Margin & Markdown Sentinel · Brief 01
The markdown wave forming under a trend launch
Women's Apparel · 2025-10-14 · forecast miss → markdown → the ads pointing at it
The one line you configure
Which trend cohorts are selling behind their own forecast today — and is any of them building the kind of markdown wave that only becomes visible at clearance?
In a nutshell
One cohort, and it is already discounting. Three SKU clusters tagged Cottagecore_Revival — sherpa outerwear, knitwear and wide-leg denim — are running 38% below sell-through forecast at week 6 of their launch window, pulling daily Forecast Variance to −65% against an expected band of −5% to +5%. Markdown dollars on the department are 2.7× their seasonal baseline, and the giveaway is where the miss originates: 61% of it traces to the AI Trend Engine vintage, not to the statistical baseline or the buyers' own numbers.
At a glance
What moved
What's happening · forecast variance fell out of its band and kept going
The cohort is at week 6 of a launch window that runs to week 9. Read at the season's end this is a markdown line; read today it is still a buying decision.
Why · the miss belongs to the forecast vintage, not to the category
Contribution analysis of the forecast variance, one value per dimension. The same category sold to forecast under the statistical vintage — so this is the trend call, not the aesthetic.
What else moved · the same day, in two other datasets
Three datasets, no join written: merchandising forecast, retail-media attribution and marketplace returns all agree on the same department on the same day. The ads are still paying to send guests to the stock that is not selling.
So what
This is the week the decision is cheap. At week 6 the exposure is still purchase orders and ad spend; at week 9 it is clearance. Cutting the next two PO drops by 30%, pulling the retail-media spend off the three clusters, and pricing them down deliberately costs a fraction of what the same units cost after the season decides for you.
The points that matter
It is one cohort, not the department
The miss concentrates on three SKU clusters under one trend tag; the rest of Women's Apparel is inside its band. The action is that narrow.
54%one trend tagThe forecast is the fault line
61% of the variance traces to the AI Trend Engine vintage. The same categories forecast under the statistical baseline are not missing — so this is a forecasting question, not a demand one.
61%one vintageThe ad spend is compounding it
Retail-media campaigns for the department run 32% below expected iROAS because they are pointing at inventory that is not moving. That budget is buying the markdown twice.
−32%iROASRecommendation
Cut the next two PO drops on the three Cottagecore_Revival clusters by 30%, move their retail-media budget to the trend cohorts running above plan, and take the clearance decision before week 9 rather than at it.
Questions it already answers
Is this weak demand or a bad forecast?
A bad forecast. Sell-through is 38% behind plan on the trend-engine vintage while the same categories forecast off the statistical baseline sit inside their band — demand did not disappear, the number it is measured against was wrong.
How much of the department is exposed?
Three category clusters under one trend tag — Outerwear, Knitwear and Wide-Leg Denim, 37% of the variance between them, 49% of it owned-brand. The rest of Women's Apparel is on band.
What does waiting cost?
The window runs to week 9. Acting now is a purchase-order cut and a media reallocation; acting at week 9 is clearance pricing on stock already bought, which is how the prior-year cottagecore overstock became a markdown wave.