This conversion tracking checklist is what I set up before a new account spends anything, and roughly two thirds of it happens before a single advert goes live. Tracking added afterwards leaves a hole you can never fill, because clicks that arrived before the measurement existed cannot be reconstructed.
Twelve items. The order is the useful part, because getting them out of sequence is how accounts end up with sophisticated infrastructure pointing at the wrong thing.
This is for you if you are launching an account, or inheriting one and want a structured way to check it.
Before anything technical
The first three items involve no code and they decide everything after.
One. Decide what counts as a conversion. The closest thing to money you can reliably record. Not a page view, not a click, not a scroll. If a form fill is genuinely the only option today, write down what would replace it later.
Two. Decide what does not count. Newsletter signups, brochure downloads and contact page visits are worth measuring and are not the thing you want bidding to chase. Naming them now prevents them creeping into the primary set.
Three. Agree which system the business will judge results on. Usually the CRM or the order system. Doing this at launch prevents the recurring argument where marketing and finance quote different figures and each assumes the other is wrong.
Those three take a conversation rather than a developer, and skipping them is why so many accounts optimise beautifully towards something nobody wanted.
The plumbing that everything depends on
Items four to six are the cheap technical work, and item four is the one people skip.
Four. Store the click identifier with every enquiry. A hidden field on the form capturing what arrives in the URL, saved alongside the record and preserved into your CRM. Fifteen minutes. Nothing connecting advertising to revenue is possible without it.
Five. Add the outcome columns. Qualified yes or no, became a customer yes or no. Even if nothing consumes them for months, data you did not collect cannot be backfilled.
Six. Fire the conversion event after validation, not on submission. If the event fires when somebody presses the button, junk is counted before it is filtered, and the algorithm learns from it.
That third one costs nothing and protects the expensive half of the spam problem. It is the highest return line on this conversion tracking checklist.
The conversion tracking checklist items for the platforms
Items seven to nine, once the foundations exist.
Seven. Set up the conversion actions and mark only the right ones primary. Bidding chases primary actions. Several measured, one or two primary, is the shape you want.
Eight. Check the count setting on each action. Count one per click for leads and calls. Count every occurrence is correct for ecommerce purchases and almost never for enquiries.
Nine. Confirm deduplication where two routes exist. On Meta, if a browser pixel and a server connection both report, your money event should show a deduplication status rather than two separate counts.
Those three are where the most common faults get created at setup, and each one takes minutes to verify against hours to diagnose later.
The conversion tracking checklist items that depend on your scale
Ten to twelve, and these are the ones a small account can legitimately defer.
Ten. Enhanced conversions or Advanced Matching, if the customer data is reachable. If the email is already on your confirmation page, this is close to free. If it needs development work, weigh it against what else that time buys.
Eleven. A server side connection, if somebody will own it. It recovers real events and it is software that has to keep working. Without a named owner it degrades quietly, which is worse than not having it.
Twelve. Outcome import back to the platforms. The single biggest improvement available in lead generation, and the one requiring the most cooperation. Items four and five exist so this becomes possible when you are ready.
Deferring ten to twelve deliberately is a reasonable decision. Deferring them because nobody thought about them is how an account arrives at year two with no way to answer which campaigns produced customers.
Verifying the conversion tracking checklist
One test covers most of it and it is worth doing properly rather than assuming.
Click one of your own live adverts. Submit a real enquiry with details you will recognise. Then follow it.
Does it appear in the ad platform as a conversion? Check your analytics next. Then confirm it reached your CRM with the click identifier attached, and that it appears once rather than twice.
If it stops anywhere, you have found your gap, and you found it in five minutes rather than in a quarterly review. If it appears twice, you have found duplication before it corrupted three months of optimisation.
I run this on every account I take over and it produces a surprise more often than not.
What to do with an inherited account
Run the same list as an audit and assume nothing.
Inherited accounts carry history nobody documented. A conversion action set up in 2022 by somebody who left. Perhaps a tag manager container alongside a plugin, both firing. Often a thank-you page reachable directly.
Verify each item yourself rather than accepting a description of it. The gap between how an account is described and how it behaves is where every inherited fault lives.
Then write down what you found. That document is worth more than the fixes, because it is what lets the next person check this without starting over.
When to run it again
After every site change, without exception.
Rebuilds cause most new faults, and they cause them invisibly. A form redesign that drops a hidden field leaves a form that works perfectly and a chain that is broken. Nothing about the visible result looks wrong.
Beyond that, quarterly is enough for a stable account. The after-changes rule matters far more than the calendar one.
Put it in the release process rather than in your own diary, so it survives you being busy that week.
Why this conversion tracking checklist front-loads the cheap work
Notice the shape of the twelve. The items costing almost nothing come first, and the expensive infrastructure comes last.
That ordering is deliberate and it runs against how these projects usually get scoped. The instinct is to start with the impressive part, because a server side container sounds like a serious answer and two columns in a spreadsheet does not.
The arithmetic disagrees. Storing a click identifier costs fifteen minutes and makes outcome import possible later. Building infrastructure before deciding what you count produces a reliable pipeline delivering the wrong signal.
There is published evidence that the sophistication of a method is not what determines whether it gets close to the truth. Gordon, Zettelmeyer, Bhargava and Chapsky compared observational attribution against randomised experiments across 15 Facebook experiments and 500 million user-experiment observations, and found estimates off by a factor of three in half the studies, using methods considerably more advanced than anything on this list.
So do the cheap things properly first. They are what the expensive things stand on, and they are the ones nobody regrets.
What to change this week
Three steps.
Run the verification test on your existing account, whatever its age. Click your own advert, submit an enquiry, follow it through every system.
Then check items four and five, the click identifier and the outcome columns, because those two are missing on most accounts and enable everything else.
Then add a tracking check to whatever process governs site releases, so this stops depending on somebody remembering.
The conversion tracking audit is the fuller version for an account already running, what conversion numbers actually measure is the reasoning behind the whole list, sending outcomes back is item twelve in detail, and the twelve checks I run on any account covers the campaign side once the measurement is sound. Getting this right at launch is where my Meta Ads work starts on every new account. To have a new setup checked before it spends, book a teardown.