A trend is only useful if you can sell into it
Every trend conversation inside a brand eventually reaches the same four words: what do we sell against this? It's the moment a trend stops being culture and starts being commerce, and it's where most trend tooling quietly hands the problem back to you. I've come to believe this last step, from a named trend to the specific products in your catalogue that fit it, is the crux of the entire job. It's also much harder than it looks.
The vocabulary gap
Here's the core of it. Trends are named by the crowd, and the crowd names feelings. An aesthetic gets called something like cottagecore or quiet luxury, words about a mood, a life you're imagining, a person you'd like to be. Products are named by merchandisers and product-feed systems, and those name attributes: "floral midi dress, pleated hem, viscose." Both descriptions are accurate. They almost never share a single word.
Once you see this gap you see it everywhere, and you also see why the two obvious shortcuts fail. The first shortcut is keyword search: type the trend name into your own product search and feature whatever comes back. For trend words, what comes back is close to nothing, or worse, the handful of products whose copywriter happened to be chronically online that month, which is a merchandising strategy built on an accident. The second shortcut is doing it by hand: a merchandiser who knows the range walks the catalogue and picks. This actually works, it's what good brands have always done. It just doesn't scale. It's one person, one trend, one afternoon, and the selection is stale the moment the range changes or a second trend shows up. Hand-picking survives at boutique scale and collapses at fifty thousand products.
What the matching actually has to do
So the real requirement is matching on meaning. Something has to understand that "cottagecore" implies florals, natural fabrics, a certain silhouette, a certain softness, and then find the products whose descriptions and images express those qualities, even though the words never touch. That's a judgement, the same judgement a fashion-literate person makes instantly when you show them a dress and ask "is this cottagecore?" The whole game is making that judgement automatically, across an entire catalogue, for every trend, every day, and having it be right often enough to trust.
And there's a second requirement I'd argue is just as important: the matching has to be honest when the answer is no. Sometimes a brand simply doesn't carry the trend. A system under pressure to always return something will dress up its weakest matches as answers, and one flagrantly wrong product on a trend page costs you the marketer's trust in all the right ones. A confident empty answer is more useful than a padded one, and, read properly, it isn't even bad news. "This trend is rising and you have nothing for it" is one of the most valuable sentences a buying team can hear.
Why I think this is the crux
Detection tells you what's happening in the world. Matching tells you what it means for you, and "for you" is the only version a working marketer can act on. It's the difference between a trend report and a trade decision, and it's exactly the step that stays manual in most trend tooling, which I suspect is half the reason those reports die in Monday meetings the way I described in the manifesto post.
Of course, claiming your system makes a fashion judgement across fifty thousand products raises an obvious question: how do you know it's any good? Answering that honestly turned out to be a project in itself, and it got its own post: I couldn't tell you if our matching actually worked.