Market validation metrics: deriving a benchmark instead of borrowing one

Every market metric worth having is a ratio between two things you can count. The trouble is that at the point you most want them, you have almost nothing to count, and the temptation is to fill the gap with a figure from somewhere else.

Do not. A borrowed ratio was produced by a different offer, at a different price, through a different channel, at a different time. It will be internally consistent and irrelevant to you. What follows is what you can measure, in what order it becomes measurable, and the arithmetic that connects them.

What to take away

The one identity that matters

Write it with letters rather than numbers, because your numbers are the only ones that apply.

Let V be what one customer is worth over their stay, after the cost to serve them. Let A be acquisition cost: total channel spend divided by customers produced.

The business works when A sits comfortably below V, and how comfortably depends on how long you wait to get V back.

That waiting period is the second number: how many payments it takes to recover A. Two ventures with identical ratios and different recovery periods are in completely different situations, because the slower one needs more cash to grow at the same rate. The recovery period is one of the fields a customer interviews framework fills before any round is booked.

Everything else here is a way of getting an honest A or an honest V.

What you can measure, and when

The last row is where most plans break. Value depends on how long people stay, and that is unknowable until enough time has passed. Anything you write down before then is an assumption and should be labeled as one in the same document.

Counting the denominators honestly

The arithmetic is easy. The counting is where the errors live.

Count everyone who entered, not everyone who finished. A conversion rate calculated over the people who reached the last step measures your last step and nothing else.

Separate warm from cold. A response rate mixing people who know you with people who do not is two numbers averaged into one that describes neither. Keep the columns apart, as in the discovery counts.

Include your own time in acquisition cost. Unpaid effort is the largest input early on and the one that will not scale. Price it at what you would have to pay somebody, and the channel comparison changes.

Count refusals. People who saw the price and declined are the most informative group you have, and they vanish from every funnel that records only successes.

Why small samples lie

A percentage built from a handful of events moves enormously when one event changes. Two out of five and twelve out of thirty are the same rate, and only one of them survives another week of data.

Two habits fix most of it. Always write the fraction, so the denominator is visible. And when a small sample looks decisive, ask what would have happened if one event had gone the other way. If the conclusion flips, you do not have a result.

This matters most at the moment you are deciding whether to make an irreversible commitment, which is when a small sample is most tempting to believe. The reasoning that keeps that honest is worked through in the readiness decision.

Numbers that feel like evidence

Each becomes useful only when paired with behavior that cost somebody something.

Deriving a benchmark instead of borrowing one

You do not need an industry figure to know whether your conversion is acceptable. Work backwards from the identity.

Take what one customer is worth to you. Decide what share of that you are willing to spend acquiring one, which gives you the acquisition cost you can afford. Divide by what it costs to put the offer in front of one qualified person, and you have the conversion rate you need. Compare that with what you are seeing.

The comparison is meaningful because every input is yours. A comparison with a number from elsewhere is not, however authoritative the source looks.

For real external figures on population or industry size, use statistical agencies, regulators, or industry bodies. Write the date beside each figure.

The federal series on business births, deaths and survival is one source. Listed companies describe their own markets in filings searchable through the SEC's full text index. The wider set worth using is laid out in market validation.

Common questions

What is the smallest useful set?

Three: cost to reach one qualified person, conversion from qualified to paying, and margin per sale. Those tell you whether the channel can work at all, and they are measurable before you know anything about retention.

When can I trust a customer value figure?

Once you have watched a full purchase cycle for a group of customers who all started at the same time, and can see how many are still there. Before that, write it as a range and mark which end the plan depends on.

How do these connect to price?

Directly. Raising price raises value per customer and usually lowers conversion, so the thing to watch is not either one but their product against acquisition cost. Testing that properly is pricing validation work.

What should I measure in the first weeks after launch?

Fewer things than you would expect, and mostly whether the customers you got behave like the ones you predicted. That is a different set: see the first hundred days.