Affiliate Reporting: Read Median and Mean Conversion Times Together

A single average can make an affiliate campaign's conversion timing appear more uniform than the underlying records. Performance marketing media buying teams can get a clearer description by reporting both the median and the arithmetic mean, with the event definition and observation window beside them.

This is a proposed reporting practice, not a benchmark for any network or an instruction to change campaign budgets. It applies only after the team has a valid set of elapsed times measured on a consistent basis.

Define the interval before calculating it

State exactly which events start and end the clock. Click to order, order to approval, and approval to payment are different intervals. Do not combine them into a single column labeled conversion time. Document the time unit, timestamp conventions and how multiple eligible clicks are resolved by the reporting system.

Also identify which records are included. A file of completed orders describes observed converters. It does not measure how long every visitor takes to decide, because visitors who have not converted have no completed interval in that file.

Use an example that exposes the difference

Consider five hypothetical, valid click-to-order intervals: 1, 2, 2, 3 and 42 hours. Their total is 50 hours, so the arithmetic mean is 10 hours. The median is the middle value in the ordered list: 2 hours. Both numbers are correct.

The mean reflects all five durations and is pulled upward by the 42-hour observation. The median describes the center of this small ordered set. Reporting only 10 hours could hide how closely four observations cluster; reporting only 2 hours could hide the long interval. Neither number alone describes the entire distribution.

Check unusual records before explaining them

An unusually long duration can be valid or can reflect a data problem. Review the event pairing, timestamp parsing and record status before assigning a behavioral explanation. Do not delete the record simply because it changes the average. If a correction is justified, preserve the original value and the reason for correction in the audit trail.

With a small sample, show the count and, where appropriate, the ordered durations. For larger samples, additional distribution summaries can help, but the calculation method and exclusions should remain explicit. More statistics do not repair inconsistent event definitions.

Avoid a misleading campaign comparison

Two campaigns may have different observed timing because they contain different products, markets or observation periods. A recently started cohort also has less opportunity to contain long completed intervals. Compare like event definitions and sufficiently observed cohorts before interpreting a lower median as a campaign improvement.

Do not use the median as a cutoff after which all remaining conversions are assumed lost. Likewise, a longer mean does not by itself establish weak traffic quality. Timing is descriptive evidence that needs the surrounding acquisition and order context.

Make the report reproducible

Include the event pair, included record count, date range, extraction time, mean, median and unresolved anomalies. Keep ecommerce CPS orders separate from virtual-product events when their definitions differ. BlueFriday's media buyer information provides the broader audience context; the report itself should let another analyst reproduce every timing figure before it informs a campaign decision.