← All fashion signals
Trend to watch

The Algorithm on the Runway

A trend vendor reads millions of social-media images a month and says it called dotted prints before the runways. The 90% accuracy behind the pitch is its own claim.

What is it?

A vendor reads fashion trends off millions of social-media images a month and sells the output as a forecast, not a research finding, to brands such as New Balance, Skims and Prada. Whether the accuracy claim would survive an outside check is unproven.

Why now?

Pattern-reading at scale got cheap, and an image pipeline can claim what a stylist used to sense. Nobody has independently verified the numbers vendors publish. The real question: is this genuine foresight, or confident pattern-matching sold as one?

Who’s building it?

Forecasting vendors like Heuritech, selling self-reported accuracy claims directly to brands such as New Balance, Skims and Prada.

HeuritechHuman forecastersBrand buyersTrend forecastersSocial image pipelines
The full read
Why now, who is building it, where the market goes, and what stays unsettled.
Get weekly →
01 · The market
Who’s in it
Builders — vendors reading millions of social-media images a month.
Buyers — brands planning sourcing and buying further ahead on a purchased accuracy figure.
Gatekeepers — nobody yet — no outside body has checked a single claim.
What’s at stake
Forecasters who used to call a season by eye now compete with a pipeline built on other people’s images — the source becomes the input.
No fashion-specific backlash over image consent is on record.
An unverified accuracy figure sits underneath real inventory decisions.
What’s unsettled
Whether a 90% accuracy claim would hold under any outside audit.
If scraping backlash from other fields eventually reaches fashion imagery too.
Whether brands treat this as genuine foresight, or simply as a hedge against being late.
02 · How the demand matureda progression, not a timeline
Novelty
Brands test a vendor’s prediction against one season, curious if the pattern holds.
Retention
A hit prediction, such as dotted prints called early, gets budget renewed.
Willingness to pay
Trend intelligence becomes a line-item brands plan sourcing and buying around.
Category
Enough purchasing volume that the accuracy claim itself becomes worth independently testing.
03 · Where the market goesthe forks, not a prediction
→Verified discipline or unchecked hype — decided by whether any brand or outlet ever audits a vendor’s published number.
→Photographer imagery could get licensed formally by vendors, or keep feeding pipelines unpaid and uncredited.
→A consent backlash over training images may eventually reach fashion imagery too.
→Forecasting could consolidate around one or two incumbents, or fragment as more vendors enter.
Everything here turns on generalisation: whether a pattern found in millions of past images actually predicts what has never happened yet, or only what already resembles it.
04 · The Taste Economy deskadjacent markets, moving now
The Filter for Slop
Trust erodes as machine output floods feeds.
Slop becomes a top priority.
CNBC
The Human Seal
Buyers now pay a premium for verified human work.
Preference for human work flips sharply.
Digiday
Machine Taste
An algorithm now curates and narrates taste directly.
Spotify’s AI DJ curates a personalized playlist.
TechCrunch
05 · The Wire
01
Heuritech says its forecasts reach 90% accuracy
FashionUnited ↗
02
Heuritech says its AI flagged dotted prints and yellow before runways
NPR ↗
The panel on this signal
They agreed on
–A named vendor sells forecasting to brands such as New Balance, Skims and Prada as a working product, not a pilot.
–No outside body has verified any published accuracy figure.
–WGSN’s forecasting chief says AI forecasting still needs rigour and process.
The retail analyst disagreed on
–Data-provenance disputes have already reshaped other creative industries under scraping litigation; fashion’s forecasting sector has so far avoided that scrutiny mainly through low public visibility, not durable legal cover.
Know more about the method →
✦How this signal was built
01 · Panel
A Delphi panel of 4 AI analysts, each reasoning independently.
→
02 · Debate
They challenge and revise each other’s reads over two rounds.
→
03 · Evidence
Every claim is tied to a cited, linked source on The Wire.
→
04 · Synthesis
A moderator writes it into one read, disagreement and all.