The short answer
Keep immediate order demand available for day-to-day steering, and estimate eventual net sales for orders still inside the return window. Deduct returns already observed plus expected remaining returns, then replace estimates with actuals as each order cohort matures. Validate forecasts by market and product mix before using them to shift budgets.
This article is part of Asked by Marketers, a series answering real questions from marketing leaders.
Sellforte's team holds more than 1,450 meetings each year with marketing leaders in Ecommerce and Retail about Marketing Mix Modeling and incrementality testing. Each week, we anonymize at least one question from those conversations and answer it in depth, based on what marketers are actually struggling with, not what keyword tools suggest. About the series
Why marketers ask this
Last week's campaign has brought in orders, and the team needs to decide whether to keep spending. Some of those purchases will come back. Finance will judge the revenue left after returns, but waiting for every return to arrive would leave marketing making decisions with old information.
The gap matters most when return rates differ across markets or product categories. A market can look attractive on order value while leaving much less revenue to cover product costs, fulfillment, and advertising. Recent sales also look unusually strong if they are compared with older sales that have already absorbed their returns.
Which sales measure should MMM use while returns are still pending?
Keep both original order demand and expected sales after returns in your reporting. Agree which is the primary outcome for your Marketing Mix Modeling analysis and which will guide the budget decision. Make that choice explicit in the results and optimization settings.
Order demand gives you a prompt view of purchasing activity. Define it consistently, including how discounts, taxes, and cancellations are treated. It can support daily or weekly campaign monitoring, but it does not tell you how much revenue the business will retain.
For recent orders, expected net sales combines what you already know with an estimate of returns still to come. Older orders provide the actual net-sales history once returns and their processing are sufficiently complete. Keep the estimate identifiable so a marketer can see which periods may still change.
A return forecast estimates how much of the order value will remain. MMM estimates how much sales activity marketing caused. Applying a return adjustment to attributed sales does not, by itself, make those sales incremental.
How should you estimate returns and replace them with actuals?
Start with completed groups of orders placed in the same period, often called order cohorts. Use their return history to estimate the additional returned value still expected from recent orders, given how old those orders are and what has already been returned.
For a consistently defined order cohort, the calculation is:
Expected net sales = original order value − actual returned value to date − expected additional returned value.
In this calculation, original order value is after discounts, excludes tax, and already excludes known cancellations. Handle later cancellations separately and consistently. Use monetary return values or rates weighted by order value; the share of items returned can differ from the share of revenue returned.
If a forecast already estimates total eventual returns, subtract that total once. Do not subtract actual returns again. If the data already supplies net sales after observed returns, deduct only the remaining expected returns from that figure. Document which version each field contains so you can check that every return is deducted once.
Start with forecasts by country and product category where you have enough mature history. Add finer segments only when they improve forecasts on unseen cohorts. An order-line prediction may be useful, but a simpler estimate from comparable historical orders can be a reasonable starting point.
Choose the maturity period from the actual return and processing pattern. A stated return policy does not guarantee that every refund is recorded on its last eligible day. Account for fulfillment timing, processing delays, and extensions to the normal policy. There is no universal number of days that makes every retailer's sales final.
Each refresh should revisit the affected order history, updating actual returns and the estimate of what remains. For an MMM measured by order date, attach the adjustment to the original order cohort. Keep a separate reconciliation to reports that record refunds on their processing date. Appending only the newest week of orders will miss later changes to earlier orders.
Why can recent performance look better than it really is?
A change in data treatment at the edge of the return window can create an apparent performance change. Check that the recent estimates and mature actuals use the same revenue definition, including the treatment of discounts.
For example, suppose a promotion's discount cost has already been reduced for expected returns, while its associated sales still include the full order value. The ratio of sales to discount cost will look too high. It may then fall as the sales are adjusted, even though the campaign itself has not changed. Inspect both sides of the calculation before interpreting that fall as weaker marketing.
