The industry makes far more than it sells, then discounts its way out. That reflex is expensive, and it is a planning problem long before it is a markdown problem.
Sources: McKinsey, Tackling retail's inventory glut (2023); McKinsey / Business of Fashion, State of Fashion 2025.
Fashion has a habit the rest of manufacturing would find strange: it deliberately builds more than it expects to sell, then relies on markdowns and outlets to clear the rest. The scale is hard to fathom. McKinsey has pegged unsold goods sitting in US retail inventory at around $740 billion, and estimates the fashion industry produces somewhere between 2.5 and 5 billion excess items every year. That is not a rounding error. That is cash, and increasingly a sustainability liability, built in on purpose.
Every unsold garment was paid for twice: once to make it, and again to discount it off the books. A full-price sale that turns into a 50-percent markdown is not half a win; after the cost of the goods, it can be no win at all, or a loss. The markdown looks like a pricing decision, but it is really the bill for a demand plan that committed to volume the market did not want, months earlier, at the buy.
Three things make it brutal. Long lead times mean the buy is committed a season or two ahead, on a guess about what will sell. Short, trend-driven lifecycles mean a style can be dead before the reorder would even arrive. And proliferation across styles, colours and sizes multiplies the number of forecasts, each one thin and noisy. Plan it on last season plus a feeling, and you are overbuying the misses and underbuying the hits at the same time, which is the worst of both worlds.
The aggregate forecast can look fine while the detail quietly bleeds. You may have planned the right total units for a style and still be sold out of the mediums and drowning in the extra-smalls. Service and markdown are decided at the size-colour level, not the style level, so the planning has to live there too. Most spreadsheets give up long before that granularity.
Breaking the markdown cycle starts upstream, at the buy, not at the clearance rack. That means forecasting demand down to the size and colour on real signals rather than a flat uplift, sizing initial buys and replenishment to what each option will actually sell, and flagging the slow sellers early enough to move them at a shallow discount instead of a deep one. It is the same reproducible, explainable approach DemandIntel brings to apparel and textile players, so the plan can be defended at the buy meeting, not just explained at the post-season review.
Long lead times force the buy to be committed a season or more ahead, so brands pad volume to avoid stockouts on hits. Short trend cycles then leave the misses unsold. The result is structural over-making, estimated at 2.5 to 5 billion excess items a year.
Fix the buy, not just the clearance. Forecast demand at the size-and-colour level on real signals, right-size initial and replenishment orders to expected sell-through, and catch slow sellers early so they move at a shallow discount instead of a deep one.
Yes, if it works at the option level and updates as early sell-through comes in, rather than relying on last season plus a gut uplift. The point is a defensible buy quantity, not a perfect long-range forecast.
Because service and markdown happen there. A style can hit its total units and still stock out on mediums while piling up extra-smalls, so the plan has to be built at the size-colour granularity, not the style.
In fashion, the markdown is not a pricing decision. It is the invoice for a buy you committed before you knew what would sell.
Picture your next S&OP meeting: one number on screen that nobody argues with. See it free, on your own data, in 45 minutes; and if it does not hold up, we part as friends.