Guide

Why High ROAS Can Hide Unprofitable Orders: Work from Product Contribution

By

A campaign can report five pounds of conversion value for every pound spent and still leave a product group with a negative contribution. ROAS describes the relationship between the value reported to the advertising platform and advertising spend. It does not, by itself, deduct product costs, refunds or the cost of getting orders to customers.

The useful commercial question is which product groups leave money after the costs included in your decision. Build a bridge from a defined sales value to contribution after variable costs and advertising. Keep that result separate from company net profit and from the platform’s attribution claims.

Quick answer

  • Document what the advertising conversion value includes before interpreting ROAS.
  • Use order and item records to establish retained revenue after the relevant refunds.
  • Subtract product costs and the variable fulfilment, payment and return costs included in your model.
  • Deduct advertising cost once, with a defensible product-group assignment.
  • Show missing costs and unresolved returns explicitly before making a spending decision.

What This Guide Does Not Solve

This guide explains a management contribution analysis for UK e-commerce teams. It does not provide tax advice, calculate statutory profit, prescribe a universal target ROAS or prove that advertising caused every attributed order. The accounting basis and cost classifications should be agreed with the person responsible for your finances.

When to Pause or Escalate

Stop the profitability conclusion if the revenue basis is unclear. If one report includes VAT, another includes delivery charges and a third excludes refunds, their values cannot be compared as if they describe the same sales. Reconcile the definitions before labelling a product group profitable or unprofitable.

Also pause if material product costs are missing or order groups cannot be matched to advertising spend. An incomplete contribution view may still expose a question worth investigating, but it should not become an exact performance claim or an automatic instruction to increase budgets.

Separate ROAS, Gross Profit and Contribution

ROAS is reported conversion value divided by advertising cost. A value of 5 means the reported conversion value is five times the spend, often expressed as 500%. The interpretation depends on what was submitted as conversion value and what the report attributes to the campaign.

Gross profit and contribution answer different questions. Google’s cost-of-goods guidance describes estimated gross profit as revenue less cost of goods sold. That narrower calculation does not establish what remains after every variable fulfilment cost or advertising expense.

The ACCA explanation of cost-volume-profit analysis distinguishes contribution from fixed costs and profit. For the management model in this guide, show two stages: contribution before advertising and contribution after advertising. Label the costs included at each stage.

A positive contribution after advertising leaves something towards fixed overheads and other excluded costs. It is not a company-profit figure. A negative result means the modelled group has not covered the costs included in that bridge, subject to the completeness and reliability of the records.

Agree the Revenue Basis Before Joining Data

Write a definition for the sales value used in the analysis. State its tax basis, whether discounts have already been deducted, whether delivery income is included and how cancellations and partial refunds are treated. Use the same definition throughout the bridge.

Do not subtract discounts twice. If the order-line revenue already reflects a promotional discount, the bridge should start with that discounted value. Likewise, if a dataset is already net of refunds, a second refund deduction would understate retained revenue.

Keep the platform value as a separate column when its basis differs. The purpose is to explain the difference, not force the order system to equal an advertising dashboard. Attribution windows, order dates and processing dates can also create legitimate differences that need separate treatment.

For the worked example below, all revenue and costs use an agreed VAT-exclusive management basis. Discounts are already reflected. Customer delivery income is zero. These are fictional modelling assumptions, not instructions about VAT treatment for your business.

Use Order Lines for Product Groups and Order Records for Shared Costs

Start with order ID, line ID, product identifier, quantity, sales value and the product group assigned at the time of analysis. Join the appropriate product-cost record and keep its effective date. A current purchase cost may not describe stock sold several months earlier.

Attach refund records to the affected order or line where the data supports it. Record the amount and date, and distinguish a full cancellation from a partial refund or returned item. A refund can occur without a physical return, while a returned product can have different recovery outcomes.

Some costs belong to the whole order, such as an outbound delivery charge paid to a carrier. If an order contains several product groups, define one allocation method and label it as an allocation. Revenue share, weight or another basis may serve different purposes; none should be presented as an observed product-level cost unless it actually is one.

Check that allocated totals still reconcile to the order total. A join that repeats the full delivery cost on every line can turn a profitable basket into an apparent loss. A join that drops orders without matching cost records can make the remaining sample look stronger than the full business.

Treat Refunds and Recovered Stock as Different Movements

A refund reduces retained revenue under the chosen model. It does not automatically mean the whole original product cost has been lost. A returned item may be restored to usable stock, written down or written off, and the treatment must follow the agreed records.

