datawebotEcommerce market research: how to use marketplace data
How to use marketplace listings for ecommerce market research: what the data can answer, a step-by-step method, the checks to run, and its honest limits.
Ecommerce market research uses the product listings on online marketplaces, with their prices, sold counts, brands and sellers, to estimate how big a category is and who is winning in it. It works because marketplaces publish this information on every listing, so you can build a market estimate from real transactions instead of from a survey.
The data is not perfect. Sold counts are reported by the platform, not audited, and a few bad records can move a total a long way. This guide covers what marketplace data can answer, a method you can repeat, the checks to run before you trust a number, and the limits you should state next to any figure you publish.
What questions can marketplace data answer?
Marketplace data answers questions about what is listed, at what price, and roughly how much of it sells. The table below lists the common questions and the fields each one needs.
| Question | What you measure | Fields you need |
|---|---|---|
| How big is the category? | Estimated monthly sales value (GMV) | Price, sold count per period |
| Which brands lead? | Brand share of category GMV | Brand, price, sold count |
| What does the price ladder look like? | Share of sales in each price band | Price per listing, sold count |
| How wide is the assortment? | Distinct products, sizes and variants per brand | Product ID, variant, title |
| Who is new? | Brands or listings with sales this period and none before | Brand, first-seen date |
| Who sells it? | Share held by official brand stores vs resellers | Seller name, store type |
| How promotional is the category? | Share of sales at a discount, average discount depth | List price, sale price |
None of these answers is exact. Each one is an estimate built from what the platform shows. Used carefully, it is still one of the fastest online market research methods available, because the data refreshes every month and covers every brand that sells, not only the brands that report to a research panel.
How do you do ecommerce market research step by step?
The method has seven steps. Skipping the last one is the most common cause of a wrong number.
1. Define the category
Write down what is in and what is out before you collect anything. "Body lotion" sounds clear until you meet body serums, hand creams, sunscreen lotions and gift sets that contain a lotion. Record the market, the platforms, the period and the exclusions. This record is your sampling frame. A reader cannot judge your number without it.
2. Collect the listings
Gather every listing in the category on each platform for the period. For each listing keep the product ID, title, brand, seller, price, list price, sold count, review count, variant and the date you retrieved it. Keep the retrieval date with every record. It tells you which period a sold count belongs to.
3. Estimate units from sold counts, and know their limits
Most marketplaces show a "sold" figure on each listing. Some show sales over a recent window. Some show a lifetime total. Some round the figure into buckets such as "1k+ sold". To get a monthly figure from a lifetime count, take the difference between two readings a month apart. Monthly sales value is then units multiplied by the price paid.
Two cautions apply. First, a sold count usually counts orders, not pieces, so a 3-pack sold once is one order. Second, the price you record may belong to one variant of a listing that has several sizes, so the multiplication can overstate or understate value.
4. Deduplicate and match SKUs
The same product is often listed by several sellers, and the same seller sometimes lists it twice. Remove duplicate listings within a platform. Then match identical products across platforms using brand, product name, size and pack count. Match only when the evidence is clear. A wrong match is worse than an unmatched record. Our guide to building a product price history database covers matching in more detail.
5. Map brands
Brand fields on marketplaces are filled in by sellers, so they are inconsistent. One brand can appear as three spellings, and a reseller can put a brand name at the start of a title for a different product. Build a brand map that merges spellings, and check any brand whose sales sit mostly in one non-official store.
6. Compute shares
Divide each brand's sales value by the category total. Keep unbranded and multi-brand listings in the total so the shares are honest, even if you leave them out of the ranking. Use the same approach for price bands and seller types.
Look at price bands two ways: share of sales value and share of orders. They often tell different stories, because many orders can sit in the cheapest bands while most of the money is spent in the higher ones. The price you record is per order, not per piece, so multipacks and bundles push orders into higher bands. A per-piece price ladder needs pack sizes taken from the title or variant first.
7. Sanity-check before you publish
Run these checks every time:
- Coverage: compare the number of listings and sellers with the previous three months. A sudden drop or jump of more than about 15% is a warning sign. If the number of listings falls but sales value holds, inactive listings were probably removed and shares are still usable. If both fall, part of the market is missing.
- Concentration: list the top sellers. If one store holds a large share of the category, check whether its numbers are real before you use the month.
- Price outliers: compare each brand's average price per order across months and across platforms. A brand whose price jumps several times over in one month usually has a pricing error, not a real price change.
- Stability: look at two or three months, not one. A brand that leads for one month only is often an error.
What are the limits of marketplace data for market research?
Marketplace data for market research has known weak points. State them next to your numbers. A reader who sees the limits is more likely to trust the figures.
| Limit | What happens | What to do |
|---|---|---|
| Platform-reported sold counts | Counts are not audited. Some are rounded into buckets, some are lifetime totals | Use differences between readings. Treat bucketed counts as ranges |
| Coverage gaps | Not every platform, seller or listing is captured every month | Compare listing and seller counts month over month. Name the platforms in scope |
| One-merchant distortions | One store with a pricing error or an unusual promotion can move a category total | Check the top sellers and any single listing above a few percent of the category |
| Bundles and multipacks | One order can contain several pieces or several products | Call units "orders". Parse pack size before computing per-piece prices |
| Placeholder prices | Some listings carry a dummy price, such as a very high price on a cheap sachet | Compare each brand's price across months and platforms. Drop listings that do not fit |
| Offline sales | Marketplace data says nothing about supermarkets or a brand's own website | Call the result "marketplace GMV", not "market size", unless you add other sources |
As our shorter guide to ecommerce data for market research puts it, a small extraction sample does not establish market share, sales volume or demand. Shares need the whole category for the period, not a sample of listings.
How does datawebot fit into this?
datawebot turns supported store URLs into structured product and variant records for people and AI agents. Today, store extraction works for Shopify stores, which makes it useful for the brand-store side of this research: catalogs, prices, variants and availability, each record kept with its source URL and retrieval time. It does not report marketplace sold counts or sales value. It has an HTTP API with an OpenAPI spec, an llms.txt for agents, and exports to CSV and JSON. New accounts get 100 trial credits with no card. See the docs, the guide for AI agents and pricing. For tracking prices over time, read our price monitoring guide.
Frequently asked questions
What is ecommerce market research?
Ecommerce market research is the study of a product category using data from online stores and marketplaces: listings, prices, sold counts, brands and sellers. It is used to estimate category size, brand share, price structure and competition.
How do you estimate category market size from marketplace data?
Multiply units sold in the period by the price paid for each listing, then add up every listing in the category. Units usually come from the difference between two sold-count readings. The result is marketplace sales value, not total market size, because it excludes offline and direct sales.
How accurate are marketplace sold counts?
They are platform-reported, not audited. They are useful for share and ranking, which depend on relative size, and less reliable for exact totals. Check them against reviews, prices and previous months before you use them.
What are the best online market research methods for a new category?
Start with marketplace listings to size the category and rank brands. Add search volume data to see demand, and customer reviews to see what buyers care about. Use surveys or interviews to explain why people buy, which listing data cannot tell you.
How many months of data do I need?
At least three. One month can be distorted by a promotion or a data error. Three months show whether a brand's position is stable.
Sources
Put product data in your agent's hands.
Connect a coding agent, or check how page-based credits work before you run anything.