measurement.no cart, so the metric gets set first.

There's an entire category of work the conversion playbook doesn't cover. That playbook was written for e-commerce and subscription software: it assumes a buy button at the end of the journey and assumes that button leaves a record. When the record doesn't exist the whole method collapses, and what usually replaces it is worse than having no method at all.

Two symmetrical mistakes follow. The first is reporting whatever can be counted, impressions, clicks, community, in the tone of someone reporting sales. Nobody lies outright; it's simply left unsaid that none of those figures is a purchase. The second is giving up: accepting that nothing is measurable and falling back on taste, where whoever talks loudest in the meeting wins.

There's a third one, quieter and considerably more expensive: measuring something different every month. When the number picked in January comes in weak in March, it gets replaced by one that looks better. A year later there's no series, no learning and no way to tell whether anything worked.

Before touching a campaign there's one deliverable that comes first and almost never gets asked for: writing down what counts as a conversion on this account. It isn't a formality ahead of the work, it is the work. Without it, optimising is decorating.

It's also the deliverable that protects everything downstream. Once the definition exists in writing, every later argument about whether a month went well has a fixed reference point instead of a moving one.

The method is three questions, in this order. First: what would the brand want to happen if the internet didn't exist? It cuts through the noise immediately, because nobody answers clicks. They answer that someone picks the product off the shelf, that someone decides to visit on a Saturday, that someone walks into a medical appointment better informed.

Second: which measurable digital event sits closest to that? Not the one that's easiest to show, the closest one. And third, the uncomfortable one: is that event tracked today? The usual answer is partly. That gap isn't a reason to postpone: it's the first finding of the audit, and often the highest-impact recommendation in the whole engagement.

Picking the intermediate signal is the real technical decision, and not every signal is equal. A good one meets three conditions: it sits close to the behaviour you're trying to cause, it can't be inflated cheaply, and it will still read the same way six months from now.

It helps to treat conversion as a ladder rather than an event. At the bottom sits qualified arrival; then shallow interest, deep interest, the primary conversion, the closest thing to the objective that can still be measured, and at the very top the action that happens off-screen, which almost never gets measured. Choosing a rung partway up is legitimate. Presenting it as if it were the top rung is not.

That produces a reporting rule worth applying everywhere: name the rung. A 12% conversion rate says nothing. That 12% of arrivals completed the tool says something, and can be argued with. In health the logic is identical but the room to move is narrower: no manufactured urgency, no pressure, and no sensitive data going anywhere near a pixel. That work is run by Champe Health, under its own rules.

A metric only starts earning its keep once it has history. Month one says nothing; month six says a lot. So the real commitment isn't choosing well, it's not redefining halfway through, precisely when the number isn't cooperating. A weak month on the same metric is information. A strong month on a brand-new metric is a change of subject.

The thing to do is agree it up front, in writing: what the number is, who looks at it, how often, and what happens when it drops. An indicator that falls and leaves everyone unsure what to do isn't a KPI, it's an ornament.

And there's one outcome worth saying early: sometimes there's nothing to optimise. If every sale happens in a physical location and there's no digital surface where a person can move forward, inventing an online conversion is inventing a problem. In that case the job is a different one, getting content to land and get consumed, and saying so in month one is worth more than three quarters of decorative metrics.

The note explains the problem; the services explain the work.