P90
P90 is the most conservative of the standard demand forecast tiers: the unit volume that actual demand is expected not to exceed 90% of the time. Where P50 is the balanced midpoint and P80 is cautious, P90 is the plan you choose when being short would be genuinely expensive.
What is P90?
P90 is the most conservative of the standard demand forecast tiers: the unit volume that actual demand is expected not to exceed 90% of the time. Where P50 is the balanced midpoint and P80 is cautious, P90 is the plan you choose when being short would be genuinely expensive.
It is the forecasting equivalent of over-insurance, and for certain situations that is exactly right: uncertain new launches, peak events with hard deadlines, categories where a stockout costs visibility and ranking rather than merely a delayed sale. The premium is real inventory; the cover is protection against the worst realistic upside.
Where the top tier belongs
Three contexts make P90 the sensible plan. Peak planning — Prime Day, Black Friday, the holiday season — where demand is genuinely unpredictable, the sales window cannot be extended, and missing it means both lost revenue now and a weaker starting position afterwards. New launches, where there is no sales history to lean on and the cost of an early stockout is a damaged launch permanently. And risk-sensitive supply chains, where lead times are long enough that a demand surprise cannot be corrected in time.
Each case shares the same structure: the downside is asymmetric. Extra stock can be discounted later; a missed peak or a dead launch window cannot be recovered at all. That asymmetry — not vague caution — is the justification for the premium.
Reading and applying the number
A P90 forecast of 2,000 units a week means the platform assesses only a 10% chance that demand exceeds 2,000 — so supply plans built on that figure cover nearly every plausible outcome. Vendors see these tiers in forecast reports and buy-plan views, and negotiate against them when risk-sharing on inventory is discussed commercially.
The discipline is deliberate selection rather than reflex. Choosing P90 for everything means carrying premium inventory across products that never needed it; the correct practice is to reserve the top tier for the SKUs where the asymmetry genuinely applies, and to hold the decision stable through the planning cycle so the buffer is not repeatedly re-argued during tight-cash weeks.
In practice
A vendor planning a holiday-exclusive coffee gift set receives a P90 forecast of 10,000 units a week for December. Because the season is fixed, the demand curve is uncertain, and a stockout would waste the entire holiday window, they plan production to the P90 figure. Demand arrives inside the band, every purchase order is met, and the peak passes with the listing at full strength — the extra units sold through rather than lingering, because the buffer was sized against a real distribution rather than a guess.
How Harpy Media helps
Peak-season and launch planning is part of our vendor work: forecast tiers read against the commercial calendar, buffers sized for fixed windows and unproven demand, and production plans locked before the season rather than revised during it.
P90 FAQ
What does a P90 forecast mean?
It is the unit volume the platform expects actual demand to stay at or below nine times out of ten — a high-confidence planning level used to protect against demand surprises in uncertain or high-stakes periods.
When should I plan at P90?
For peak events with fixed windows, new product launches without reliable history, and supply chains whose lead times cannot react to a surprise. It is the tier for situations where being short is far more expensive than carrying stock.
Does P90 guarantee I will not stock out?
No — it covers roughly nine outcomes in ten. It materially reduces the risk of a stockout at the cost of carrying more inventory, which is why it is chosen deliberately per product rather than applied across an entire catalogue.
Want these numbers watched for you, every week?
Book Free Consultation