Forecasting May 26, 2026 · 5 min read · All posts

Why your reorder point is probably wrong (and the seasonal fix).

Most reorder points are off by 30–60% in any given month. Not because the formula is bad — the formula is fine. Because everyone runs it on a flat 12-month average in a business that isn't flat.

Most reorder points are wrong by 30–60% in any given month. The formula is fine. The problem is that almost everyone runs it on a flat 12-month average, in a business that has never once behaved like a flat 12-month average.

Every inventory textbook teaches the same formula:

Reorder Point = (Average Daily Demand × Lead Time) + Safety Stock

It's correct. Run it against the trailing 12 months for any product and you'll get a number that tells you when to reorder. Use that same number every month and you'll stock out twice a year: once heading into your seasonal peak, once on the way back down.

Why?

The flat-average trap

Imagine you sell a product that moves 10 units a day in normal months and 30 units a day in November. Trailing 12-month average daily demand is 11.7 units (10 × 11 months + 30 × 1 month, divided by 365 days, give or take).

If your lead time is 30 days and you carry one lead-time of safety stock, your reorder point is 702 units.

In October that number is way too high. You'll order too much, too early.
In November-December it's way too low. You'll stock out 12 days into your peak.

Nothing wrong with the math. The inputs are the problem. You're using yesterday's average to predict tomorrow's demand, in a business where tomorrow looks nothing like yesterday.

The seasonal fix

A reorder point that adjusts for seasonality needs a few things working together.

1. Detect seasonality per SKU

Category-level and brand-level patterns are too coarse. Every SKU has its own. A coffee mug might peak in November for Christmas gifting while a coffee bean peaks in May for Mother's Day. Flat averages mash the two together. Per-SKU detection gives each product its own seasonal profile.

2. Project demand for the lead-time window, not the past 12 months

If your lead time is 30 days and you're in October, you don't care about January's average. You care about what November is likely to bring. Pull the same 30-day window from each of the last two years, blend with current trend.

3. Floor the seasonal value at the standard one

The seasonal calculation is a best estimate, and a best estimate can be wrong low. It should only push your reorder point above the trailing average when the season warrants, never below it. If the seasonal projection comes in under the standard, use the standard.

What this looks like in practice

In Stocura, every SKU has two reorder points side by side: Standard (the textbook trailing-average number) and Seasonal (the lead-time-window-aware number). When seasonal is higher, the dashboard uses seasonal as the trigger, so you get the alert earlier going into peak. When seasonal isn't materially different, the standard wins.

For a Mother's Day-heavy SKU in our own range, the difference looks like this:

Stocura uses 26. We get the alert with three weeks of buffer instead of three days.

Why this matters more than people think

Stock-outs in your peak season cost you 3–5x more per dollar of missed sale than stock-outs in shoulder season. By then your customer-acquisition spend is already sunk. You've paid Google and Meta to drive the traffic, and if you can't fulfill, you've handed your margin to the ad networks.

Fix the seasonal trigger and you stop paying for traffic you can't fulfill.

Forecasts that learn your business.

Stocura watches 24 months of demand, finds the seasonal pattern per SKU, and adjusts your reorder point automatically.

Request beta access