Use the smallest clean sample that has stabilized
There is no universal minimum because product frequency and variation are different. A rare record and a current phone should not use the same threshold. Start with the closest matches, calculate the median and range, then add more results and see whether the conclusion changes.
| Clean sample | How to treat it | What to check |
|---|---|---|
| 1 to 4 clean sales | Directional evidence only | Show the individual comps. Avoid presenting one precise market value. |
| 5 to 9 clean sales | Useful for rare products | Use a range and explain differences in condition, variant, or completeness. |
| 10 to 19 clean sales | A workable early sample | Check whether one cluster, lot, or condition dominates the result. |
| 20 to 39 clean sales | A practical baseline | Usually enough to inspect a median and p25 to p75 range for a specific product. |
| 40 or more clean sales | Helpful for variable markets | Use more evidence when prices vary by condition, bundle, capacity, grade, or region. |
Relevance, recency, and comparability decide whether the sample is useful
Match the product identity
Keep the same model, variant, capacity, card number, grade, format, or other required identifier.
Separate condition and completeness
A sealed product, used product, parts-only item, loose game, and complete-in-box game are different markets.
Use a recent, local market
Choose the eBay marketplace and recent date window that match the valuation decision you are making.
CompSniper's published 100-product study observed 19,220 raw priced rows and 11,942 rows after relevance cleaning. The median moved by at least 10 percent in 34 of the 100 fixed searches. That study measures the effect of cleaning, not a universal accuracy guarantee, but it demonstrates why raw row count alone is a weak confidence measure. Read the complete method and limitations.
Stop when the valuation stops moving materially
- 1
Define the item
Write down the model, variant, condition, completeness, and marketplace before searching.
- 2
Inspect the first clean set
Remove clear conflicts and record the median, p25 to p75 range, and sample size.
- 3
Add another page or broader clean set
Recalculate using more genuinely comparable sales, not merely more rows.
- 4
Compare the answer
If the median and typical range stay close enough for the pricing decision, the sample is adequate. If they move materially, inspect the mix and keep collecting.
Calculate a transparent median
function median(values) {
const sorted = values
.filter(Number.isFinite)
.sort((a, b) => a - b);
if (sorted.length === 0) return null;
const middle = Math.floor(sorted.length / 2);
return sorted.length % 2
? sorted[middle]
: (sorted[middle - 1] + sorted[middle]) / 2;
}
const prices = response.items
.filter((item) => !item.bestOfferAccepted)
.map((item) => Number(item.soldPrice));
console.log({ sampleSize: prices.length, median: median(prices) });The example excludes accepted Best Offer markers because the exact accepted amount is not publicly disclosed. It still requires a relevant set of listing rows before calculation.
This is practical valuation guidance, not a claim that one sample size guarantees accuracy. The available sold window and fields depend on the source. See eBay's official Marketplace Insights search documentation, the CompSniper methodology, and the 100-product study for collection and cleaning details.
Sold-comp sample size FAQ
Are five eBay sold comps enough?+
Five genuine comparable sales can be useful for a rare item, but they should be shown as a small evidence set rather than a definitive market price. Condition, variant, completeness, and accepted Best Offer markers matter more when the sample is small.
Is a larger sold-comp sample always better?+
No. Forty mixed listings can be less useful than ten close matches. A sample should contain the same product or a deliberately defined comparable group. Accessories, parts, lots, wrong models, and different grades can move the median away from the product being valued.
Should Best Offer accepted sales be included?+
Exclude them when you require exact transaction prices. eBay can mark that a Best Offer was accepted without disclosing the accepted amount, so the visible crossed-out price is not reliable evidence of what the buyer actually paid.