A nearly perfect seller score can conceal a large difference in delivery quality. Buyers who never leave feedback disappear from the public percentage, even when their experience affects whether they return.
Show understandable seller history, but validate it against complete transaction outcomes. The purpose of reputation is to help customers find reliable sellers, not merely to decorate listings.
🔎 What research says
Preserve useful seller history and test whether it predicts fulfilled promises. A field experiment found value in an established identity, while richer transaction measures revealed information that public feedback missed. Validate ranking by later customer outcomes as well as immediate sales.
🗺️ In this guide
1. Show seller history with its transaction context
📈 Recommendation
Keep seller history visible, including the volume and period behind a score. Explain whether the history concerns the seller’s service or the product itself.
For a marketplace listing, separate seller dispatch/description feedback from product quality reviews.
Test the presentation while tracking disputes, fulfilment and repeat purchase. Do not promise a fixed price premium for earning a badge.
🎓 Findings
Resnick and colleagues (2006) randomized matched lots of vintage postcards between a dealer’s established eBay identity and new identities operated by the same dealer. They estimated 8.1% higher willingness to pay under the established identity. That means a relative auction-value premium, not 8.1 percentage points of conversion.
A smaller subsidiary experiment found no statistically clear price effect from one or two negative feedbacks on new identities. That does not mean complaints are harmless.
🧠 Why it works
A history gives buyers information about whether a seller keeps promises. The experiment held the operator constant, reducing the chance that the premium merely reflected better fulfilment by different people.
✋ Limitations
Historical collectible auctions do not supply a forecast for modern fixed-price retail. Identity history was a bundle of signals, not an isolated badge. The subsidiary null is not proof of no harm.
2. Validate seller ranking against complete transaction outcomes
📈 Recommendation
Check whether public seller scores predict disputes, delivery failures and later purchasing when transactions without feedback are included. Test quality-aware ranking before changing the whole marketplace.
Keep nonresponse separate from explicit negative feedback. Investigate its predictive information without calling silent buyers dissatisfied.
Measure immediate conversion and later repeat purchase by assigned ranking group. Audit exposure for newer and lower-volume sellers.
🎓 Findings
Nosko and Tadelis (2026) calculated seller quality using positive feedback divided by all transactions, including those with no feedback. In their randomized eBay ranking experiment, assigning users to a ranking that favored this measure increased purchasing again within 180 days by about 0.3 percentage points: roughly three more returning purchasers per 1,000 assigned users. These were actual platform purchases, not survey intentions.
The reported group rates were about 61.85% versus 61.55%. Adding controls reduced the estimated difference to about 0.1 percentage points, still statistically clear. The researchers found no corresponding reduction in immediate conversion. The ranking changed the whole search environment; it did not randomly assign a particular seller-quality score.
🧠 Why it works
Public feedback can miss customers who quietly stop buying. Complete transaction records help a platform test whether a popular score actually separates sellers’ performance.
✋ Limitations
These were historical eBay users. Exact group confidence intervals were not recovered in this extraction. The smaller adjusted estimate shows that the headline result is not a universal forecast. Nonresponse contributes predictive information but does not itself prove dissatisfaction; changes in ranking may affect more than seller quality.
💡 The takeaway
A seller score is useful when it helps buyers find reliable service.
Show understandable history and validate it against complete transaction outcomes. Test ranking changes through both immediate orders and later customer retention.
📚 Sources and study types
Paul Resnick et al. (2006). The value of reputation on eBay: A controlled experiment. Experimental Economics,9,79–101. Randomized matched postcard-lot auctions under established versus new seller identities operated by same dealer.
Chris Nosko et al. (2026). The limits of reputation in platform markets: An empirical analysis and field experiment. Quantitative Marketing and Economics,24,9. eBay transaction cohort and randomized search-ranking field experiment.
