Black Friday demand: how a shop can size its stock with a probability, not a guess
Order too much Black Friday stock and you eat markdowns; order too little and you lose sales. Here is how a probability-based forecast turns that gamble into a calculated bet.
Outcomer Team · Aug 1, 2026
Every retailer faces the same November headache: how much to stock for Black Friday. The buyer commits to an order weeks in advance, long before anyone knows whether the promotion will draw a crowd or a shrug. Order too much and the surplus gets dumped at a markdown in December. Order too little and the shelves clear by mid-morning while customers walk out empty-handed. This is a case study in replacing that gut-feel bet with a number you can actually reason about.
The problem: two ways to be wrong
Picture a mid-sized electronics shop deciding how many units of a discounted headline product to buy in for Black Friday. It pays €40 per unit and plans to sell at €70, a €30 margin. Whatever does not sell on the day gets cleared afterwards at €35 — a €5 loss per leftover unit.
The trap is that both mistakes cost real money, and they pull in opposite directions. Every unit that sells earns €30; every unit that sits earns minus €5. So the whole decision hinges on one uncertain quantity — demand — and the buyer is effectively taking a position on it whether they admit it or not. The useful move is to make that position explicit and price it.
The forecast: turn "will it sell" into a probability
This is exactly the kind of uncertain, resolvable question a market is good at pricing. Whether it is a public prediction market on a demand driver, an internal market where staff and suppliers trade on the season, or simply market-style probabilities pulled from last year's data, the output is the same: a number between 0 and 100% for "will we sell at least X units?" If the idea of a price standing in for a probability is new, our primer on what a prediction market is covers it in a couple of minutes, and it rests on the same wisdom-of-crowds effect that makes aggregated forecasts beat a single manager's hunch.
Suppose the forecast says there is a 70% chance of a strong day (roughly 500 units sold) and a 30% chance of a weak one (roughly 200 units). The buyer no longer has to pick one story and pray — they can weigh both.
The unit economics: pick the order that maximises expected profit
Compare two order sizes against those probabilities.
Order 200 units (play it safe). In both scenarios all 200 sell, because even a weak day clears 200. Profit is a flat 200 × €30 = €6,000, guaranteed. No leftovers, but also no upside if the day goes well.
Order 500 units (back the strong day). On a strong day, all 500 sell: 500 × €30 = €15,000. On a weak day, 200 sell for €30 each (€6,000) and 300 are cleared at a €5 loss (−€1,500), netting €4,500. Now weight those by the forecast:
Expected profit = (0.70 × €15,000) + (0.30 × €4,500) = €10,500 + €1,350 = €11,850.
The bigger order carries real downside — €4,500 on a bad day is worse than the safe €6,000 — but its expected profit of €11,850 is nearly double the cautious plan. The probability is what lets the buyer see that trade-off instead of arguing about it. And you can find the break-even directly: ordering the extra 300 units pays off as long as the chance of a strong day clears the ratio of the leftover loss to the total swing — here, comfortably below the 70% the forecast gives.
Hedging the driver, not just guessing it
Sometimes demand hangs on one identifiable event: a cold snap that sells out coats, a console launch landing before the sale, a rival's stock-out. When that driver trades on a market, the shop can do more than forecast it — it can hedge. Buy the outcome that would hurt you, and a bad-news day is softened by a payout, the same mechanic a venue uses when it hedges weather risk for an outdoor event or a pub uses to run a free-drinks-if-we-win promotion without an open-ended bar tab. The promotion still fills the room; the tail risk just stops being a surprise.
Where it breaks — and how to keep it honest
A few caveats. A forecast is a probability, not a promise: 70% strong days still mean three weak ones in ten, so size orders you can survive on the bad draw. Feed the model real signals — last year's sell-through, web traffic, supplier lead times — rather than optimism. And keep the question sharp: "at least 500 units on Black Friday" resolves cleanly, "a good season" does not. The tighter the question, the more trustworthy the number.
None of this requires risking real stock to learn. On Outcomer you can practise reading market probabilities and sizing a position with virtual money — pick a market, work out the expected value, and watch how it settles. It is the cheapest way to build the instinct before a real purchase order is ever on the line.