Facebook and Instagram advertising account audit

For the most part, no one makes a mistake on purpose — several effects converge. First and most important: double accounting, when browser and server events are not deduplicated. Second: post-impression conversions that count without a click. Third: different attribution windows. We divide these reasons and show how much each accounts for - after that the conversation becomes substantive.

Deduplication is when Meta understands that the browser event from the pixel and the server event from the Conversions API describe the same purchase, and counts it once. It is maintained on two conditions: both copies have the same event name and the same unique event ID. You can check for yourself: in the event manager, open the desired event and see if the system sees duplicates and in which part. If the server events come without an identifier, they are simply added to the browser events, and the cabinet shows a duplicate picture.

The most common reason is creative burnout with an unchanged set of materials. The audience sees the same thing, reacts worse, the cost of showing increases. The second reason is the intersection of audiences: campaigns compete for the same people and raise each other's prices. Both defects are visible in the layout, and both do not appear to be broken.

How many pixels are there, what events are triggered, whether the amount and currency are transferred, whether there is a server channel and whether the events are deduplicated. This is the technical part, and without it the audit of the cabinet will be incomplete: most of the problems with numbers are born here, and not in the campaign settings. In our own sample, we wrote out the composition of counters by name on 13 sites out of 33 — the pixel stood on six, and on two of them there was nothing else at all nearby.

Without access, you can check only what is visible on the site: the presence of a pixel, the composition of events, duplication. It's useful, but it's half-baked: costs, audiences, and campaign breakdowns remain invisible. We take on such a limited format, but we clearly indicate which conclusions are drawn blindly.

A minimum of a month of expenses or several hundred results. On the weekly data, any conclusions will be about the settings, and not about the efficiency - the fluctuations in this channel are large. If the campaigns have been launched recently, we would honestly advise you to come back later instead of doing work for work's sake.

Our recommendation is to do it in two steps, and the first step is free. First, switch the attribution window to seven days after the click, excluding views, and look at the same dates again: some of the difference disappears without any edits. Then take a month, not a day. The office puts the conversion on the date of interaction with the ad, your system - on the date of the order, and in short segments these two logics do not coincide in principle. What will remain after these two actions, it is already worth sorting out technically.

We will write what the settings cost and how much it costs in money. There are no personal evaluations in the document: some of the defects arise due to changes on the site side or due to decisions made in a different situation. A list of corrections is more useful than a sentence and can be given to the same person to execute.

Go through the points in the order in which they are listed: the first two or three usually have the most impact and most often relate not to campaigns, but to data. If our participation is required, we name the cost for specific items. It often turns out that the most important thing is to fix the deduplication, and this is a multi-day job.