Revenue attribution
From click to contract.
The board does not care about CTR. It wants to know which click became a contract. That answer exists, and it does not require handing personal data to the advertiser.
The problem
The campaign optimises for forms. The company lives on contracts.
Google learns from what you report. If what goes back is "someone filled in a form", forms are what it will go and find, and it will get very good at it. A pile of bad leads is a possible outcome, and it is the default one.
The board is not asking how many forms. It is asking which ad paid for which contract. Between those two things there is a path, and it is almost never built.
The path
Five stops, and none of them assumes the previous one.
The click identifier enters the site, travels alongside the lead, waits in the CRM for as long as the sale takes, and goes back to Google when the stage changes.
- 01
Ad
The click carries an identifier issued by Google itself: gclid, gbraid or wbraid.
- 02
Site
The identifier is stored only for visitors who consented. Without consent it is not stored.
- 03
Platform
The lead arrives with the identifier alongside it, in the same send, not in a second call that can fail on its own.
- 04
CRM
The opportunity keeps the identifier and signals when the stage changes: qualified, won.
- 05
Google
The outcome returns as an offline conversion, with the identifier and the date.
What travels, and what does not
Google receives the click identifier and the outcome, nothing else.
Travels
- The click identifier, issued by Google itself
- The stage the opportunity reached
- The date and time, with the timezone declared
- The value, when there is one and when you authorise it
Does not travel
- Name
- Phone
- Any identifier of the person, hashed or not
This is not anonymisation or aggregation. It is that the click identifier is already enough: Google issued it, and Google is the one who knows which campaign it belongs to.
Consent
Whoever revokes stops being reported.
The identifier is stored only for people who accepted. Without consent it never enters the browser, never travels with the lead, and does not exist to be reported later.
And it is not one switch for everyone: it is per person. When someone revokes, what was stored is removed and their opportunity stops feeding Google. Consent is not a state of the campaign. It is a state of the person.
What usually breaks
Five defects that measure wrongly without saying so.
- 01The conversion fires on the button click, not when the server confirms. A form that failed becomes a recorded conversion, and the campaign learns from an outcome that never happened.
- 02The CRM has no field for the click identifier. The lead arrives, the sale happens, and there is nowhere to write down where it came from. The bridge breaks at the cheapest stop to fix.
- 03Two stages with the same name in the funnel. The signal fires on the wrong stage, and the conversion goes back to Google with its meaning swapped.
- 04The date goes without a timezone. The conversion lands a day early or a day late, and the report closes differently from the CRM every month, with nobody finding out why.
- 05Consent is treated as a default instead of a state. It works until someone revokes, and then the system keeps reporting, because nobody wrote the revocation path.
The proof
We built this on our own operation before selling it.
The whole path, from the ad to the CRM and back to Google, runs at Code and Soul. It is not an architecture drawing: it is what measures our own media.
It was built and tested end to end before it became an offering, and every stop was checked by reading what arrived on the other side, never the status code of the response. A test that only looks at the status proves something left, never that something arrived.
Tell us how your media is measured today.
A traceability diagnosis: where the path from click to contract is cut, and what it costs to reconnect each stop.
Engineering is not decoration. It is direction.