Show the product-cost adjustment separately from the customer refund. If the model reverses the cost of an item returned to saleable inventory, use the supported cost adjustment, not the retail refund value. Otherwise the bridge can manufacture margin by treating the selling price as recovered stock cost.

Return postage, inspection and other variable handling costs may remain even when the goods are recovered. Keep those in their own included cost lines. Record where a fee is non-refundable or where a cost is already included elsewhere, so it is neither omitted nor deducted twice.

Choose a consistent maturity point for comparing order cohorts. Recent orders may have fewer recorded refunds simply because less time has passed. Define the review cut-off and distinguish observed refunds from any estimate of future returns. Do not call an estimate an actual result.

Make Delivery and Payment Economics Visible

Separate what the customer pays for delivery from what the business spends fulfilling the order. If delivery income is included in revenue, subtract the relevant full delivery cost. If the model instead uses a net delivery subsidy, explain that approach and avoid also subtracting the full cost.

Decide which pick, pack, packaging and handling costs vary with the orders being analysed. Use supported rates or actual records and mark estimates. A warehouse salary does not automatically become a variable cost merely because it is divided by the number of orders.

Payment fees can also differ from a simple percentage of retained sales. Use the actual charges or a documented approximation suitable for the analysis. Do not assume a refund reverses every associated fee unless the records support that treatment.

Keep a reconciliation column for excluded or unallocated costs. It tells the reviewer what the contribution figure does not cover. Hiding those costs outside the report makes a narrow measure look more comprehensive than it is.

Worked Example: Equal ROAS, Different Contribution

All figures and product groups in this example are hypothetical. They are not client results or retail benchmarks. Assume two separate campaigns serve two non-overlapping product groups, with no mixed baskets in the illustration. Each has £2,000 advertising cost and £10,000 reported purchase value, giving reported ROAS of 5.

For simplicity, the reported purchase value initially matches the order revenue on the agreed basis. It has not been reduced for the later refunds shown below. The cohorts are reviewed at the same stated maturity point, and all listed costs are assumed available. The model is a teaching example, not a claim that a real account reconciles this neatly.

Hypothetical contribution bridge, GBP on a consistent VAT-exclusive management basis
Bridge line Product group A Product group B
Reported purchase value and initial order revenue £10,000 £10,000
Less customer refunds £500 £2,000
Retained revenue £9,500 £8,000
Dispatched product cost £4,750 £5,000
Less supported product-cost reversal for recovered stock £250 £1,000
Net product cost deducted £4,500 £4,000
Less outbound fulfilment costs £1,000 £1,800
Less payment fees £200 £200
Less return-handling costs £100 £400
Contribution before advertising £3,700 £1,600
Less advertising cost £2,000 £2,000
Contribution after advertising £1,700 −£400
Reported ROAS before these refund deductions 5.0 5.0

For A, retained revenue of £9,500 less £4,500 net product cost and £1,300 other variable costs leaves £3,700 before advertising. Deducting £2,000 leaves £1,700. For B, £8,000 less £4,000 product cost and £2,400 other variable costs leaves £1,600 before advertising and minus £400 afterwards.

The refund amounts equal 5% and 20% of initial order revenue respectively. Those are refund-value ratios, not order return rates: the table contains no count of orders or returned units from which to calculate those rates.

Combining the two groups produces £20,000 reported value against £4,000 spend, still ROAS 5. Their combined contribution after advertising is £1,300. That positive total hides the negative £400 from B, while still saying nothing about whether the business as a whole made a profit.

The product-cost reversals are already included in the net product-cost line. Do not add £250 or £1,000 again at the end. The table shows those intermediate lines to make the treatment auditable, not to create extra deductions or credits.

Show What Missing Costs Could Change

Pause the spending recommendation when uncertainty can change the conclusion. Suppose B’s net product cost has not yet been confirmed and could plausibly be between £3,600 and £4,400 under a documented estimate. Hold every other input constant and show the effect.

Hypothetical sensitivity to B’s unresolved net product cost
Assumed net product cost Contribution before advertising Contribution after £2,000 advertising
£3,600 £2,000 £0
£4,000 £1,600 −£400
£4,400 £1,200 −£800

This range is a scenario, not a statistical confidence interval. It shows that even the lowest assumed product cost only covers the included costs, leaving nothing towards excluded overheads. The next action is to confirm the cost and investigate the economics, not announce an exact company loss.

If the costs are completely unknown, do not invent a plausible range to make the report look finished. Show the affected revenue and orders as unresolved. An incomplete group should not be blended into a supposedly exact total with its costs set to zero.

