What should a marketing report help you decide?

Name the decision before choosing the chart

Before your next reporting meeting, finish this sentence: “After this review, we need to decide whether to…” You might need to keep a campaign running, test a different landing page or investigate why new customers are requesting refunds. Those decisions need different evidence.

A question such as “How did marketing perform?” is too broad to guide the report. A more useful question is: “Did the customers acquired in this period justify another test at the same budget?” It identifies the population, the commercial concern and the decision. It also leaves room for an honest answer: the evidence may not be mature enough yet.

Agree what would change your mind before discussing the result. For a measurement problem, that might be a reconciled set of transactions. For an acquisition decision, it might be a comparable cohort with an agreed observation window and cost boundary. A click-through-rate chart can help explain what happened, but it cannot replace the evidence needed for the business decision.

Give every outcome a usable definition

Write a short definition beside each decision metric. “Customer,” “conversion” and “revenue” are not precise enough on their own. A trial starter and a first-time payer are different people for reporting purposes. A reservation and a completed hotel stay represent different stages of the same journey.

• Unit: what exactly is counted, and what makes it unique? Count customers, orders or room nights deliberately.
• Source: which records establish the outcome? Name the relevant billing, CRM or booking records rather than calling every dashboard the source of truth.
• Dates: when did acquisition occur, how long can the outcome mature, and what is the reporting cutoff?
• Adjustments: how are duplicates, refunds, cancellations, test transactions and unmatched records treated?
• Cost boundary: which expenses enter the calculation, and which are excluded?

For example, “unique first-time customers acquired during the selected week whose first payment has not been fully refunded by day 30” is a usable definition. It is a proposed reporting choice, not a universal customer definition. It would need adaptation for partial refunds, multiple products or an advertising-funded app.

Use a shared time zone and reporting currency. If one system records a payment shortly before midnight and another places it on the next day, a daily comparison can appear inconsistent. Keep original amounts and document the currency conversion rule when combining markets. Resolve these choices once in the metric definition instead of rediscovering them in every meeting.

Calculate the cost without changing the denominator

Consider a hypothetical acquisition cohort observed through day 30. Media spend was $10,000, with another $2,000 of creative production allocated to this cohort. There were 250 unique first-time payers. Fifty of them received full refunds, leaving 200 customers with a retained first payment. Assume no partial refunds for this example.

All three results are arithmetically correct. The first describes initial payment acquisition. The second incorporates the chosen refund adjustment. The third includes an additional expense. Reporting “acquisition cost is $40” without the definition hides two choices the reader needs to understand.

The $60 figure is not necessarily an all-in customer acquisition cost. It excludes any expenses outside this example's boundary, such as sales costs or agency fees. It also says nothing about profitability without knowing the value and cost of serving those customers. If production serves several cohorts, document the allocation method; do not charge it to one period and omit it from another without explanation.

Keep the same definitions when comparing campaigns or periods. A cohort at day 7 has not had the same opportunity to pay or request a refund as one at day 30. AppsFlyer's cohort documentation distinguishes complete from partial periods and explains how calendar boundaries can differ between systems. Select a window appropriate to your customer journey and mark immature observations as incomplete.

Reconcile systems before explaining performance

An ad platform can attribute a transaction to an interaction while a billing system records that the transaction happened. These are related questions, but the totals need not describe the same population or period. Google Analytics explains attribution as assigning credit along a user's path and identifies the reporting model, eligible channels and lookback window as settings to review.

When systems disagree, investigate in a consistent order:

1. Check whether the event and unit match. A repeated purchase event is not automatically a second customer.
2. Check acquisition dates, transaction dates, time zones and the report cutoff. Export timestamps with the records.
3. Compare inclusion rules: new versus existing customers, full refunds, test orders and unobserved activity.
4. Inspect attribution settings and the scope each platform can observe. Keep overlapping platform claims separate from deduplicated business outcomes.
5. Where permitted and technically available, compare stable transaction identifiers. Record unmatched items and investigate their cause rather than silently dropping them.

If identifiers are unavailable or privacy restrictions prevent a join, state the reconciliation limit. You can still compare aggregate trends with appropriate caveats; you cannot claim an exact campaign-to-customer match that the data does not establish. Avoid summing overlapping platform conversion totals into a supposedly unique customer total.

A report should show the size and nature of the unresolved gap. “Payment records are complete; campaign matching is incomplete” is more useful than either declaring all measurement broken or choosing whichever dashboard reports the strongest result.

Check whether the comparison explains the change

Suppose cost per retained first payer fell. Before increasing spend, check what changed around that number. Did the campaign reach a different country? Did a discount change payment conversion? Was more spending directed toward people already familiar with the product? Did refunds have enough time to appear?

Break the comparison into groups only where the distinction can change your decision. A country or product split may reveal that an overall improvement came from a shift in mix rather than better performance within each group. Keep the same cost and outcome definitions for each comparison. Avoid creating dozens of tiny segments that look precise but contain very little evidence.

Separate a recorded observation from an explanation. “The retained-payer cost fell” is an observation under a stated definition. “The new creative caused the improvement” is a hypothesis unless your evaluation supports that causal conclusion. Record simultaneous changes so the team can see competing explanations.

When the result depends on a handful of customers, show the counts next to the percentage or cost. There is no universal sample threshold that makes every marketing decision safe. The evidence required depends on variability, the proposed test and the consequence of getting it wrong. A small exploratory test and a large budget commitment should not be treated as equivalent decisions.

Show the decision the evidence actually supports

Return to the hypothetical cohort above. The report establishes the scoped costs and the refund adjustment. It does not establish customer profitability, explain why refunds occurred or prove which ad caused a payment. A completed decision record could look like this:

This example deliberately stops short of a campaign verdict. That is the useful result: it identifies what the team can conclude and what would be required to go further. Another business with a complete contribution model and different evidence could reach a different decision.

Make the review repeatable without adding reporting work

Use one short decision page, supported by the source detail needed to challenge it. Before the meeting, the person responsible for measurement updates the definitions and cutoff, checks the arithmetic and flags missing data. The person responsible for the commercial outcome supplies the context the dashboard cannot establish on its own.

During the meeting, start with the previous action: what was done, what was observed and whether the original question was answered. Then examine the current decision. Move diagnostic charts into the discussion when they help resolve a disagreement, rather than giving every chart equal time.

After the meeting, retain the decision, owner and review condition with the report version used to make it. If the source later changes, preserve the earlier snapshot and explain the correction. Otherwise, a later reader may judge an old decision against information that was not available at the time.

When a required input is missing, assign its resolution explicitly. “Wait for more data” needs a reason and a condition for returning to the question, such as completion of the observation period or reconciliation of the affected transactions. Adapt the review routine to the people who own those decisions.

Bring a clear record to your next reporting meeting

Copy these fields into your next report and fill them with your own agreed definitions:

• Decision: What exactly are we deciding?
• Evidence: Which outcome, source, cohort, window and cost boundary are we using?
• Observation: What changed, with the underlying counts and relevant comparison?
• Uncertainty: What is incomplete, unmatched or still only a hypothesis?
• Action: What happens next, who owns it and what would change the decision?
• Review: When will we revisit it, and which new evidence should be available?

If those fields cannot be completed, the next task is to resolve the missing definition or evidence. If they can, the report is ready to support a discussion about action. Keep the supporting records available so someone outside the reporting team can follow the reasoning.

For help connecting campaign measurement to business outcomes, see how MORE approaches measurement, or bring your reporting question to us through the contact form.