Traffic is the expensive half. Conversion is the half most brands leave alone. If a listing turns 12% of clicks into orders and it could turn 22%, no bid strategy will make up the difference.
Variation structure comes first
Splitting a range across separate listings splits reviews, splits sales velocity and confuses shoppers about which page is the real one. Merging them properly — under one parent, with review pooling tested rather than assumed — is usually the single highest-return change available in a catalogue. On one supplements account, merging two size variants and rewriting the parent took the post-merge conversion rate to roughly 30–35%.
Writing for how search reads now
Retail search no longer matches strings. It reads listings the way a shopper would, in whole sentences, and answers questions the shopper never typed. That changes what belongs in a bullet: not a keyword list, but a plain statement of who the product is for, what it replaces and what happens after the first order.
We rewrite catalogues to be legible to both — the crawler that still indexes backend terms and the model that summarises the page.
Where the images do the work
Buyers scan, they do not read. The first three images carry the argument; the copy confirms it. We brief imagery off the same competitive research the ad plan uses, so the claim you win on in search is the claim the image makes.
Every rebuild ships alongside the ad program, never after it. See MagLite for a catalogue rebuilt across 46 SKUs in phases.