When a report lacks a source, do not automatically label it direct, organic, or AI traffic to make the totals look complete. Preserve the missing evidence and explain what the selected analytics system actually recorded.
Separate platform classifications from your own reconciliation categories. Google Analytics, for example, defines its default channel groups through documented rules. Its Direct and Unassigned states are not interchangeable labels for every record your separate analysis cannot explain. Those rules were checked September 9, 2026.
The method below is an original reporting framework with invented counts. It does not change tracking, infer individual identity, or claim a complete view of any customer journey.
Keep evidence types in separate fields
| Evidence | What it establishes | Limit |
|---|---|---|
| Recorded source or campaign | The value captured under the system's rules | May not describe every earlier influence |
| Self-reported discovery | What the respondent says influenced discovery | Can overlap with recorded evidence |
| Modeled attribution | An estimate under a specified method | Is not the same as a directly observed event |
| Unknown in this review | The available records do not resolve the source | Does not prove absence of marketing influence |
Retain the source system's original values alongside your analysis fields. That makes the report reconcilable when definitions or joins change.
Work through overlapping evidence
Suppose a fictional cohort has thirty enquiries. Eighteen have usable recorded acquisition evidence and twelve remain unknown in that dataset. Ten people also answer a discovery question; eight of those responses belong to the eighteen already observed, and two belong to the twelve unknown.
The ten survey answers do not create ten additional enquiries. They add another view of existing records. The report still contains thirty enquiries. Two previously unknown records now have self-reported evidence, while ten remain without either type in this example.
Do not convert the two self-reported sources into a claim that the tracking system observed those paths. Preserve the distinction between recorded and reported evidence even when both are useful for the business decision.
Investigate a concrete gap
Choose a small sample of unresolved records and inspect the actual data path. A missing campaign value could reflect an untagged link, a broken join, a scope mismatch, or unavailable data. Identify the evidence before assigning a cause.
Check whether the report combines session, user, and event-level dimensions inconsistently. A record can appear unresolved because the analyst joined incompatible units, not because the website failed to collect every field.
Review consent and collection behavior through the appropriate current implementation documentation when it is relevant. Do not treat missing data as permission to bypass user choices or add tracking that has not been authorized.
Show coverage with the outcome
Report the eligible record count, observed-source coverage, self-report response coverage, overlap, and unresolved count. This lets a reader judge how much of the business outcome the analysis can explain.
Keep changes in measurement coverage visible over time. An increase in identifiable referrals can result from a repaired tag or classification change as well as a real acquisition change. The release record should help distinguish those possibilities.
Use the cohort review to connect evidence with mature outcomes. A credible report can be incomplete. It becomes misleading when it hides the incompleteness behind a confident channel label.