Postback or Pixel: Why Two Attribution Reports Disagree by a Fifth

Each mechanism sees a different slice of the conversion funnel, and the discrepancies stack up fast.

How Pixels and Postbacks See Different Events

A pixel is a snippet of JavaScript that executes in the user's browser. It fires when the conversion page loads—if the user loads the page. Network interruptions, ad blockers, or a user who closes the tab before the pixel loads can all prevent the fire from registering. The merchant's backend, by contrast, has already processed the order. The postback originates from there, traveling server-to-server to the tracking platform.

Trackier's documentation notes that postbacks must include correct click ID mapping to record conversions at all. Without this identifier, the system cannot attribute the sale to the originating affiliate. Trackier also flags that incorrect or missing click IDs and expired attribution periods cause postback recording errors. The pixel, meanwhile, reports what the browser saw. These are fundamentally different data sources.

Six Verified Causes of Divergence

Gaps between systems emerge from several identifiable sources, each pushing numbers in a predictable direction.

Deduplication mismatches run in both directions. An advertiser may deduplicate conversions using an order ID while a platform deduplicates on click ID. If the same user converts twice—perhaps a purchase and a subscription upgrade—one system may count two events where the other counts one. Postbacks can retry after network failures, and without stable deduplication keys, these retries inflate totals. Track360 documentation emphasizes that repeated deliveries of the same conversion must be treated as a single event using a stable key.

Timing and date boundaries split totals when systems credit conversions to different dates. A user clicks on March 31st and converts at 11:55 PM Pacific Time. If one system runs on UTC and the other on Eastern Time, the conversion may fall into April 1st for one report and March 31st for the other. Cometly's guidance on conversion tracking discrepancies identifies timing mismatches as a direct cause of attribution differences.

Attribution window mismatches produce different credited totals even when raw click and conversion data are identical. A 30-day window credits conversions that a 7-day window rejects. Track360 defines the attribution window as the period after a click during which a conversion still counts toward that affiliate. TrackNow confirms that different windows produce divergent credited totals from identical underlying data.

Event definition gaps occur when "conversion" means different things. The pixel fires on checkout page load. The postback fires on payment confirmation. A user who loads the checkout but abandons payment triggers the pixel, not the postback. Outbrain DSP's documentation notes that settings can treat received events as already-attributed conversions—a configuration that changes what gets counted.

Click ID failures strand conversions in the wrong bucket. When postbacks arrive without valid click IDs, they may be discarded entirely or attributed to direct traffic rather than the affiliate source. This undercounts affiliate-reported conversions against internal totals.

Retry and delivery failures affect postbacks uniquely. Postbacks can fail in transit and require retry logic. Without retries, conversions disappear from the affiliate report entirely. With retries but without deduplication, they may double-count.

Windows, Dates, and Configuration Alignment

The attribution window is typically the largest lever in any discrepancy. Extending from 7 to 30 days can capture significantly more conversions, depending on your sales cycle. But window mismatches are invisible in aggregate totals. You must compare configurations directly.

Offer18's guidance on tracking discrepancies specifies four fields to check: click window, view window, attribution model, and deduplication key. Microsoft Advertising recommends verbose logs of pixel fires to investigate gaps with server-side implementations. The logs reveal what the browser attempted versus what the server received.

Date handling deserves explicit attention. Systems may timestamp on event time, processing time, or click time. A conversion clicked January 1st and processed January 3rd appears in different periods depending on which timestamp governs. Cometly flags this timing mismatch as a frequent source of confusion.

What Postback Solves, and Where It Fails

Server-to-server transmission fixes the pixel's visibility problems: ad blockers, network drops, and incomplete page loads. But it introduces its own failure modes. The postback depends entirely on click ID integrity passed through the funnel. If the affiliate link sets a click ID that doesn't survive to checkout—stripped by URL parameters, lost in redirects, or overwritten by another tracking system—the postback arrives unattributed.

Deduplication keys matter enormously. The advertiser using order ID and the platform using click ID will disagree on what constitutes a duplicate. A single transaction with two line items might trigger two postbacks. Without a shared key, one system sees one conversion, the other sees two.

Postbacks also cannot resolve attribution conflicts upstream. If two affiliates both claim credit for the same click—through partner-side click deduplication failures or competing view-through claims—the postback carries whatever click ID the backend last recorded. It does not adjudicate which affiliate earned it.

Stopping the Argument: A Reconciliation Protocol

Disputes over totals resolve only when configurations are aligned. The practical approach: select one conversion event as the agreed source of truth—payment confirmation, not checkout load or subscription creation. State the attribution window explicitly in both systems. Review deduplication keys to ensure they match.

Microsoft Advertising's guidance on server-side pixels emphasizes verbose logging for investigation. Outbrain DSP's documentation highlights configuration settings that treat received events as pre-attributed. These are checkable fields, not opinions.

When gaps persist after alignment, the remaining difference is measurable noise: pixel blockers the postback captures, postback retries the pixel avoids, timing edges that fall on different dates. Document the monthly percentage gap. If it holds steady, you have a calibrated adjustment. If it widens, you have a drift in configuration to chase.

The argument over whose numbers are "right" misses the point. The systems are built to count different things. The work is making them count the same thing, then understanding why they still differ.