How to reconcile marketing attribution with real revenue

Start with the source, not the score

A dashboard total is the end of a chain. At the beginning are source accounts, exports or API responses, timestamps, identifiers, and the rules used to reconcile records. If the chain is invisible, a precise-looking total can conceal stale data, duplicate orders, a disconnected account, or a date-range mismatch.

For each result, name the source and the period it covers. Distinguish platform-reported conversions from orders or revenue recorded in a commerce system or CRM. Those systems answer related questions, but they do not observe the same events in the same way. A platform's reported conversion is not automatically a finance-confirmed sale.

Make the calculation reproducible

A result should state its definition. Is revenue gross or net of discounts, refunds, and tax? Are conversions deduplicated across sources? Which currency and time zone are used? What is the attribution window? Those choices can materially change a comparison even when every underlying event is correct.

Keep the model name and its rule beside the output. First-touch, last-touch, linear, and time-decay models distribute credit differently across the same recorded journey. The calculation should be understandable from its inputs and rule, not presented as an unexplained score.

• Source and account used
• Date range and data freshness
• Revenue and conversion definitions
• Attribution window and model
• Known exclusions, gaps, and unmatched records

Show coverage as part of the answer

Coverage describes how much of the relevant activity is visible and usable for the question. A missing ad account, incomplete CRM history, or low rate of matchable journeys limits what a report can say. A team may still use a partial view, but it should not mistake that view for a complete picture.

Treat unknown and zero as different states. Zero means the source was checked and no qualifying activity was found under the stated rules. Unknown means the data needed to decide is absent, late, or could not be matched. Replacing unknown with zero quietly changes the evidence and can distort a channel comparison.

Use attribution for the decision it can support

Attribution describes how a chosen rule assigns credit to recorded touchpoints before an outcome. It can help teams compare journeys consistently, identify where to investigate, and make their assumptions explicit. It does not, by itself, prove that an ad caused an order or estimate what would have happened without the ad.

For causal questions, use an appropriate experiment or incrementality design. For operational budget reviews, use attribution as one clearly labeled view alongside spend, confirmed outcomes, margins, and business context. Keep the distinction visible in the report and in the conversation around it.

A five-minute evidence review

Before sharing a result, ask a colleague to follow it from the headline back to its basis. If they cannot tell what is observed, what is calculated, and what is unavailable, the report needs more context.

• Confirm the source and last successful sync
• Check the date range, currency, and time zone
• Read the attribution rule and conversion window
• Look for duplicate, refunded, or unmatched records
• State what the result cannot establish

Work through a revenue reconciliation

Consider a fictional store reviewing a full calendar month. The store recorded £50,000 in gross order value. Discounts accounted for £3,000 and refunds for another £2,000, leaving £45,000 under this example's agreed net revenue definition. Meta reported £32,000 in conversion value and Google Ads reported £28,000. Adding the platform figures would suggest £60,000, but those figures can contain overlapping claims and different valuation rules. They do not describe £60,000 of unique store revenue.

The team next links £36,000 of its net revenue to eligible recorded journeys. The remaining £9,000 has no eligible matched journey under the chosen window and identity rule. It still belongs in the store's revenue total. It should sit in an unmatched category in the journey report rather than disappear or be spread across paid channels to make the report look complete.

Turn a report into a decision brief

A performance marketer needs to know whether to increase a budget, investigate a drop, or wait for a delayed source. A business owner needs to know what the company spent, what customers paid, and how much confidence to place in the comparison. Put that decision at the top of the report. A table of metrics becomes more useful when the reader knows what choice it is meant to inform.

For the store above, a useful brief would say: net revenue was £45,000, £36,000 was linked to recorded journeys, and £9,000 remained unmatched. Credit was calculated with a stated model over a stated window. The next step is to inspect the unmatched group and compare model allocations before changing channel budgets. That tells the team what happened and what to do, without claiming more than the records support.

Include the comparison period and a reason for choosing it. A month with a promotion or a stock shortage is not directly comparable to an ordinary month. If the team uses a prior period, annotate the commercial differences. Evidence includes the conditions under which the business operated, not just the events a connector returned.

Keep an evidence trail when a number changes

Reports can change after publication because orders are refunded, conversions arrive late, permissions change, or historical data is backfilled. A changing result is not automatically a failure. The problem is a change with no explanation. Save the reporting period, source update time, definition, and calculation version with the exported result.

When a later report differs, reconcile the difference into recognizable causes: newly imported records, removed duplicates, revised order values, changed channel classifications, or a different model. Avoid rewriting the earlier report as though the corrected information had always been available. A revision note helps colleagues distinguish a data update from an actual shift in marketing performance.

• Archive the report with the period and source update times.
• Record the revenue definition and attribution settings.
• Explain material corrections before comparing exports.
• Keep the same conversion set when testing a different model.

Make the weekly review short and repeatable

Start with source health, then review spend and outcomes, then inspect attribution. This order prevents a disconnected account from becoming a performance story. Check that every material source has reached the end of the reporting period and that the same revenue definition applies across the current and comparison periods.

Choose two or three material movements to investigate. Open a sample of journeys behind each movement, confirm campaign classification, and look for changes in customer mix, offers, and landing pages. End with an owner and a next action for each finding. The best report is the one that helps the team decide what to check or change, then makes it possible to review whether that action helped.