Defining a conversion event precisely enough to measure

Ask a marketing team what their conversion is and you will get an answer immediately. Ask them to define it precisely enough that a machine could count it without a human adjudicating edge cases, and the conversation slows down considerably. That gap is where most measurement projects quietly fail.

The failure is not technical. The tracking works. The number appears on a dashboard. It is just that the number is counting something slightly different from the thing the business cares about, and nobody notices until someone tries to reconcile it against finance.

The most common answer we hear is some variant of "leads". It is not a definition. A lead is a business concept that sits on top of an action, and the action is what gets measured.

The action might be a form submission. But which form? The contact form, the demo request, the gated download, the newsletter box in the footer? These have wildly different commercial value and, in most implementations, they all fire the same event. Once they are pooled, source ranking becomes meaningless: a channel that drives newsletter signups outranks a channel that drives demo requests, because there are more of the former.

The fix is not clever attribution logic. It is picking one action and being willing to say the others do not count for this purpose.

The second break is qualification. A form submission is an action a visitor takes. Whether it was a good one is decided later, often by a human, often in a different system.

If you measure raw submissions, you will optimise toward whichever channel produces the most submissions — and that is reliably the channel producing the worst ones. Cheap traffic converts on low-friction forms at impressive rates. The submissions are real. They are also worthless, and your channel report will recommend you buy more of them.

There are two honest ways out. Either define the conversion as the qualified state and accept the reporting delay that comes with it, or define it as the raw action and pair it with a second, slower measurement of quality by source. What does not work is measuring the raw action and talking about it as though it were the qualified one.

A surprising number of conversion events can fire repeatedly for the same visitor: a subscription renewal, a repeat purchase, a second demo booking after a stalled deal. Whether those count as one conversion or several changes the ranking materially, and the answer depends on the commercial question being asked.

For ranking acquisition channels, first conversion per visitor is usually the right unit — you are measuring which source produced a customer, and the second purchase was not produced by the source. For measuring account value, it is not. Decide explicitly, because the default in most tools is to count every occurrence and nobody will tell you that is what happened.

The last piece is the state change that makes the event commercially real. A purchase event fired on the checkout success page counts orders that will later be refunded, cancelled, or fail payment capture. In some businesses that is a rounding error. In others — high-ticket, subscription, anything with a trial — it is the difference between a channel looking profitable and being profitable.

You do not always need to solve this on day one. You do need to know which version you are measuring, and to say so on the dashboard, so that nobody builds a budget on a number that was never claiming to mean what they assumed.

A conversion event is defined well enough to measure when you can answer four questions without hedging. What is the specific action? Which visitors' instances of it count, and which are excluded? Does it count once per visitor or every time? And what later state, if any, has to hold for it to still count?

Answer those four and the tracking is straightforward. Skip them and no amount of tracking sophistication will save the number.