The usual way to judge a category is by its total: monthly units, market size.
A total tells you the ceiling. It cannot answer the actual question — whether there is still room for you underneath it. The total is also the shakiest number in the set, for reasons covered in where the market size number comes from.
A category doing a million units a month, where 80% of those units sit with listings over three years old, is harder for a new product than one doing a hundred thousand with a loose structure. The total cannot show that difference. The distributions can.
Start with the statistics call to size the room
The market statistics endpoint returns base metrics for a category node:
| Field | Meaning |
|---|---|
totalProducts | Total products |
brands | Number of brands |
sellers | Number of sellers |
nodeLabelPath | Node name path |
currency | The marketplace's currency |
brands and sellers against totalProducts is your concentration measure. Five thousand products across four brands and five thousand across four hundred are completely different markets — an oligopoly versus a fragmented one, with completely different entry strategies.
Note that products and totalProducts are two separate fields: the first is the sample size, the second the total. Keep them straight when computing ratios; it is the easiest thing to misread on this endpoint.
Distribution one: listing age, answering "can a new product get in"
Of the four, read this one first.
The listing age endpoint groups by how long listings have been up, returning label (such as "over 3 years"), products, units, revenue and unitsRatio.
unitsRatio is the field that matters. The documented example for the "over 3 years" band is 0.834 — 83.4% of that category's units go to listings older than three years.
That number answers a new product's situation directly:
- Old listings hold a large share → positions are locked in, and a new listing needs a long time to move anything
- Listings under a year hold a meaningful share → the structure is still fluid and there is an opening
- The share spreads across several age bands → new entrants keep succeeding, which is the healthiest signal
A large total with a small new-listing share is the classic trap — the data looks inviting, and those units were never available to you.
To see what getting in looks like in practice, our monthly rising new products ranking lists the US listings under six months old that sold the most over the last 30 days. The latest issue's top ten are below; the ranking page splits them by top-level category, so you can check whether new listings are breaking through in yours. How sellers based in China fare in that ranking is broken down in Chinese sellers on Amazon.
| # | Product | Units (30d) | Price | Ratings | Growth |
|---|---|---|---|---|---|
| 1 | FNTCASE for Google Pixel 11-Pro Case: Pixel 11 Case Compatible with Magsafe FNTCASE · Cell Phones & Accessories | 170,340 | $14.99 | 27 | +7% |
| 2 | FNTCASE for iPhone 17e/ iPhone 16e Case, Fit for Magsafe, Screen Protector FNTCASE · Cell Phones & Accessories | 152,517 | $11.99 | 5 | +3% |
| 3 | Coca-Cola Bottles, 16.9 fl oz, 15 Pack Coca-Cola · Grocery & Gourmet Food | 127,942 | $16.43 | 537 | New listing |
| 4 | FNTCASE Magnetic for Motorola Razr Ultra 2026/2025 Case, Clear FNTCASE · Cell Phones & Accessories | 120,694 | $15.99 | 32 | +11% |
| 5 | INABA Salmon Churu Cat Treats, Lickable, Squeezable Creamy Purée Cat Treat with Green Tea Extract & Taurine, 0.5 Ounces Each Tube, Salmon Variety Box, 1 Count (Pack of 480) INABA · Pet Supplies | 82,483 | $32.96 | 2,405 | -10% |
| 6 | Apple AirPods 5 Wireless Earbuds with Active Noise Cancellation Apple | 80,092 | $129.00 | 231 | 0% |
| 7 | Valvoline Advanced Full Synthetic SAE 5W-30 Motor Oil 1 QT (Pack of 6) Valvoline · Automotive | 56,406 | $53.82 | 9 | — |
| 8 | medicube PDRN Collagen Gua Sha Neck Wrinkle Cream - Firming, Contour Care medicube · Beauty & Personal Care | 56,210 | $29.90 | 1,181 | -21% |
| 9 | Clorox Disinfecting Wipes Variety Bundle, 2 Crisp Lemon & 2 Cherry Blossom, 300 Total Wipes Clorox · Health & Household | 54,116 | $22.08 | 3 | +32% |
| 10 | MiraFIBER Gummies, from The Makers of MiraLAX, Vegan Formula, 8g Per Serving, Gut Health for Men & Women, Prebiotic Fiber Gummies for Adults, Mixed Fruit Flavor 72Ct MiraLAX · Health & Household | 53,933 | $15.97 | 779 | +16% |
Sales, revenue and search volumes are third-party estimates, not figures published by Amazon.
View rankingDistribution two: seller type, answering "who am I competing with"
The seller type endpoint returns the mix by fulfilment and seller type, each group carrying label (such as "Amazon retail"), asinNum, asinRatio, units, unitsRatio, ratings and rating.
