Invisible to every filter on the left-hand side of the page

A large share of marketplace shoppers never type a second search. They land in a category and narrow it down the left-hand rail — by sport, by size, by type, by material — and every one of those refinements is driven by the browse node a product sits in and the attributes filled in behind it.

The brand's catalog had been uploaded over six years by four different people through three different templates. Running shoes sat under a generic accessories node, half the apparel had no size or gender attribute, and two hundred ASINs were in nodes for a sport the brand does not sell. None of this produced an error message, so nobody ever saw it. The effect was simply that a shopper filtering a category never met the brand — not because it ranked poorly, but because it was not in the set being filtered.

We treated catalog data as a ranking surface rather than as paperwork.

• Node-by-node audit of all 1,240 ASINs against the category tree, including the nodes competitors occupy for the same product type.
• Attribute completion at scale — the specific fields each node exposes as a filter, filled for every product rather than the minimum required to publish.
• Template discipline — one maintained flat file per product family, so the next upload cannot reintroduce the drift.
• Listing text realigned to the node, because a product indexed in the right place still has to read like it belongs there to convert.

Sessions grew from 148k to 204k a month and revenue rose 59%, with no new products, no price changes and no additional advertising.

The traffic came from a source the account had never touched: shoppers browsing and refining rather than searching. It is also traffic that cannot be bought — there is no bid that places a product into a filter it does not qualify for — which is why six years of advertising optimization had never surfaced the problem.