Tracking Funnel

Conversion Tracking in Media Buying: Where Data Gets Lost Between the Click and the Postback

Discrepancies between the tracker and the affiliate network are one of the most common problems in traffic arbitrage. Yet few teams actually break down where exactly the data gets lost.

It usually gets lost at very specific points. Most of these issues do not require complex solutions – you just need to know where to look.

What the Full Data Transfer Chain Looks Like

A basic conversion tracking flow looks like this:

Something can fail at every step. Let’s go through the chain.

Redirect: Speed and Fault Tolerance

The delay between the click and the landing page load affects conversion rate. Even 500–800 ms can already be noticeable on mobile traffic, especially in geos with slower internet connections.

But there is an even more important issue: if the server goes down, traffic will not reach the landing page until someone notices and fixes it. With properly configured technical infrastructure, domains automatically switch to a backup server, and the team receives a notification instead of discovering the incident an hour later through zeros in the stats.

Domain monitoring is a separate topic. An expired domain, complaint, blacklist issue – all of this should be flagged in advance, not discovered when the traffic has already been wasted.

Landing Page: Pixel, CAPI and Lost Attribution

A conversion on the landing page is usually captured in two ways: a browser pixel and server-side CAPI. The pixel can be blocked by browser extensions and privacy policies – iOS, Safari ITP, and most desktop ad blockers. A pixel alone, without CAPI, does not cover a significant share of conversions.

CAPI works through the server, so it does not depend on browser restrictions. But it needs to be configured correctly. If the event_id, email, or other data is not passed properly, Meta cannot reliably match the event to the user, and optimization algorithms perform worse.

Another common issue is manually adding code to each landing page. When pixel, CAPI, and other scripts are inserted manually every time a landing is uploaded, something eventually gets missed or added with an error. Global macros that are automatically inserted into every landing page uploaded to the system solve this problem — this is exactly how it works in AIO.

Landing Page A/B Test: When the Result Cannot Be Trusted

The split test is running, the data has been collected, but the result still cannot be interpreted. This happens more often than it seems.

Uneven traffic distribution. One variant received three times more impressions, so the data is not comparable. This usually happens when weights are distributed without accounting for caching or traffic source specifics.

No link to a conversion metric. You can see that one variant received more button clicks. But which one brought more leads is unclear, because the conversion event is not tied to a specific split variant.

The test was stopped too early. There is no statistical significance yet, but visually one variant looks better. The decision is made before enough data has been collected.

A properly configured split test requires clear weight distribution, one target metric for the entire test, automatic winner selection based on conversion, and display rules by geo or device. Heatmaps should also be used to understand user behavior, not just the final numbers.

Postback: Where Discrepancies with the Affiliate Network Come From

The tracker shows 100 conversions, while the affiliate network shows 80. Several scenarios are especially common:

• The postback was not sent because of a server failure at the moment of conversion. If postbacks are not logged with detailed statuses, it is impossible to reconstruct the picture afterward.
• Duplicate conversions. Pixel and CAPI both sent the event, but deduplication failed because the event_id was missing or mismatched – and Meta counted the conversion twice.
• Time zones. The conversion ended up in a different reporting day because the tracker and the affiliate network use different time zones.
• Incomplete postback parameters. The affiliate network could not match the event to the click, so the conversion was not counted.

How This Works in AIO

The issues described above are only part of what AIO helps media buying teams solve. If you want to see how it works in practice, book a demo.

You can also read more about how infrastructure problems affect overall team efficiency here.