Costs also need their own treatment. Some fulfillment or return-handling costs remain even when the customer sends the entire order back. Multiplying every revenue and cost field by the same retention percentage can therefore misstate profitability. Agree with finance which costs reverse and which remain, and reconcile the resulting figures with the business's reporting definitions.
Check forecast accuracy by saving what you predicted for each cohort and comparing it with the eventual result. Recreate historical forecasts using only the information available at the time. Look for persistent overstatement by market or product category, and check whether changes in the product mix or return policy explain it. A good overall average can hide an error in the market where you are about to increase spend.
How should returns affect budget allocation?
Use a comparable sales-after-returns or contribution-margin measure when moving budgets between markets with different economics. A high return on gross order demand may leave less retained revenue than a lower gross return in another market. Sellforte's guide to optimizing ecommerce gross profits by market explains how revenue adjustments and costs enter that comparison.
If orders associated with different channels have different product or customer mixes, one business-wide return rate may hide a meaningful difference. A reliable channel adjustment needs enough mature data and a clear connection to the sales outcome being modeled. The mix of purchases credited to a channel is not automatically the mix of purchases it caused.
Before making a large budget change, compare the recommendation under plausible higher and lower return assumptions. If the preferred market or channel changes, use tighter spending limits while the estimates mature. Return-adjusted historical ROAS alone does not establish how much additional budget a channel can absorb.
How this looks in practice
Consider a hypothetical retailer tracking one cohort with $100,000 of order value after discounts, excluding tax and known cancellations. The figures below are illustrative snapshots of the same orders. They show revenue retained after returns, not incremental revenue attributed to marketing.
| Measure | First estimate | Later update | Mature actuals |
|---|---|---|---|
| Original order value | $100,000 | $100,000 | $100,000 |
| Actual returned value to date | $10,000 | $24,000 | $32,000 |
| Expected additional returned value | $20,000 | $7,000 | $0 |
| Net sales after returns | $70,000 estimated | $69,000 estimated | $68,000 actual |
At the first snapshot, subtracting only the returns already received would report $90,000. That would overlook the further $20,000 expected to come back. The working estimate is $70,000, and both the actual and remaining-return amounts change as the cohort develops.
When the cohort matures, net sales are $68,000. The first estimate overstated retained revenue by $2,000. The team should record that forecast error and check whether similar cohorts show the same pattern before relying on the next estimate for a budget increase.
Keep those historical forecast snapshots alongside the refreshed MMM input. The snapshots show what the team knew when it made the decision; the updated input shows the best current estimate of the cohort's outcome.
Related questions
What if we cannot forecast returns reliably yet?
Use recent order demand for timely monitoring and mature net sales for financial comparisons, with the different reporting periods clearly labeled. Test a simple forecast on completed cohorts before introducing it into current budget decisions. Until it is reliable enough for the decision, limit reallocations that depend heavily on an uncertain return adjustment.
Does net-sales ROAS tell us whether more media spend is profitable?
No. Net-sales ROAS accounts for returned revenue, but profitability also depends on product margin and relevant costs. For an additional spending decision, assess the expected incremental contribution from that increase against its media cost. See How much should I invest in media overall? for the budget-threshold decision.
How Sellforte helps
Sellforte can work with demand and sales-after-returns measures, using the sales data and return assumptions agreed with your team. Its MMM and budget-planning tools help compare marketing outcomes, while the data setup must keep estimates, actuals, and the selected optimization target consistent. Book a demo.
Authors
Lauri Potka is the Chief Operating Officer at Sellforte and has over 15 years of experience in Marketing Mix Modeling, marketing measurement, and media spend optimization. Before joining Sellforte, he worked as a management consultant at the Boston Consulting Group, advising some of the world's largest advertisers on data-driven marketing optimization. Follow Lauri on LinkedIn, where he is one of the leading voices in MMM and marketing measurement.