Affiliate Reporting: Keep an Outlier Exclusion Register

Document excluded campaign observations so media buyers can compare filtered results with complete performance data.

Make exclusions visible

Performance marketing media buying often produces observations that look unusual: a large order, a brief spend spike, or a campaign with very few clicks. Unusual does not automatically mean invalid. This proposed reporting method keeps exclusions visible so a cleaner chart does not quietly become a different business story.

Before filtering, write down the question the report should answer. A view of typical daily behavior and a reconciliation of all recorded revenue serve different purposes. Preserve an unfiltered baseline alongside any analytical view.

Record each decision

Maintain a register containing the observation identifier, period, metric, reason for exclusion, evidence, decision date, and reviewer. Avoid copying personal customer data into that register. Reference an access-controlled source record when supporting details are sensitive.

Separate confirmed data errors from valid exceptional events and unresolved anomalies. A duplicate import can warrant correction after verification; a genuine large order may need annotation rather than removal. An observation awaiting investigation should not silently disappear from an advertiser report.

Show the effect on the result

For a hypothetical example, suppose five campaigns record approved commissions of 40, 50, 60, 70, and 280 units of the same currency. The total is 500 and the mean is 100. Removing the largest observation leaves 220 and a mean of 55. Neither number is useful without explaining what population it describes.

When a filtered result appears in a table or chart, show the excluded count and the corresponding unfiltered figure. Keep currency, date range, approval status, and attribution scope consistent. If a denominator changes, recalculate the ratio from its underlying components rather than averaging displayed rates.

Use a review rule before budget decisions

Define exclusion rules before assessing a campaign whenever practical, and record changes to those rules. Require a second review when a discretionary exclusion would materially change a scale-or-stop decision. This is a governance suggestion, not a statistical guarantee.

For media buyers evaluating CPS offer scaling, the register creates a route back from a summary to the evidence. Review confirmed corrections separately from hypotheses, and retain enough history to explain why the current report differs from a previous version.