Affiliate Reporting: Prevent Overlapping Time Windows

A boundary event can appear in two reports

Performance marketing media buying depends on consistent comparisons. Two adjacent exports can each look reasonable while counting the same event at their shared boundary. This article proposes a reconciliation method for publishers, advertisers, and media buyers assembling recurring reports. Apply it only after confirming how the source system defines its timestamps and filters.

Write the interval rule in plain language

A convenient reporting convention includes the start instant and excludes the end instant. Under that convention, the first window contains events at or after its start and strictly before its end. The next window begins at that same end instant. This convention is often called a half-open interval. It is a proposed rule for your reporting layer, not a claim about how an advertising platform implements its export.

Check the source before changing filters

Identify whether the export uses click time, order time, approval time, or another event timestamp. Record the timezone, timestamp precision, and whether date filters represent calendar days or exact instants. A date-only export may not provide enough detail to reconstruct a boundary event. Preserve that limitation instead of fabricating a precise allocation.

Reconcile with a tiny example

Consider a hypothetical event recorded exactly at 10:00. A window from 09:00 inclusive to 10:00 exclusive omits it; a window from 10:00 inclusive to 11:00 exclusive includes it. If both exports include 10:00, the event can be duplicated. If both exclude it, it can be lost. Use a stable event identifier where available to check membership, rather than relying solely on totals that can conceal offsetting errors.

Separate overlap from later revisions

An event can legitimately change approval status after its original timestamp. Define whether a report represents events occurring in a period or the latest status of an earlier cohort. Keep extraction time alongside the event-time rule. Otherwise, a revised commission total may be mistaken for a boundary problem, leading a buyer to adjust bids for the wrong reason.

Build a repeatable handoff

Attach a short interval specification to each reporting dataset: event basis, timezone, precision, start rule, end rule, and revision policy. Compare the union of adjacent windows with a combined-period export when the source permits it, accounting for revisions between extraction times. Teams working on media buying can use this check before interpreting a change as campaign performance. A clean boundary validates counting logic; it does not by itself prove attribution accuracy or profitability.