Most conversion tracking problems do not look like problems. The account reports conversions, the numbers are plausible, and the reporting looks healthy. That is precisely why the fault survives — it produces confident decisions built on a measurement nobody has checked.
And the decisions are not small. Automated bidding optimises toward whatever conversions you supply. Give it the wrong signal and it will pursue that signal efficiently, at scale, with your budget.
The four faults, in order of frequency
1. Double counting
The same action recorded twice — by a thank-you page and a form-submit event, by two tag implementations left in place after a migration, or by a tag firing on a page that reloads.
It inflates conversions, deflates cost per conversion, and makes every campaign look better than it is. Because the ratio between campaigns is often preserved, the account can look internally consistent while every absolute number is wrong.
How to find it: submit one test conversion and confirm exactly one is recorded. Not "roughly one".
2. The counting setting
Google Ads offers Every or One conversion per interaction. The choice should follow the business model, not the default.
| Business | Setting | Why |
|---|---|---|
| Lead generation | One | One person enquiring three times is one lead |
| Ecommerce | Every | Three purchases are three purchases |
| Subscription signup | One | The second signup is usually a duplicate, not value |
Lead-generation accounts left on Every systematically over-report, and the over-reporting is concentrated in whichever campaigns attract repeat form-fillers — which is exactly the audience you would prefer to bid down.
3. Measuring the step instead of the outcome
A conversion should be something the business values. Common substitutes that are not:
- Page views on a contact page — arriving is not enquiring
- Scroll depth or time on page — engagement, not intent
- Clicks on a phone number without confirming the call connected
- Newsletter signups counted alongside sales enquiries at equal weight
The last is worth dwelling on. If a newsletter signup and a qualified enquiry are both "a conversion", the system will find you the cheaper one — and it will be very good at it.
4. Consent handling
Where consent is required, a share of users will decline, and their conversions are not recorded. That is correct behaviour, and it means your reported numbers are a subset of reality.
The problem is not the gap — it is not knowing the gap exists, then comparing periods across a consent implementation change and concluding performance dropped.
Because every downstream decision inherits whatever tracking reports, verifying it is the first stage of
the account work that starts with tracking rather than bids.Verifying properly
Tag-assistant tools tell you a tag fired. They do not tell you a conversion was recorded correctly, deduplicated, or attributed. Those are different claims and only one of them matters.
The verification that counts:
- Submit a genuine conversion — real device, real form, ideally mobile data rather than office wifi
- Wait for the interface to update
- Confirm exactly one conversion recorded
- Confirm the value is right, if you pass one
- Confirm it is attributed to the campaign you clicked from
- Repeat for every distinct conversion path — form, phone, chat, checkout
Do this after any change to the site, the tag manager, the consent tool, or the forms. Tracking breaks silently and most often during work that had nothing to do with tracking.
Why Ads and GA4 disagree
They will disagree. That is expected, and understanding the direction matters more than closing the gap.
| Google Ads | GA4 | |
|---|---|---|
| Attributed to | The ad interaction | The session, by its own model |
| Reported against | Date of the click | Date of the conversion |
| Scope | Ad-driven only | All traffic |
A conversion from a click on the 1st that converts on the 5th appears on the 1st in Ads and the 5th in GA4. Over a fixed window that alone produces a discrepancy.
Pick one as the decision source. For bidding decisions, use the platform doing the bidding — Google Ads optimises against its own conversion data, so that is the number governing behaviour. Use GA4 for cross-channel understanding.
Reconciling them exactly is not a solvable problem and chasing it wastes time.
The change that matters most
Import the outcome, not the enquiry.
Optimising toward form submissions teaches the system to find people who submit forms. That audience overlaps with buyers, but it is not the same audience — and the difference compounds as the system gets better at its objective.
Offline conversion import sends the real result back: which enquiries became qualified leads, which became customers, and what they were worth. The system then optimises toward customers.
What it requires:
- Capturing the click identifier when the enquiry arrives, and storing it against the record
- A CRM or spreadsheet where outcomes are recorded honestly
- A regular upload — automated where possible
That is a genuine integration and it is the single highest-return tracking change available to most lead-generation accounts. It moves the optimisation target from a proxy to the actual objective.
The prerequisite is a CRM that reliably captures every enquiry with its source. If that is not true today, the CRM automation checklist covers capture and hygiene before anything downstream is worth building.
A short checklist
- Every conversion action reflects real business value
- One primary action per meaningful outcome
- No double counting — verified by a real test submission
- Counting setting matches the business model
- Conversion window reflects a realistic decision period
- Attribution model chosen deliberately
- Consent handling understood, and its effect on reported numbers known
- Offline outcomes imported where the real result happens after the form
- Verified on a real device after every site or tag change
The full account review, ordered by dependency, is in the Google Ads audit checklist — where tracking is stage one for exactly the reasons above.
Want this handled properly?
You now know what the work involves. If you would rather not do it yourself, that is what we do.