In fashion, the most important economic decision of the year — the Spring/Summer buy — is made six months before anything sells, when the season is still a hypothesis. When the forecast is wrong, and it often is, the error can't be undone: it turns into markdowns, unsold stock and trapped capital. The Vedrai Observatory flips the perspective: competitive advantage doesn't come from forecasting demand better, but from building a buy that stays economically sound even when the forecast is wrong. The report introduces four proprietary KPIs — Buy-at-Risk, Margin-at-Risk, Flexibility Ratio and Time-to-Correct — and the Risk × Flexibility map that shows where capital is truly at risk and where it's still correctable. Backed by industry data ($1.7tn in inventory distortion in 2024, the H&M case) and a concrete example on a €50M seasonal buy. A tool for buying, merchandising, CFOs and leadership who want to defend margin six months out — not report on it after markdowns.
KEY INSIGHTS
- Buy risk can be measured before the season, not just observed at markdown. With far-off production, 3-6 months pass between order and shelf: capital is committed exactly when demand visibility is lowest. The right question isn't "how well do we forecast" but "how much margin have we exposed, and how fast can we still protect it?"
- Real competitive value is optionality, not forecast accuracy. Every euro committed six months out lowers the unit cost but erodes the right to change course. The Risk × Flexibility map pinpoints the "Avoid" quadrant — high uncertainty, low flexibility — where the genuinely at-risk capital concentrates: on a €50M buy, 15% in Avoid is worth up to €7.5M of Buy-at-Risk.
- AI creates value by turning signals into decisions, not by promising perfect forecasts. Its job isn't to "know sooner" but to make better use of decision time: converting early sell-through, weather, online search and competitor pricing into scenarios and concrete supply-chain actions. Early adopters see up to -35% inventory levels and +65% service level.



.jpg)