Open any monthly ads report and you will find impressions, clicks, click-through rate, cost per click, maybe a conversion count. Every one of those numbers can go up in a month where the business made less money than the month before.
That is not because the platform is lying. It is because the platform can only count the events it can see, and the event that matters to you, the invoice that got paid, happens somewhere it has no access to.
The gap between a conversion and a sale
A conversion, by default, is a form submission or a phone tap. For an eCommerce store the gap is small: an order is close to real revenue, minus returns. For a trade, a clinic, a professional service, the gap is enormous. A submitted form is a stranger who typed their name in a box.
Some of those strangers were never going to buy. Some wanted something you do not sell. Some are worth twenty times more than the others. Optimise the campaign against form fills and the algorithm will faithfully find you more of the cheapest ones.
The algorithm optimises for whatever you tell it success looks like. Tell it badly and it will succeed at the wrong thing, efficiently.
Wiring the loop closed
The fix is unglamorous and mostly one-off. You give every enquiry a stable identifier at the moment it arrives, carry that identifier through whatever you use to track jobs or deals, and send the outcome back to the ad platform when it is known.
What has to be in place
- A click identifier captured on landing and stored with the enquiry
- One place where enquiries live. A CRM, a spreadsheet, a job board; it does not need to be fancy.
- A status that gets updated when a job is won or lost, by someone whose job it is
- An offline conversion upload back to the ad platform, with the real value attached
- Server-side event sending where browser tracking is unreliable
Once that loop exists, the monthly conversation changes completely. Instead of arguing about whether a click-through rate of two percent is good, you can say: these keywords produced work we won, these produced work we did not want, and this one produced nothing for three months.
What to do while it is not built yet
Do not wait for the perfect setup before making decisions. Two crude measures beat none: ask every caller how they found you and write the answer down, and count enquiries by type rather than in total. Both are manual and both are biased. They still tell you more than click-through rate does.
The rule underneath all of this: never let a number be reported that nobody can act on. If a line in the report has never once changed what you did next, take it out of the report.