Amazon search is modeled as a "spiderweb of keywords": a single buyer action (a click or a purchase) hands out partial ranking credit across hundreds of related keywords at once, rather than crediting one exact keyword and waiting for a distinct signal on every other related term.
Because Amazon can't wait for a separate click or sale on each of hundreds of related search terms, one conversion event nudges the rank of many related keywords simultaneously, but only a little each. That partial "broad" credit only converts into a durable rank gain for keywords already written in exact-match relevant form in the listing (see A9 Title Keyword Weighting (Position & Phrase Match)) — credit is distributed broadly, but capturing it depends on relevancy already being in place.
Don't assume a sales spike ranks every synonym a buyer might use — it gives all of them a small boost, but only the ones present in exact form in the listing turn that boost into a lasting rank improvement. This is why listings written only in broad/plural phrasing leave ranking credit unclaimed even on keywords that "obviously" should match.
Из тем: Unsorted