Offline conversions: how to feed real deals back into your advertising
You return them by uploading back to the account the outcome of each enquiry: which one became a deal, when and for how much. An offline conversion is a real result outside the website, linked back to the click that brought it. With service businesses the platforms see only the beginning: the submitted form or the call. What happened afterwards stays with you, and bidding learns from an incomplete picture. For the return to work, you need a click identifier recorded in the system where you manage your deals, and a regular upload of statuses. In Google the upload window is up to 90 days after the click, and the benchmark for stable learning is about 30 conversions per 30 days. The aim is for bidding to learn from customers, not from filled-in fields.
My name is Hristiyan Panov — a performance marketing consultant with over 5 years’ experience and over €2 000 000 managed advertising budget for over 90 companies on Google and Meta. Below is how the mechanism works, what has to be in place and when it is worth it.
Why does optimising for enquiries mislead?
Because the algorithm optimises exactly what you give it. Tell it “I want forms” and it will find you the people who fill in forms most easily. Some of them are not your customers and never will be.
The effect builds up quietly: the cost per enquiry falls, the number grows, the report looks better and better, while sales stay flat. This is one of the most common scenarios in B2B accounts.
How does feeding the data back work?
When the form is filled in, an identifier is recorded — for Google this is the GCLID (or an anonymous identifier with Enhanced Conversions for leads), for Meta — an event identifier and contact details. It is stored together with the enquiry.
The identifier arrives in the page address at the moment of the click and lives only for the duration of the visit. That is why it is captured at the moment of submission: a hidden field in the form passes it along with the name, the phone and the time, and the row goes to the system where you manage your deals.
When the deal closes, you send the same identifier back together with the value. The platform links the revenue to the specific ad interaction from weeks earlier.
The upload is a file with several columns: identifier, conversion name, closing date and time, value and currency. The time of the conversion must be after the time of the click, and the conversion action is created beforehand as imported, otherwise the rows are rejected. In Google the upload window is up to 90 days after the click.
What has to be in place?
- The identifier is recorded in a hidden field of the form and reaches the system where you manage your deals.
- A place where the status is kept — a CRM, a spreadsheet, whatever you use, as long as it is consistent.
- Enough volume. Below a certain number of closed deals per month, the algorithm has nothing to learn from.
- Discipline. If the statuses are not updated regularly, you send back noise instead of signal.
The last point has sunk more implementations than the technology has.
What changes afterwards?
Bidding starts to chase the queries that historically lead to deals and pulls away from those that bring only forms. The result is usually fewer enquiries at a higher price — and more revenue.
In metrics the movement looks like this: the number of enquiries falls, the cost per enquiry goes up, and the revenue per enquiry goes up by more. If you look only at the first two numbers, you will decide the account has got worse.
That is why the change has to be explained in advance. The report after it looks worse on every metric you have followed so far.
In practice this means three things. The headline metric in the report becomes closed deals and revenue, and enquiries drop one level lower. Comparisons are made over a window commensurate with the sales cycle, not week against week. And changing the conversion action puts bidding back into learning — during this period the data is unstable and is not grounds for further changes.
When is it worth it and when not?
It is worth it with a long sales cycle, a big difference in value between deals, and sufficient volume. It is not worth it with a small number of enquiries per month or when every deal is roughly the same value — there the simpler solution is to optimise for a qualified enquiry.
The check is short and is done before the implementation, not after it:
- Volume. The benchmark for stable learning is about 30 conversions per 30 days. If closed deals are noticeably below that, the returned signal is too sparse to change bidding.
- Spread of values. Compare the largest and the smallest closed deal over recent months. If they are of the same order of magnitude, the gradation adds little.
- Length of the cycle. If deals close within days, the standard website conversion already captures the result.
- Coverage. If most enquiries come by phone without call tracking, the identifier is missing for a large part of the traffic — this is solved first.
What is a simpler first option?
Values by stage. Instead of waiting for a closed deal, assign a different value to “enquiry”, “call held” and “quote sent”. The algorithm gets a gradation from the very first days, and you save yourself half the integration.
The values are not arbitrary. They are ordered by the probability that the stage reaches a deal: the further along the customer is, the higher the value. If you have kept statuses before, this probability can be read straight from the history — what share of enquiries reach a call and what share of calls reach a quote sent.
The technical part is small. Each stage is a separate conversion action, all of them are returned with the same identifier, and only one stays primary for bidding, so that the same customer is not counted several times. The currency and the scale are the same everywhere.
This is also the sensible start, before you take on the full return of revenue. When real revenue starts to flow, the stage values are replaced, not added on top of it.
FAQ — Frequently asked questions about offline conversions
How far back can a conversion be returned?
The window is limited and depends on the platform. With a very long sales cycle some of the deals drop out — which is why the intermediate stages are used as well.
Do I need a CRM?
Not necessarily. It also works with a spreadsheet, as long as the identifier and the status are kept consistently.
What about calls?
Call tracking records its own identifier on the same principle. For businesses where most enquiries come by phone, this is a mandatory part.
How does it fit with website tracking?
It complements it. See also Enhanced Conversions — the two work together, they do not replace each other.
How long until there is an effect?
Until enough returned deals accumulate for the model to change — usually several weeks to months, depending on volume.
Conclusion
Advertising platforms optimise towards the goal you give them. If that goal is a “form”, you will get forms. Offline conversions are the way to change the goal to “customer” — and that is the change that most often unlocks growth for service businesses.
If you are getting enquiries but not deals, start with an independent review.
Request a quote for a digital audit →
Related articles: enquiry form · CAC and LTV · Enhanced Conversions
