Case study · Market and competitive research
A client with a very large catalog wanted to know whether to cut it. Everyone assumed the answer was yes. Ten years of their own order data, which nobody had opened, said the opposite — and the competitor analysis explained why.
The brief was to work out why a very large product catalog was not converting on the web, and whether it should be cut back. Everyone involved assumed the answer was yes — that a catalog of three hundred thousand items was unmanageable and the sensible move was to prune it to a few thousand of the best.
Before recommending that, we went and looked at what had actually sold.
The obvious comparison was a direct competitor selling similar material at a similar price. Rather than glance at their site, we mapped it — every collection page, the three axes they organize on, the anatomy of a category page, their filter structure, their price bands, their channels, and the fact that they were mid-migration on product titles with two formats live at once.
The finding that reframed the whole engagement
Source: full crawl and enumeration of both catalogs, July 2026. The competitor organizes on format, geography and theme, with a written introduction on every collection page that doubles as its meta description.
The competitor did not have a better catalog. They had roughly a twenty-fifth of it. What they had was doors — hundreds of browsable, indexable, written category pages that gave search engines and customers a way in. The client had the inventory and no way to reach it.
The client had a decade of marketplace order data sitting in spreadsheets nobody had analyzed. It answered questions the website analytics could not.
What people actually bought
Source: analysis of 6,900 marketplace orders across a ten-year window, reconciled against a complete profit-and-loss log.
The best-selling theme in the company’s entire history — city and place panoramas, over a thousand orders — had no category page on the website. The categories had been built at some point, switched off, and left empty. Nobody had noticed because nobody had put the sales history next to the site structure.
We had been ready to recommend cutting the catalog. Then we looked at the distribution.
The long tail was the business
Source: order-level analysis of ten years of marketplace transactions, deduplicated by product.
Nearly nine in ten products that ever sold, sold exactly once. Those one-time sellers accounted for three quarters of all orders and three quarters of all revenue. Cutting the tail would have cut the business.
Keep the whole catalog on the marketplaces, where search retrieves the tail. On the website, do not cut the catalog — cut the index, and build the doors. Roughly 286 customer-facing categories in the language a customer would actually type, rather than the archival collection names the catalog had inherited.
| The usual approach | What was done here |
|---|---|
| Look at the competitor’s homepage | Enumerate their entire structure and count it |
| Trust the brief | Test the brief’s assumption before acting on it |
| Use website analytics | Use ten years of order data the client already owned |
| Deliver the expected answer | Reverse our own recommendation when the data said so |
The most valuable data in this engagement was already inside the company, in spreadsheets nobody had opened for years. That is usually true. Before buying a research subscription, it is worth finding out what the client already has and has never looked at.
Marketing Analytics Consultants
We map it properly — structure, categories, pricing, channels and change history — and set it against what your own sales data already says. Then we tell you what it means, including when it contradicts what you hoped.