Affiliate Reporting: Do Not Add Overlapping Cumulative Exports

Two export files can contain different totals while still describing much of the same activity. A Monday month-to-date report and a Tuesday month-to-date report are snapshots of overlapping periods. Adding them together treats the earlier activity as if it happened twice.

For teams doing performance marketing media buying, the first question when receiving a new file should be what it represents. This proposed reporting workflow helps affiliates and advertisers distinguish a replacement snapshot from an incremental delivery of new records. It does not assume that every reporting system uses the same export semantics.

Label the window and the extraction time

Keep the reporting start, reporting end, time zone, extraction timestamp, filters, currency, and metric definition with each file. The reporting window describes the activity being measured. The extraction timestamp describes when the system produced that view. Those dates can differ when approvals or corrections arrive later.

Use a filename for convenience, but do not rely on it as the sole definition. A file called daily-report may contain an entire month. Confirm the source interface or documentation and retain the original export unchanged.

Work through the overlap before calculating a total

In a hypothetical example, a report covering September 1 through 10 contains 120 tracked orders. A later report covering September 1 through 11 contains 135 tracked orders. Adding the files gives 255, which counts the shared period twice. If the later file is a complete replacement snapshot, its 135 orders are the appropriate snapshot total for that window.

The difference of 15 is only the net change between those snapshots. It is not automatically proof of 15 orders occurring on September 11. A later export could also contain late records, removed duplicates, or corrections to earlier dates. Confirm the data semantics before describing the difference as new daily activity.

Choose the combination rule explicitly

• Complete snapshots: Use one suitable snapshot for the selected window, preserving earlier versions for comparison.
• Disjoint periods: Combine only after confirming that boundaries, filters, and metric definitions align.
• Incremental records: Apply the source's documented record and revision rules, using stable identifiers where available.

A stable identifier helps detect repeated records, but it does not by itself resolve whether a later row replaces an earlier state or represents a separate adjustment. For example, an approved-status update might require replacing a pending record rather than adding another order. Make that rule part of the import design.

Keep revisions visible

Create a small reconciliation note with the prior snapshot total, current total, net change, and known explanation. Where the reason is not available, label it unresolved. Do not quietly assign all revisions to the latest day merely to make a daily chart balance.

Before using the result for campaign decisions, check whether the same issue affects spend, commission, and order counts independently. A correct order total does not prove that a commission export uses the same dates or status rules.

Give the next analyst a reproducible handoff

Retain the source filename, selection rule, excluded overlapping files, and final reporting window. This makes a handoff between media buyers and advertisers easier to audit without exposing customer-level information in shared summaries.

The useful result is a total whose coverage can be explained. Once the team knows whether a file replaces earlier data or adds new activity, it can compare ecommerce CPS and virtual-product campaigns without introducing an avoidable arithmetic error at the import stage.