Add up the sales reported by your advertising platforms and the total exceeds the orders in your store. Before assuming someone has made a mistake, check what each system is counting.
When the figures do not match, questions are reasonable. Start with the definitions and trace an order through the systems. Then separate explainable differences from implementation errors.
An advertising report connects an action with previous interactions. The store’s records track orders. These systems can describe the same customers using different rules.
What does attribution mean?
Attribution is how credit is assigned to the interactions that precede an action. The model and attribution window affect the result. Google explains this logic in its Analytics documentation. Google Analytics: introduction to attribution
For a specific comparison, we check the settings of the systems involved: which action is tracked, which interactions are eligible and within what window. We do not assume that all platforms apply the same rule.
One person can encounter several ads
The following example is fictional. We assume that the rules of each system allow the order to be attributed to its interaction.
On Monday, someone clicks an ad and discovers a bag. On Thursday, they see another ad and return. On Friday, they buy. The store records one order with one unique identifier.
Two attribution reports can describe one order.
Fictional example: both platforms’ attribution rules allow them to claim the order.
Monday
Interaction with ad A
Thursday
Interaction with ad B
Friday
One order
Report A
1 attributed conversion
Report B
1 attributed conversion
Store
1 unique order
| Source | What appears in the example | How we read it |
|---|---|---|
| Platform A report | One attributed conversion | A credits its own interaction |
| Platform B report | One attributed conversion | B credits its own interaction |
| Store records | One unique order | The business has received one order |
Two platforms taking credit does not turn one order into two sales. Simply adding their totals can therefore overstate the business’s actual volume.
Differences may also have other explanations
We check the time period and time zone, reporting time, data delay, cancellations and returns. Some reports may date the result by the interaction, others by the action, depending on configuration.
There may also be implementation problems: the same action sent twice, wrong values or events that do not represent a completed purchase. They must be investigated separately from the natural differences between the attribution models.
Missing data can result from technical and measurement limitations. A difference between systems does not, on its own, prove fraud, and it should not be ignored either.
Build a consistent reporting framework
For unique orders, we choose a reference commercial record and define the stages: placed, paid, delivered, canceled or returned. For advertising, we keep separate reports and their rules.
We compare transaction identifiers where they are available and where their use is permitted. We explain what can be reconciled exactly and what remains estimated or incomplete. We do not artificially adjust numbers just to fit.
On this basis, we can discuss both the total outcome of the business and the information each platform provides about its own campaigns.
What we assume in interpretation
At NOT MORE., every metric should have a clear definition and purpose. We flag overlaps and limitations, investigate implementation errors within our scope, and avoid presenting combined attribution totals as unique orders.
Even after reconciliation, a separate question remains: how much did advertising add? Answering that can require evaluation beyond routine platform reports.
What’s worth asking
- Which source records unique orders, and which order statuses does it include?
- Which rules and attribution windows do the advertising reports use?
- Which differences have been explained, and which still need investigation?
What does each of your reports count?
Start with the metric names and the period in which the differences appear. If inquiries or transactions need analyzing, we agree on what information is necessary and how to use it without collecting unnecessary personal data.
A shared source needs an implementation and an owner
For unique orders and customers, we agree on which system supplies the commercial record and who confirms its accuracy. NOT MORE. can implement the events, integrations and shared report included in the project, then check that data connects according to the agreed definitions. Your team confirms order status and changes arising from operations.
If we continue monitoring, we report platform attribution separately from commercial results, alongside any errors or gaps we observe. A documented difference is something to investigate. Adding overlapping results together would conceal it.