Conversion Tracking: The Setup Most Get Wrong
Marketing measured badly is worse than marketing not measured at all, because it produces confident decisions in the wrong direction.
Almost every account we audit has conversion tracking installed.
Rather fewer have tracking that reports what anyone thinks it reports.
And that matters more than it sounds, because the numbers are not just describing the past. They are instructions for the future — the platform uses them to decide who to show your ads to next.
Tracking is an instruction, not a report
This is the part that changes how carefully you set it up.
When you tell Google or Meta that an action is a conversion, you are not only asking to see it in a dashboard. You are telling the delivery system: find me more people who do this.
Define the conversion as "form submitted" and it will get very good at finding people who submit forms. That is a different population from people who buy — and the gap between them is exactly the problem described in Meta lead ads and cost per lead.
The three ways it usually breaks
| Failure | What it looks like | Consequence |
|---|---|---|
| Double counting | Platform revenue meaningfully exceeds your back office | Every ROAS and CPA figure is flattered. Budget goes to the wrong campaigns. |
| Wrong action counted | Lots of conversions, no matching revenue | The algorithm optimises toward people who do the cheap thing |
| Never verified | Everything looks fine and has been broken for months | Decisions made on a frozen or partial picture |
| Wrong or missing values | ROAS targets behave unpredictably | Value-based bidding cannot work at all |
| Thank-you page refreshes | Conversions slightly above reality, consistently | Small, persistent inflation nobody investigates |
Double counting is the most common and the most damaging, because it is invisible unless you reconcile against your own records. The usual cause is the same action reported twice — a tag on the page plus an imported conversion from Analytics — with nothing telling the platform they are the same event.
How to set it up properly
- Decide what actually counts. One primary action closest to money. Write it down before anyone opens a tag manager, because this is a business decision, not a technical one.
- Count few things. Every additional conversion action is something the platform might optimise toward. A cluttered list makes it easier to optimise for the wrong outcome.
- Choose one source per action. Either the platform tag or the Analytics import — not both for the same event unless they are properly deduplicated.
- Send real values. Actual order value, not a flat placeholder. Without this, ROAS targeting is meaningless.
- Handle the edge cases. Refreshed thank-you pages, back-button revisits, test orders from your own team. All of them inflate quietly.
- Verify with a real transaction. Buy something or submit a real enquiry, then follow it through the reporting. This is the step that separates working setups from installed ones.
Why Google Ads and GA4 never agree
They are counting different things and this is expected, not broken.
- Attribution timing. Google Ads credits a conversion to the day of the click. GA4 credits it to the day it happened. A conversion from a click ten days ago lands in different columns.
- Attribution model. The two apply credit across touchpoints differently by default.
- What counts as a session. GA4 and the ad platforms disagree on when a visit begins and ends.
- Consent and blocking. Browser restrictions affect each differently.
Pick one as your source of truth for decisions, and use the other as a sanity check. Trying to reconcile them to the unit is a waste of everyone's afternoon.
What good tracking unlocks
Everything downstream depends on it, which is why it comes first in every engagement we run.
Smart bidding. Target CPA and target ROAS optimise against this signal. Feed them a wrong or unstable signal and they will pursue the wrong outcome confidently — see bidding strategies.
Automated campaign types. Performance Max is an optimisation engine with no manual controls to fall back on. Without verified tracking it is spending on guesswork.
Audiences. Every remarketing list and every lookalike is built from event data. Broken tracking produces thin audiences, usually misdiagnosed as a targeting problem — the Meta-specific version is in Meta Pixel and Conversions API.
Honest reporting. The ability to say what marketing earned, in a number your finance team recognises. That is what CPA is for.
The quarterly check
- Complete a real conversion and follow it end to end.
- Confirm it appears once, not twice.
- Check the value is correct and not zero.
- Reconcile a month of platform conversions against your own records.
- Review the conversion action list and remove anything that has crept in.
- Re-verify after any site change — theme updates, checkout changes and plugin updates all break tracking silently.
Frequently asked questions
What is conversion tracking?
It is how an ad platform learns what happened after someone clicked. You define an action that matters — a purchase, a qualified enquiry — and report it back. That signal is what campaigns optimise against, so tracking is not just reporting; it decides who the platform goes looking for next.
Why do Google Ads and GA4 show different conversion numbers?
Because they count differently. Google Ads credits a conversion to the day of the click, GA4 credits it to the day it happened, and their attribution models differ. Some difference is normal and expected. A large or growing gap usually means something is genuinely misconfigured.
Why are my conversions being double counted?
Most often because the same action is being reported twice — a tag on the page plus an imported conversion from Analytics, both counting the same purchase. The other common cause is a thank-you page that can be refreshed or revisited, firing the tag each time. Both inflate everything downstream.
What should I count as a conversion?
The action closest to money that happens often enough to learn from. A purchase for e-commerce. A qualified enquiry for services — not a raw form fill, which teaches the platform to find form fillers rather than customers. Count few things, and count the right ones.
How do I verify conversion tracking is working?
Complete a real conversion yourself and follow it through. Confirm it appears, appears once, carries the right value, and is attributed to the right source. Checking that a tag is installed is not the same as checking that it reports correctly, and the difference is where most broken setups hide.
We fix measurement before we spend a rupee
Tracking verified with real transactions, deduplicated across sources, and reconciled against your own records every month.