How does F-Predict forecast colours?

The problem this solves

The obvious way to build an emotion-led colour forecast is a lookup table: confidence returns red, calm returns blue. It is also the reason such tools return the same palette forever. If the emotion determines the colour, then the market, the category and the region cannot — and two completely different briefs produce identical swatches.

F-Predict inverts that. The market supplies the colours. The emotion supplies the reading. Confidence in a market whose evidence runs warm resolves toward burnt orange and brick; the same confidence in a cool-running market resolves toward inked blues. Same emotion, different answer, because the evidence differs.

How a palette is actually built

1. 1

   1. Collect colour evidence

   Every evidence stage contributes colour: designer campaign imagery, street signals, catwalk direction, material and trim signals, and the narrative lane. Two kinds of evidence are captured — literal hex values observed in imagery, and colour language published in brand and press copy.

2. 2

   2. Cluster and weight

   Raw colour evidence is clustered into families and weighted by how often and how independently it appears. A hue confirmed by three unrelated stages outranks one that appears once.

3. 3

   3. Derive themes

   Clusters become named colour themes with a stated rationale — what the theme is responding to, and which evidence supports it. A run guarantees a minimum of four themes so a range has real breadth to plan against.

4. 4

   4. Build harmonies

   Around each theme, full harmonies are generated mathematically — the swatch ladder is computed, not asked of a language model, which is what keeps it internally consistent and reproducible.

5. 5

   5. Temperature-correct the neutrals

   Neutrals are resolved warm or cool to match the theme rather than defaulting to a generic grey, because a neutral that fights the palette is the fastest way to make a colour card unusable.

6. 6

   6. Match to Pantone TPX

   Each resolved shade is matched to a Pantone TPX reference so the output is directly usable in a colour card, a tech pack and a lab-dip conversation.

Where emotion enters

Emotion enters as a lens at interpretation time. F-Predict uses a 25-emotion taxonomy organised into three groups — Empathy, Enjoyment and Assurance — across nine clusters. Each emotion carries a product profile covering personality, purpose, silhouette, fabric, detail and pattern tendencies.

That profile biases which evidence is treated as most relevant and how a theme is framed and named. It does not select the hex values. When colour evidence for a scope is genuinely thin, the emotion profile is used as a documented fallback so the output is still coherent — and the grounding indicator shows you that this is what happened.

Why the output is traceable

• Each colour theme reports the evidence stages that fed it, so you can see whether a direction rests on runway, on street, on material, or on all three.
• Corroboration across independent stages is scored, so a well-supported call is distinguishable from a single-source one.
• Evidence is anchored to the present day. The target season is the destination of the forecast, never the date of the evidence used to build it — a rule enforced across every stage.
• Sources are cited per stage rather than at the end, so a specific claim can be checked rather than a whole report.

What else the colour stage produces

• A curated season palette tray you assemble by hand from the generated themes.
• Harmony sets for each theme rather than a single flat swatch row.
• Palette export for handoff, plus PDF output of the full direction.
• A record of which evidence stages contributed to the colour direction shown on the colour tab itself.