Statistics, lightly
Median, outliers and reading a price spread
The median sold price is the middle value of your sorted sold comps, and it is the number to price from because one odd sale barely moves it. In a set of 11 sales, ten between $40 and $55 and one at $400, the median is $47 and the average is $79. The average would leave the item unsold.
Pricing does not need statistics, but it needs three of its ideas. Each one takes a minute to understand and prevents a specific expensive mistake.
Why use the median instead of the average for pricing?
Sort your comps and take the middle value. That is the median. It answers the question you have, which is 'what does this typically sell for', and it is barely affected by extreme values.
An average is pulled hard by extremes. In a set of eleven sales, ten between $40 and $55 and one at $400, the average is $79 and the median is $47. The $79 is arithmetically correct and would leave the item unsold.
| Comps (sorted) | Average | Median |
|---|---|---|
| 40, 42, 44, 45, 47, 47, 49, 51, 53, 55, 400 | $79 | $47 |
| Same set with the 400 removed | $47 | $47 |
The median hardly notices the outlier. That stability is what you want from a pricing summary. eBay leans on the same number: its price guide for sports cards shows the median sold price, per eBay's pricing help page.
What should you do with outlier sold prices?
The instinct is to delete extremes. The better move is to open them first, because an outlier is usually a fact about a listing rather than an error.
- A high outlier may be a rare variant, a bundle, exceptional photographs, or a bidding war. If it is a variant, check whether yours is that variant. That is a considerably more valuable finding than a median.
- A low outlier may be damage, missing parts, a poorly written listing, or a seller who needed a fast sale. If it is a bad listing, it tells you there is room above the median for a good one.
Open the highest and lowest comp before trusting either. Thirty seconds each, and it is the highest-value thirty seconds in the whole pricing process.
What does a wide spread in sold prices mean?
The width of the middle of your data matters as much as its center. A useful way to read it is the range covering the middle half of your comps, ignoring the top and bottom quarters.
| Middle half spans | Means | What to do |
|---|---|---|
| A narrow band | A well-established price the market agrees on | Price near the median with confidence. Little upside in going high |
| A wide band | Presentation, condition or variant is driving price | Identify which. If it is presentation, photograph well and price high |
| A very wide band | Your comp set is probably mixed | Split it. You are likely looking at two products or two condition bands |
How do you apply this to your own sold comps?
Sort the list, look at the middle, open the ends, notice the width. That is the whole discipline. The comp analyzer does the sorting and flags the outliers so you can go straight to reading them.
How many sold comps do you need?
Enough that one unusual sale cannot move your answer much. Eight is a reasonable working minimum, fifteen is comfortable, and beyond about twenty you are usually adding older and less relevant sales rather than improving accuracy.
Below eight, treat the number as provisional and be more willing to adjust it after the market responds. That is the thin-comp situation.
Why use the median instead of the average for pricing?
What should I do with outlier sold prices?
What does a wide spread in sold prices mean?
How many sold comps should I collect?
Disclosure What's It Sold For comes from the team that builds FlowLister, an AI tool that turns your photos into a complete eBay listing and prices it from sold listings, with the sales behind the number shown. It's AI and it misses things, which is why every listing waits for your OK before it goes up on eBay. Read this site as a guide from a tool maker with a stake in the topic, not an independent review.
Keep reading
When to lower your price on eBay, and by how much
A schedule that avoids both panic discounting and listings that sit forever.
How to price an item with no sold comps
Proxy comps, floors, the descending-price method, and when to walk away.