Two people can export an advertising report for the same campaign and produce different purchase totals without either person making an arithmetic error. The reports may use different attribution windows, conversion actions, or date conventions.
The solution is to make the report definition explicit before debating performance. This guide gives a reconciliation method. Use the window labels and options available in your current Meta account rather than assuming that a remembered default applies to every campaign or report.
Create a report identity card
Record the platform, account ID, campaign scope, date range, timezone, conversion action, attribution view, and value definition. Include whether the report groups results by interaction date, conversion date, or another date basis exposed by the tool.
Save the export time. Recent periods can continue changing as events are processed or attributed. Two exports made at different times may not be identical even if the filters match.
Put these definitions in a shared metric dictionary. A report labeled only “Meta ROAS” leaves too many choices invisible.
Understand what changing the window changes
An attribution window is an eligibility rule for assigning credit to interactions before a conversion. A longer eligible period can include conversions that a shorter period excludes. A view-based component and a click-based component can also describe different interaction paths.
This does not mean the longer window creates additional business orders. It changes which orders the reporting system can associate with the advertising interaction under that definition.
Meta's holistic measurement course outline includes reporting, experiments, lift studies, and marketing mix modeling as different measurement approaches. Keep the distinction visible when interpreting attribution views.
Do not add overlapping views
Consider an illustrative set of ten business orders. A short click window credits four of them, while a longer click window credits seven. If the four are included in the seven, adding the two reports would produce eleven attributed orders from ten actual orders.
The correct interpretation is that the longer window includes three additional credited orders under that example's rules. It is not a separate group to add to the shorter-window total.
The same caution applies when comparing platform reports. Two advertising platforms can each claim credit for an order that involved both channels. Summing their attributed revenue does not necessarily produce unique business revenue.
Reconcile one definition at a time
| Step | Hold constant | Inspect |
|---|---|---|
| Conversion action | Purchase or another exact event | Whether both reports count the same action |
| Scope | Account, campaigns, and ads | Excluded or missing objects |
| Time | Period and timezone | Boundary differences and partial days |
| Attribution | Window and interaction type | Additional eligible credited conversions |
| Value | Revenue basis and currency | Shipping, tax, discounts, and refunds |
| Export maturity | Data collection time | Later arrivals or processing changes |
Change one report setting at a time and record the difference. If several settings change together, the team cannot tell which explains the discrepancy.
Use the timezone reconciliation guide when transactions near midnight appear on different days.
Compare with the order ledger carefully
The order system can establish that an order exists and its commercial status. It does not automatically establish which marketing interaction caused it. Advertising attribution is a separate assignment of credit.
For ecommerce, preserve transaction identifiers and consistent value definitions in the measurement path. Google's GA4 ecommerce documentation describes purchase and refund event data that can support a structured reconciliation. GA4 event collection still does not guarantee that Meta or another platform will assign the same credit.
If gross attributed revenue exceeds retained order revenue, inspect refunds, cancellations, value definitions, and overlapping credit before calling the integration broken.
Use the right view for the decision
A consistent attribution view can help compare campaign trends within an account. A business-level revenue and contribution view can help the owner understand overall economics. A suitable experiment can help answer whether advertising produced outcomes that would not otherwise have occurred.
These views can inform one another without being forced into a single number. For example, a campaign may have strong attributed ROAS while the business still needs to test how much of that demand was incremental.
Read the attribution-versus-incrementality guide before using an attribution export as proof of causal lift.
Explain discrepancies in plain language
A useful report note might say: “This export uses the longer click-attribution view, so it includes orders outside the shorter view used in last week's report. The order ledger is unchanged; the credited set differs. Future comparisons will use the same view.”
If the difference remains unexplained, say which definitions were matched and which source is still being investigated. Do not hide the gap with a blended average or choose whichever report looks more favorable.
Keep the original exports. A reconciliation record should let another person reproduce the comparison rather than depend on an undocumented sequence of interface clicks.
Lock the convention for recurring reporting
Once the team agrees on a reporting view, use it consistently and document deliberate changes. A new attribution convention may be useful, but it can create a break in the historical series.
For that transition, present a period under both definitions so readers can understand the effect. Do not silently restate an old report or imply that the difference represents newly generated sales.
The purpose of reconciliation is not to make every system agree. It is to explain what each system counts, where the differences come from, and which view is appropriate for the decision being made.