Every channel claims the sale. Who deserves the next pound?

The business took £500,000 in revenue. Add up the channel reports and marketing appears to have delivered £700,000. Nobody's necessarily invented a sale, but somebody now has to explain the extra £200,000 to finance. That conversation gets uncomfortable quickly, especially when the same reports support next month's budget requests. Before accusing a team or platform of inflating its contribution, work out what the numbers actually count.

Attribution gives credit to marketing interactions using particular rules and observations. Google Analytics describes attribution models in those terms: methods for assigning credit to touchpoints along a path.\[1\] A customer can encounter several channels before buying, and separate platforms can each count the purchase within their own reporting rules. Adding those claims together doesn't produce a deduplicated view of business revenue. It certainly doesn't establish how much revenue would disappear without the advertising.

Start with definitions

In a mock example, paid search claims £300,000, paid social claims £250,000 and an affiliate report shows £150,000 against £500,000 of business revenue. Although the £700,000 total could include overlapping purchases, differences in periods, currencies, refund treatments or conversion values could also explain part of the gap. The numbers alone don't tell you which explanation dominates, and the £200,000 difference isn't automatically the exact value of duplicated sales.

To make the comparison useful, establish whether the platforms report by interaction date or purchase date, whether they include view-through credit and how long an interaction remains eligible. Then check whether refunds are removed and whether the reported values represent revenue, predicted customer value or something else. Until those definitions are clear, the team could be arguing about numbers that were never comparable. Google Analytics' documentation confirms that attribution settings affect how credit is allocated, and its cross-channel reporting guidance distinguishes conversion-time and interaction-time views.\[2\] Checking those details can resolve an argument that initially sounds like a disagreement about channel effectiveness.

Build a disagreement ledger

A shared ledger makes the uncertainty manageable by giving each discrepancy a row, a likely explanation and an action to investigate it. With those checks visible in one place, marketing and finance can work through a common investigation instead of each trying to prove their own report is right.

Each check needs an owner who can report what changed and explain the implications. Where the available data can't resolve a discrepancy, that uncertainty needs to stay visible in the decision. “Unresolved, with these consequences for the decision” is a more useful answer than repeatedly switching between systems until one produces a comfortable result.

A reconciled total can still lead you astray

Although deduplicating transactions is valuable for reporting, it doesn't determine who caused the purchase. A single order might follow a social impression, a search click and an affiliate visit, so assigning 100% of the credit to the last eligible interaction gives you a tidy answer under that allocation rule. What it leaves open is the question the budget owner actually needs to consider: what would've happened if one of those activities hadn't been there?

That distinction matters when the budget starts following the credit. Activity close to the purchase can look excellent because it reaches people already likely to buy, while activity further from the transaction can be harder to observe. Since neither pattern proves that a channel deserves more or less money, we'd need to investigate its role, the customers it attracts and what happens when investment changes before recommending a move.

You can haircut every platform's numbers until the spreadsheet balances, but you still haven't established what caused the sales. If that adjustment is an accounting convention, label it accordingly so it doesn't quietly become an estimate of causal return. Otherwise, the next budget decision will rest on a reporting fix dressed up as evidence.

If it's working, where are my sales?

When someone argues that a channel might be driving sales we can't see, I'd ask what we'd expect to observe in the business if that were true. If sales volume hasn't improved, we need a credible explanation for why investing more would help. Measurement gaps exist, although their existence doesn't automatically turn a disappointing result into hidden success. The person asking for more budget needs to explain the mechanism, the expected commercial outcome, the time horizon and the evidence that would make them reconsider.

“If it's working, where are my sales?” is a reasonable challenge to an investment claim. Flat sales alone don't prove a channel had no effect: underlying demand might have weakened, customers might convert later, or a product constraint might have offset a gain. Those possibilities deserve investigation, which means naming the evidence each explanation would predict and comparing it with what happened. That gives us a basis for deciding whether a bounded test is worth funding. An explanation that can accommodate every possible result gives the business no way to decide when to stop.

Give each source a defined job

Reconciled business records establish the outcome you've achieved within their reporting scope, while attribution helps explain recorded paths and support campaign operations. Customer economics then lets you judge whether that outcome is worth paying for. Where the investment warrants further work, a suitable model or feasible experiment can help investigate incremental response. Each source needs its own quality checks, including the commercial records. Replacing several imperfect views with one unchallenged number doesn't remove uncertainty.

For the immediate decision, write down the investment at stake, the evidence supporting it and what remains unresolved. A modest, reversible change might be reasonable while the team fixes the reporting, whereas a major reallocation based entirely on conflicting platform claims deserves a higher bar. Agreement around the table doesn't lower it. If we can't support the recommendation, the honest answer is to hold the decision or propose a bounded test, with the commercial cost of waiting made explicit. If you need an experiment, establish whether it can answer the question at a commercially useful level before promising the board a definitive result.

Marketing loses trust when every report seems to justify more spend and nobody can explain the total. You can rebuild that trust by being precise about what each number means, admitting its limits and connecting the evidence to an action someone can review. Finance shouldn't have to choose between blind faith and cutting the budget.

If competing attribution claims are holding up an investment decision, GV Marketing can help reconcile the evidence and work out what deserves your next pound.

Read next: work through conflicting budget evidence or check what your ROAS target rewards.