A return policy is several decisions packaged as one promise. More days, a fuller refund and an easier process can affect customers differently.
Choose the component that addresses the buying risk, then check whether retained contribution and customer trust justify its cost.
🔎 What research says
Separate the policy levers, preserve the option customers value, and measure sales and return economics together. A tighter deadline can reduce returns while leaving the business worse off.
🗺️ In this guide
1. Separate refund value from return hassle
📈 Recommendation
Choose the specific promise you are changing: days allowed, refund amount, effort, eligible items, or cash versus exchange. For a purchase-confidence problem, start by testing the refund amount and the work needed to return.
Show all five terms together before checkout. A generous deadline does not compensate automatically for a costly or difficult return.
Change one component at a time where feasible. Track paid purchases, retained revenue, return cost and complaints.
🎓 Findings
Janakiraman, Syrdal and Freling (2016) pooled 21 papers. The association between leniency and purchase-related outcomes was 0.217, with a 95% interval from 0.175 to 0.261. The corresponding return association was 0.080, with a 95% interval from 0.021 to 0.137. These are small positive correlations on a scale from −1 to 1, not conversion lifts.
Their component analysis linked more generous refunds and less effort to stronger purchase outcomes. The other purchase components were not clearly different from zero in that model.
🧠 Why it works
A shopper evaluates the possible loss and work if an item disappoints. Reducing those obstacles can make an uncertain purchase easier to consider.
✋ Limitations
The studies mix intentions and behavior, and vary substantially. The purchase and return correlations cannot be subtracted to calculate profit. The component results compare studies; they do not identify one universally optimal policy.
2. Protect the return option where fit is uncertain
📈 Recommendation
Evaluate categories separately before tightening a sitewide policy. Shoes or unfamiliar fits may depend more on a return option than predictable repeat purchases.
Compare contribution after refunds, handling, shipping and resale loss within each category.
Pair policy tests with better sizing or fit information. Treat category-level thresholds as hypotheses to test.
🎓 Findings
In a historical catalog-retailer model, Anderson, Hansen and Simester (2009) estimated the return option at $15.81 for women’s footwear versus $3.19 for men’s tops. These are inferred customer values, not fees people paid. The model used 987 customers over 126 months.
For footwear, allowing returns added 3 percentage points to the modeled monthly probability of ordering and keeping an item: roughly 5.7% without returns versus 8.7% with them. That is the source of the paper’s roughly 53% relative increase; it is a model comparison with a return ban.
🧠 Why it works
When customers cannot establish fit before delivery, the ability to reverse a mistake has value. Removing it can reduce retained demand as well as returns.
✋ Limitations
The retailer did not randomly ban returns. Estimates depend on the structural model and historical customers, and Table 4 does not report an interval for these contrasts. They do not justify a $15.81 fee, a current forecast, or different treatment of individual customers.
3. Judge shorter windows by channel and profit
📈 Recommendation
Before shortening a deadline, inspect when customers actually return and whether behavior differs online and in stores. Judge a change on retained contribution, not the return-rate dashboard alone.
Attribute later refunds to the original purchase cohort. Follow cohorts long enough to see returns and resale outcomes.
Check sales, refunded value, recovery value and support cost by channel. Look for earlier returns replacing later ones.
🎓 Findings
The public author manuscript describes a jewelry retailer shortening its window from 100 to 60 days in 2013. Ertekin and Agrawal (2021; online 2020) compared affected brands with an unchanged 30-day control using about 12.5 million transaction records. Store sales fell by about 8% and the share of sales value returned fell by 2.7 percentage points. Their cost-based calculation estimated 7.3% lower store profit.
For the four online stores, the sales and return-rate changes were not statistically clear. The analysis does not establish that the same restriction works equally across channels.
🧠 Why it works
A shorter window can prevent refunds, discourage purchases, or simply bring returns forward. Lower return expense may be outweighed by lost contribution.
✋ Limitations
This was one nonrandom change, not a randomized deadline test. The return rate measures refunded dollar value, not the percentage of orders. Profit was calculated using estimated margins and return costs. Jewelry resale economics and the small online comparison limit transfer.
4. Explain a restriction as part of the policy change
📈 Recommendation
If the business case supports a tighter policy, explain the operational reason alongside the new terms. Plan the customer response as part of the policy decision.
State what changes, which purchases it affects and the real reason. Give support teams the same explanation.
Test understanding and trust as well as contribution. Do not assume an explanation removes the commercial cost of a restriction.
🎓 Findings
In the accepted author manuscript of Abdulla and colleagues (2025; online 2024), a simulated retailer replaced an open-ended policy with 30 days. In the severe-change groups, post-change trust averaged 5.72 with an explanation versus 5.06 without, on a seven-point scale: a descriptive 0.66-point gap. Those groups contained 84 and 86 people.
The analysis retained 538 of 640 participants after attention checks. The paper’s adjusted mediation analysis supported a smaller negative reaction with explanation for severe and medium restrictions, but no clear moderation for the mildest restriction.
🧠 Why it works
A restriction can look like a broken promise. An explanation gives customers information about why the offer changed; that may reduce the damage to trust.
✋ Limitations
People judged a hypothetical brand and reported trust and intentions; they did not buy. The 0.66 gap is descriptive, without a reported confidence interval for that raw contrast, and is not a sales percentage. The cited public manuscript may differ slightly from the final typeset paper.
💡 The takeaway
Make the return promise precise, and optimize retained contribution rather than return counts.
Start with one category and a defined policy lever. Keep the original purchase cohort visible through refunds and resale, compare channels, and explain any restriction customers will experience.
📚 Sources and study types
Anderson, E. T., Hansen, K., & Simester, D. (2009). The Option Value of Returns: Theory and Empirical Evidence. Marketing Science, 28(3), 405–423. — Historical transaction model and counterfactual simulation.
Ertekin, N., & Agrawal, A. (2021; online 2020). How Does a Return Period Policy Change Affect Multichannel Retailer Profitability? Manufacturing & Service Operations Management, 23(1), 210–229. — Nonrandom retailer policy change and cost-based profit calculation.
Abdulla, H., Ketzenberg, M., Abbey, J. D., & Heim, G. R. (2025; online 2024). The point of no return? Restrictive changes to lenient return policies and consumer reactions to them. Journal of Operations Management, 71(1), 81–108. — Randomized simulated-retailer studies; accepted author manuscript reviewed.
