Your order system records a sale. An advertising platform credits a campaign. Your analytics report credits a different channel. These records can describe the same purchase while answering different questions. Before deciding that a report is wrong, find out what each system is counting and which rules it uses to assign credit.

A useful attribution process makes those rules visible. It gives your team a consistent basis for decisions without pretending that every tool has an identical view of the customer journey.

Name the outcome before comparing the count

Start with the actual business action. Does “lead” mean a submitted form, a verified contact or an enquiry accepted by the sales team? Does “purchase” mean an order created, payment received or an order that remains after cancellations? Two dashboards can use the same label for different events.

Write a short measurement definition: the trigger, the system recording it, the identifier used where available, the value and the treatment of repeats or reversals. Include diagnostic actions, such as viewing a product, separately from the outcome used to judge campaign success.

Product terminology also matters. Google Analytics distinguishes important behavioral key events from conversions used for advertising measurement. Do not assume that every number labeled “key event” is the same action or scope as the conversion column in an ad account.

This definition is a working agreement for the team. It helps a media buyer, analyst and sales colleague identify whether they are looking at the same stage of the journey.

Separate counting rules from duplicate events

A customer may submit a form twice or place two legitimate orders after one interaction. You need a rule for meaningful repeated actions and a separate way to identify accidental duplication in the collection process.

Google Ads, for example, offers one or every conversion counting options for a tracked action after an interaction. The “one” setting is not a universal count of unique people across all campaigns and time. Its meaning belongs to that platform’s conversion-action rules.

In your implementation review, ask whether a confirmation page reload can resend an event, whether multiple integrations report the same transaction and whether two different actions have been grouped into one total. A count setting does not replace these checks.

Record exclusions as deliberately as inclusions. If an internal sales report removes cancelled orders but an advertising report retains the original conversion until an adjustment is processed, the difference needs an explanation before it needs a campaign change.

Understand which interactions are eligible

A conversion window defines how long an action remains eligible after an advertising interaction. Google Ads documents separate windows for clicks, engaged video views and impressions. Supported settings and definitions depend on the product and conversion type.

Imagine a purchase eight days after an ad click. A hypothetical seven-day click window would exclude it; a fourteen-day window could include it, assuming the other requirements were met. The extra reported credit comes from a different rule, not an additional order.

Also distinguish the conversion setting from a report’s lookback control. One can govern which conversions are recorded for an interaction; another can govern which earlier interactions are considered in a particular analysis. Read the definition attached to the exact field you are comparing.

Choose settings that serve your measurement purpose and reflect the decision cycle you want to examine. Do not lengthen a window simply to make results look stronger. Keep a record of changes so that an apparent improvement is not mistaken for new customer activity.

Treat the model as a credit-allocation rule

An attribution model distributes credit among eligible interactions. A last-interaction rule concentrates credit at a particular point in the observed journey; a model using additional information may distribute it. Google Analytics explains that data-driven attribution can assign fractional credit, which is why reports may contain decimal conversion totals.

A fraction does not mean that a customer completed part of an order. It means that a report allocated part of the credit for an action to the selected channel or campaign.

Keep the report’s scope in view. The same Analytics guidance distinguishes event-level attribution from user- and session-scoped traffic dimensions. Changing a reporting model does not make every acquisition table answer the same question.

Attribution is also not a complete causal explanation. Receiving credit under a rule does not establish how many purchases would have disappeared without the advertisement. Use attribution to organize observed performance, and treat questions about additional impact as a separate measurement task.

Follow one fictional order through the reports

Consider a fictional desk-accessory shop. A customer clicks an advertisement from Network A, later clicks an email and then completes one $60 order. Assume that all three interactions are correctly recorded and that both marketing interactions fall inside the relevant windows. The following simplified rules are illustrative, not descriptions of a specific vendor’s default configuration.

One fictional order, three reporting perspectives
Report Question and rule Illustrative result
Order system How many completed orders exist? 1 order, $60
Network A report Was a purchase recorded after an eligible Network A click? 1 credited conversion, $60
Cross-channel report Which eligible channel received the last interaction before purchase? Email: 1 credited conversion, $60

Adding the last two rows would produce two credited conversions and $120 of attributed value. The shop still has one $60 order. The reports overlap; they are not independent sales ledgers.

A different cross-channel model could divide credit between Network A and email. That changes the allocation, not the order. Real implementations may observe different portions of the journey, so this example explains the logic rather than promising perfect reconciliation at the individual level.

Investigate differences in a consistent order

Start with a stable reporting period and capture the settings alongside the numbers. Google’s conversion-management documentation highlights eligible channels, counting methods, models and account time zones as settings that affect comparisons. A cross-channel report and a paid-channel report may legitimately include different credit.

  1. Match the action. Compare the same event and business stage, not purchases against all important actions.
  2. Match the rules. Check counting, windows, eligible interaction types and attribution scope.
  3. Match the reporting period. Confirm time zones and whether rows follow interaction time or conversion time.
  4. Inspect collection. Check event triggers, duplicate integrations, missing identifiers and known tracking interruptions.
  5. Allow for processing. Separate a provisional report from one that has had time to incorporate later activity.

Google’s discrepancy troubleshooting guide discusses tagging, conversion delay, count settings and other causes. It also points to conversion-time columns for more aligned comparisons. These are diagnostic options, not a promise that all systems will match exactly.

Choose the view that fits the decision

Use the business system to establish the defined outcome total. Use platform reports to inspect delivery and the conversion signals available to that buying system. Use cross-channel reporting to examine credit under a consistent shared framework.

Put the definition beside the number: “completed orders,” “platform-attributed purchases” or “qualified enquiries,” rather than a bare “conversions.” Note material setting changes in the same review record. A reader should be able to understand a movement without reconstructing the account’s configuration.

When deciding whether to increase spend, connect attribution to lead quality, order quality and a deliberate testing approach. The goal is an explainable decision. A well-understood difference between reports is more useful than apparent agreement produced by comparing the wrong fields.

Sources & further reading

Explore advertising with Promoweb.
Meet our self-serve platform