The pair to compare is asinRatio against unitsRatio:
| Pattern | Meaning |
|---|---|
Low asinRatio, high unitsRatio | This group takes a lot of volume with few products — efficient and hard to displace |
High asinRatio, low unitsRatio | Many products that do not sell; this group is making up the numbers |
A category with a high retail share deserves its own consideration. It does not necessarily mean staying out, but it does raise the bar on price and delivery.
rating and ratings come back here too, so you can read the rating level of each seller group in passing — a group with a visibly low average rating is a quality opening.
Distribution three: price, answering "at what price point"
The price distribution endpoint groups by price band, each returning label (such as "5-10"), products, units, revenue and unitsRatio.
What you are looking for is a mismatch between products and unitsRatio.
- A band with many products and a low share of units → crowded with products that do not sell, a dead zone inside a red ocean
- A band with few products and a high share of units → demand concentrates here and supply has not caught up
The second is the position you want. It is not necessarily the highest or lowest band — it is often an unremarkable one in the middle.
Market statistics endpointCore statistics per category node: total products, brand count, seller count and the node pathFour more cuts in the same shape
Four further endpoints return the same shape as the distributions above: one label per group, with products, units and a share of units. Each answers a narrower question.
| Question | Endpoint | Grouped by |
|---|---|---|
| Where are the competing sellers from? | Seller country distribution | Seller country |
| How good are the products? | Rating distribution | Rating band |
| How much social proof do sales need? | Ratings count distribution | Rating-count band |
| How complete are the listings? | Listing content distribution | A+ content and video |
The documented examples appear to describe one market: the unit totals implied by their shares agree at about 5,238 (4,496 ÷ 0.8585, 548 ÷ 0.1046 and 3,599 ÷ 0.6872), so they can be read together.
- Seller country: the 44 products from China sell 4,496 units, 85.85% of the total. The 24 from the US sell 362, 6.91%.
- Ratings count: the 36 products with 1–50 ratings still take 10.46% of units, so new listings are getting some sales here.
- Rating: the 14 products in the lowest rating band (2.0 or below) sell 5 units between them, 0.1% of the total. A group with a low average rating is a quality gap; it is not a cheap way in.
- Content: products with A+ content and video are 18.18% of products and 68.72% of units, so content is doing a lot of the work in this market.
Two traps. The share of units is a fraction in three of these endpoints and a percentage in the last: unitsRatio is 0.8585 for seller country and 68.72 for listing content, and productsRatio is 18.18. Normalise before comparing bands across endpoints. And the labels in the documented examples are Chinese strings such as 4.5以上 and 有A+有视频, so match on the label you receive instead of typing your own.
Reading the four together
Any one distribution alone will mislead. The conclusions come from combinations:
| Combination | Reading |
|---|---|
| Low old-listing share + fragmented brands + an underserved price band | Worth entering |
| High old-listing share + concentrated brands | Locked in; stay out without clear differentiation |
| High new-listing share + generally low ratings | Demand exists and nobody executes well — a quality opening |
| Large total + old listings taking everything | The classic trap: good numbers, no way in |
The third row is the most frequently missed opportunity. A flood of new entrants means demand has been validated, and generally low ratings mean nobody has done it well yet. That category suits a seller with supply chain capability.
Then filter for products
Category-level judgment comes first; filtering products inside it comes second. The product research endpoint exposes close to twenty min/max conditions, so the conclusions above translate straight into filters — if the age distribution says avoid established listings, gate on availableMonth; if the price distribution gives you a target band, gate on minPrice and maxPrice.
The full three-step pipeline is in Bulk product research, and getting candidates into a comparable sheet is in The Amazon product research sheet.
Three things this data cannot do
Everything sales-related is an estimate. The units, revenue and unitsRatio in these distributions all rest on sales estimates, and the marketplace does not publish unit sales. Comparing relative structure with them is sound; treating them as absolutes is not. What Amazon sales data can and cannot tell you covers that layer.
It cannot tell you whether your supply chain can do it. The data says a position is open; it does not say you can fill it. Whether you can source it or whether there is patent exposure is not something the endpoint knows.
You have to choose the category level. Top-level and sub-level distributions can differ completely, and the same judgment can invert between a first-level node and a third-level one. Record nodeIdPath rather than writing "this category" loosely.
Questions
Which category level should I read? The level you will actually compete in — usually the subcategory your product gets classified into, not the parent above it. A parent's distribution is diluted by unrelated subcategories.
What new-listing share counts as "can get in"? There is no universal threshold, because categories vary widely. What works is comparing across: line the same figure up for several candidate categories and rank them, rather than staring at one absolute number.
How often does distribution data update? Go by the computation time the endpoint returns rather than assuming it is live. When judging trends, hold one basis and do not mix samples from different points in time.
Can this data judge seasonality? The age distribution cannot — it describes the standing structure. For seasonality use the time series from the market trend endpoints; the full method is in seasonal product research.