ready is a list: posts published, the site refreshed, two campaigns live, a conference booked. All of it is true, and none of it tells you whether to spend the next quarter the same way. You can describe the quarter in detail and still not know whether it worked.
A number earns its place by changing a decision.
The test is small and unforgiving. If the number came back bad next month, what would you do differently? If the honest answer is nothing, or try harder, the number is decoration. It is on the slide because the slide has a space for it.
Most reporting fails here before it fails anywhere else. It is assembled to show effort rather than to settle an argument. Effort is not in question at a startup; nobody thinks the founder is idle. What is in question is whether this quarter’s work should be repeated, and no amount of activity answers that.
The first test is whether you can move it this quarter.
Revenue is the number everyone respects and the wrong instrument for this. Between pre-seed and Series A, the lag from a marketing decision to money in the bank is longer than the gap between board meetings, and the line moves on one deal closing or slipping. A number that takes three quarters to respond cannot tell you what to do in this one.
So you want the earliest point in the chain that is genuinely yours: the step where your work shows up first, and where a bad month is visible while you can still act on it. Qualified conversations started. Demos booked from a named source. Trials that reached the step where the product does something for the person using it.
The opposite failure is a number that moves only when you spend more. Impressions do that. So does traffic bought at a fixed cost per click. Those are spend meters. They go up whenever you feed them, which is another way of saying they will never tell you to stop. A budget set against a spend meter is a standing order, not a decision.
The second test is whether a stranger could check it.
Write the definition down and give it to someone who was not in the room. Give them access to the source and ask for the same number. If they come back with a different one, or if they need you to explain the method first, you do not have a number. You have an interpretation.
This kills more dashboards than founders expect. Qualified is the usual casualty. Qualified by what: company size, stated budget, a problem they named out loud? Two people at the same company will count differently, and the count drifts in whichever direction the person counting needs it to go.
It matters outside the building too. The first thing a good investor asks after a metric is how you count it. If the answer takes a paragraph, the metric was decoration, and the paragraph is the tell.
Most of the numbers a founder reaches for fail both tests.
Followers, impressions, and sessions behave like spend meters, so they fail the first. They fail the second because the definitions belong to the platform, and a session or an impression means a different thing in each one. Awareness fails the second outright; there is no source a stranger could open.
Pipeline influenced fails it too, because influence is a modeling choice rather than an observation. A north star copied from a company two stages ahead fails both, having been built for decisions you are not making yet.
The trap is that the honest candidates pull in opposite directions. Revenue clears the stranger test perfectly and fails the quarter test. Activity clears both tests and still changes nothing, because a post count does not tell anyone what to do next. So founders report effort: the number they trust is too slow, and the fast ones are free.
The number you want sits between them: close enough to money that a bad quarter would show in it, close enough to the work that this month’s decisions move it.
The number that survives is smaller than the one you wanted.
For most companies at this stage, it is one line, and it is unglamorous. The first half is how many qualified conversations started this month, and where each one came from. The second half is what makes the number worth having: it tells you which channel to add next. What you end up with is a number, a written definition, one source, and a fixed date each month when someone looks at it.
Everything else becomes diagnostic. Traffic, rankings, and open rates explain the number; they are not what you answer for. That sounds like a small distinction. It changes how the month runs, because diagnostics can be messy when nobody is being judged on them.
Accept what a real number costs. It can go down. A number that can go down in front of your team and your board is the only kind worth reporting, and the first month it does is the month it starts being useful. The reports we see founders understand and act upon had three numbers at the top and their definitions underneath. One of the three is the number the quarter is judged on; the other two explain how it moved.
It should not be permanent either. Change it when the constraint moves, when the problem stops being finding conversations and starts being closing them. Change it on purpose, say why, and keep the old one visible for a quarter so the change does not read as a rescue.
Analysis is the fifth thing we do for a startup, after Brand, Positioning, Website, and Demand, and it is what decides which of the four gets repeated. Before we write a plan, we agree on the targets—actual numbers—define them in one sentence, and put a date on the first review.
Tell us what you’re building.