The Conversion Window You Set in Your Campaign Is Not the Window Your Attribution Partner Counts

When a media buyer sets a 7-day click conversion window, the reasonable expectation is that the attribution report will count conversions that happened within 7 days of a click. That expectation is almost always wrong, and the gap is not a bug in any one system. It is a structural feature of how attribution vendors build their reporting layers on top of campaign configuration.

What You Set and What Gets Counted Are Two Different Things

Conversion windows in a campaign platform control which impression and click events are eligible to be paired with a downstream conversion event. They do not control how the attribution vendor ingests, deduplicates, or timestamps those events before reporting.

Most attribution vendors receive event data through a pipeline that introduces latency. A click that happened on day 6 of your 7-day window may not arrive in the attribution system until day 8 or 9, depending on batch cadence, API polling frequency, and how each platform exports its event stream. When the attribution vendor processes that click, it may apply its own lookback window from the moment of data ingestion rather than from the moment the click occurred. The result is that a click well within your configured window gets counted as outside it, or dropped entirely.

The inverse also happens. Some attribution vendors apply a rolling lookback from the conversion event backward, not from the click event forward. If your platform is counting forward from the interaction and your attribution vendor is counting backward from the outcome, the two windows are not measuring the same interval even when the numbers match on paper.

Why This Rarely Gets Caught

The discrepancy is easy to miss because both systems report with authority. Your campaign platform shows a conversion count. Your attribution vendor shows a conversion count. Neither report flags that the underlying windows are misaligned. Buyers typically only notice the gap when they try to reconcile the two numbers and find them consistently off by a percentage that does not explain itself through standard deduplication.

The standard explanation offered by vendor support teams is deduplication logic, and that explanation is usually partially true. Deduplication does account for some variance. But deduplication does not explain directional bias. If your attribution vendor consistently reports fewer conversions than your platform across multiple campaigns and multiple time periods, window misalignment is a more likely explanation than deduplication alone.

The Timestamp Problem Underneath It All

Both systems depend on timestamps to define windows, and timestamps are not standardized across the ad tech stack. Some platforms record the timestamp of the event in the user's local time zone. Others record in UTC. Some record when the event was fired client-side. Others record when the event was received server-side. A conversion that happens at 11:58 PM in one time zone is a next-day event in another. Across a 7-day window, a consistent one-hour timestamp offset accumulates into a meaningful population of events that fall on opposite sides of the window boundary depending on which system's clock you trust.

This is not a hypothetical edge case. Any campaign running across geographically distributed inventory is generating events whose timestamps vary by how each publisher's ad server, the buyer's tracking pixel, and the attribution vendor's ingestion layer each record time. The window you set is only as precise as the least consistent timestamp in that chain.

What Practical Alignment Looks Like

Buyers who want the window they set to approximate the window that gets counted have a few practical options, none of them perfect.

First, ask your attribution vendor to document explicitly how it defines the start of a lookback window. Is it from the event timestamp as recorded by the source platform? From the timestamp as received by the attribution system? From the timestamp of the conversion looking backward? That single answer narrows the range of possible misalignment sources significantly.

Second, request a sample-level event log from both your campaign platform and your attribution vendor for the same short time period, such as a single week. Compare event timestamps at the individual event level, not at the aggregate report level. Aggregate reports obscure timestamp variance. Individual event logs expose it. If your team does not have access to event-level logs from both systems, that access gap is itself a measurement risk worth documenting.

Third, set your campaign window longer than your reporting window. If you intend to report on 7-day conversions, set your campaign window to 14 days. The buffer absorbs pipeline latency without changing your reporting logic. You will attribute more conversions than your window strictly requires, but you will undercount fewer. For most buyers, overcounting at the margin is less damaging than systematically undercounting because pipeline lag pushed events outside a tight window.

Fourth, standardize on UTC across every platform you use for event recording where configuration allows it. Local time zone timestamps are a source of variance that serves no measurement purpose for most media buying use cases.

What This Means for Campaign Optimization

Conversion window misalignment has a downstream effect on optimization that is easy to overlook. Many DSPs use the conversion signals reported through attribution integrations to train their bidding algorithms. If the attribution vendor is systematically undercounting conversions due to window misalignment, the optimization signal the DSP receives is noisier than it should be. The algorithm is not being told that fewer conversions happened. It is being told that conversions happened, but with gaps and delays that corrupt the signal-to-impression relationship the model is trying to learn.

This means campaigns that look like they are optimizing normally may be learning from a degraded signal for their entire flight. The campaign can still perform adequately, but the bidding model is operating with less accurate feedback than the buyer believes it has.

A Reasonable Starting Point

No buyer can fully control every timestamp and pipeline latency across a multi-vendor measurement stack. The goal is not perfect window alignment. The goal is knowing how misaligned your windows actually are so that the variance you observe in reconciliation reports has an explanation you can act on rather than a variance you absorb as unexplained noise.

Starting that conversation with your attribution vendor by asking for their specific definition of a lookback window, before your next campaign launches, is the lowest-cost diagnostic step available. The answer will tell you more about your measurement stack than most post-campaign audits do.