Trend Score
One number per product, refreshed weekly, from what people are posting and wearing. Feed it into ranking, ads placement, or cataloguing decisions.
Your ranking knows what shoppers clicked.
Gist knows what they’ll search next.
Discovery is trained on clicks, add-to-carts, and purchases — all of which arrive after a trend does. Gist reads what creators are posting each week and hands your stack a signal that moves first.
Three products, one weekly signal
Built for the discovery stack you already run.
One for each stage you already own — ranking, query, and browse.
Emerging Keywords
The terms gaining momentum before they surface in your search logs. Adapt listings to the language shoppers are actually using.
Trendy Homepage Collections
Themes shoppers are already responding to, shaped into collections for the homepage and category pages. Refreshed weekly, so they never go stale.
One API call. One weekly refresh.
Ships as an attribute before it ships as a model.
All three run on the same API. Fetch weekly, map it onto your catalog, let merchandisers boost on it — then move it into ranking once the test clears.
Fetch
GET /v1/context/fashionMapSignal onto catalog
IndextrendScore attribute
BoostRank and merchandise
TypeScriptknowledgeCutoff: 2026-07-26
// Weekly job — nothing runs in the request path.
const context = await fetch(
"https://gist-api.the-alt.co/v1/context/fashion",
{ headers: { Authorization: `Bearer ${GIST_API_KEY}` } }
).then((response) => response.json());
if (!context.noSignal) {
await index.updateAttributes(
scoreCatalog(catalog, context.read)
); // → { sku, trendScore, trendTerms[] }
}Where the signal lands
Six places it changes what shoppers see.
Cold-start ranking
Score new arrivals before they have collected a single click.
Boost and bury rules
Index it as an attribute and merchandisers can boost on it. No model change.
Query understanding
Resolve the aesthetic names shoppers type that map to nothing in your catalog.
Zero-result recovery
Turn null searches into synonyms pointing at products you already stock.
Attribute enrichment
Add current aesthetics, cuts, and occasions to product metadata weekly.
Ads and placement
Weight sponsored slots toward what is rising, not what converted last quarter.
A leading signal beside a lagging one
Behavioural data tells you what already happened.
Behavioural signalsGist signal
Exists only once a product has been surfacedThere on the day it is listed
Blind to new arrivals and the long tailScores SKUs with no click history
Learns a trend after shoppers find itReads it where it starts, on social
Limited to how your catalog is taggedSurfaces terms before your logs do
FAQs
How does this get into our ranking?
Start by mapping the weekly signal onto your catalog as a product attribute — merchandisers can boost or bury on it with no model change. Once it proves out, pass it as a feature into second-stage ranking.
Does this add latency to the search path?
No. You fetch on your schedule and write to your index, so the signal is already resident when a query arrives. Nothing is generated while a shopper waits.
What about products with no behavioural data?
That is the sharpest use of it. Clicks only exist after a product has been surfaced, which leaves new arrivals unrankable. A cultural signal is there on day one.
How fresh is it, and how do we know?
Weekly. Every response carries a knowledge cutoff and a date window, so your systems know exactly which period a signal describes.
What happens when there is not enough evidence?
The response returns noSignal. Your ranking falls back to its existing logic instead of ingesting an invented trend.
Can we query a specific market, category, or attribute?
Yes. Live Query handles regions, demographics, sub-categories, garment types, colours, and cuts — so the signal matches the catalog you run.
Are you competing with our discovery platform?
No. We do not rank, merchandise, or sit anywhere near your shopper. You own discovery; we supply the input your behavioural data cannot see yet.
Culture moves first. Your ranking should follow it.
Index it this sprint, A/B test it the next.