Report both order coverage and revenue coverage where useful. A small number of high-value orders without costs may matter more than many low-value orders with complete records. Give the reviewer the missing-data exposure rather than only one reassuring percentage.

Distinguish Cost Attribution from Advertising Incrementality

The bridge describes the economics assigned to a defined set of orders and advertising costs. It does not prove all of those orders would disappear without advertising. Platform attribution is not the same as a controlled estimate of incremental sales.

Real campaigns can advertise one product and generate a basket containing another. Do not automatically assign the entire basket’s contribution to the advertised item’s group without documenting that analytical choice. Keep purchased-product economics separate from campaign attribution when the data cannot support a clean join.

If spend can only be established at campaign level, calculate contribution at a compatible level or clearly label any allocation. Avoid creating precise product-level advertising costs by dividing spend arbitrarily and presenting the result as observed fact.

The example’s £1,600 contribution before advertising for B is the amount available to cover advertising within that fixed historical bridge. It does not mean that reducing future spend to £1,600 will preserve the same orders or make the next period break even. Volume, product mix and acquisition costs may change.

Use Platform Reporting as Supporting Evidence

Merchant Center’s cost_of_goods_sold attribute can support additional estimated gross-profit reporting when the required basket and item-level conversion information is provided. Google’s documentation also allows approximate COGS for that reporting. Label that estimate appropriately; it is not a substitute for a complete contribution bridge.

GA4’s e-commerce documentation provides a refund event with a transaction reference and item information. A documented feature is not proof your store sends complete refund data. Verify the implementation against known records before relying on an analytics refund figure.

Keep the order system and agreed cost records as the evidence for the management analysis. Then explain how the advertising and analytics reports relate to it. A neat dashboard match is less important than knowing which orders, values and costs each report includes.

Do not change conversion-value logic or bidding targets merely to make a report look closer to contribution. Such changes affect measurement and campaign operation and need their own documented design, validation and review.

Turn the Bridge into a Review Decision

Record the product group, cohort dates, maturity cut-off, revenue definition, cost coverage and advertising assignment alongside the result. Add the principal uncertainty and the action that could resolve it. This makes the analysis usable in a commercial conversation.

Completed decision record for the hypothetical groups
Group What the evidence supports Next action
A Positive £1,700 contribution after included costs and advertising Check excluded costs, capacity and stability before considering greater spend
B Negative £400 on the base case; £0 to −£800 in the stated cost scenario Confirm product cost and review refunds, fulfilment and offer economics before scaling
Any unresolved group No reliable complete contribution figure yet Assign a cost-data owner and keep the missing exposure visible

Investigate the source of the difference before blaming advertising alone. An expensive delivery promise, unsuitable product expectation or high refund exposure may need an operational or merchandising response. Changing a bid cannot directly repair those underlying issues.

Repeat the bridge with consistent definitions and suitable cohorts. Track whether the improvement came from revenue, retained sales, product cost, fulfilment or advertising rather than attributing every movement to the campaign manager.

How to Get This Done

Bring order-line exports, refunds, dated product costs, fulfilment and payment records, campaign spend and the current conversion-value definition. Ask for one reconciled product-group example before commissioning a large dashboard.

KAP SEO Services can support Google Ads measurement and campaign review so commercial decisions are grounded in the available evidence. The finance owner should agree revenue and cost treatment, while operations confirms delivery and return records. Keep missing inputs visible in the handover.

Summary

High ROAS can coexist with negative contribution because reported conversion value is not the amount left after costs. Reconcile retained revenue, product costs and variable order expenses, then deduct advertising once. Compare groups on a consistent basis and keep uncertainty, overheads and causal claims separate.

Frequently Asked Questions

Can identical ROAS produce different contribution?

Yes. Product costs, refunds, delivery and other included expenses can differ even when reported conversion value and advertising spend are identical.

Is contribution after advertising the same as net profit?

No. It excludes whatever fixed overheads and other costs are outside the stated model. Label those exclusions and use the appropriate financial records for company profit.

Should missing product costs be entered as zero?

No. Mark them unresolved and show the affected orders or revenue. Use a sensitivity range only when there is a defensible basis for the assumptions.

Should refunds be deducted from the purchase month or refund month?

Define the purpose of the analysis. A purchase-cohort view follows later refunds back to its orders; a period cash or accounting view answers a different question. Do not mix the two silently.

What ROAS target should every retailer use?

There is no universal target. Revenue definitions, product economics, returns, attribution and business requirements differ. Establish those inputs before judging a target against contribution.

Review Your Advertising Economics

Connect campaign measurement with the product costs, refunds and fulfilment records behind your commercial decisions.