Use store records to establish what was sold, platform reports to understand credited interactions and a defensible experiment to investigate added sales. MER, CAC and ROAS each answer a narrower question. None is a complete profit calculation or automatic proof that a channel caused an order.
A fictional overlap that does not add up
A fictional store records $50,000 of defined revenue in a period. Two advertising platforms attribute $35,000 and $25,000 respectively. Adding them produces $60,000, which is more than the store total.
That does not automatically prove a tracking defect. The platforms may credit the same order under different windows and rules. Check dates, currencies, refunds, view-through treatment and actual event counts before deciding what is wrong.
The example is fictional. Its lesson is to keep source totals labeled and reconcile the definitions. Do not present the sum of attributed revenue as additional store revenue.
Define the three ratios in your own report
For this guide, MER means defined store revenue divided by paid media spend for the same period. Write which revenue and spending are included. A $50,000 numerator and $10,000 denominator produce 5.0. That is an aggregate ratio, not a causal result.
Media-only acquisition cost means the defined acquisition media spend divided by qualifying new customers. A broader CAC can include additional sales and marketing costs. Give them different labels so a change in scope does not look like improved performance.
Platform ROAS means attributed conversion value divided by the platform spend under the selected reporting settings. Record the attribution model, window and conversion action. A value marked as revenue may still include tax, shipping or duplicated events unless the implementation defines it correctly.
Reconcile events with orders before buying another tool
Choose a completed reporting window and compare qualifying orders with the purchase events intended to represent them. Inspect a permitted sample for stable transaction identifiers, value, currency, refunds, test records and repeated submissions.
Distinguish event receipt from processed reporting. A debugging view can show that an event arrived without proving it has been included in a final report. Delays and consent-related gaps need labels, not fabricated zeroes or bypassed consent.
If browser and server integrations describe the same purchase, verify the provider’s current deduplication method. Do not manufacture a purchase event from an unfinished checkout to make a report look active.
Use incrementality as a separate investigation
Attributed orders are orders credited under a model. Additional orders are the estimated difference caused by an intervention. A holdout or another suitable experiment can help investigate that difference, but design, sample size, contamination and uncertainty matter.
Do not treat MER as a replacement for that experiment. It can move because of promotions, returning customers, organic demand, stock or product mix. Nor does another attribution tool automatically reveal “true” channel contribution.
Keep a weekly note with the metric definitions, dates, source totals, unresolved differences and next decision. Your next step is to reconcile one disputed number before changing a budget. Use the scorecard guide to keep that decision visible.
Editorial revision, October 20, 2026: unsupported CAC targets, product-feature claims and attribution-lift promises have been removed. Numerical examples are fictional arithmetic; no measurement method is presented as certainty.