Days on hand, measured against the date rather than the calendar

Days on hand is the simplest useful number in inventory: at the rate you are selling, how long does what you are holding last? It takes one division. The part worth getting right is what you compare the answer against, because measured against the calendar it means very little and measured against the code date it means everything.

This page covers the formula, the comparison that makes it actionable for dated goods, and where the number misleads.

How do you calculate days on hand?

Divide average inventory by cost of goods sold per day. For one SKU, divide units in stock by units sold per day, which is the same calculation without the accounting wrapper. The result is a duration: the number of days the current position covers at the current rate.

Use a recent rate rather than an annual one. An annual average silently assumes this month looks like every other month, which is false for almost anything seasonal and catastrophically false in the weeks after a listing goes live. The number is most useful computed on the last few weeks and recomputed often, because it is a forward-looking estimate wearing the clothes of a historical measure.

What should you compare it against?

Remaining shelf life, not a target. Days on hand only becomes a decision when it sits beside the days of selling window left: 60 days of cover on a product with 90 days of life is fine, and the same 60 days on a product with 45 left is a write-off already in progress.

That table is why a company-wide days-on-hand target is close to meaningless for food and drink. The same figure is healthy on one SKU and a loss on another, and averaging across a catalog with mixed shelf lives produces a number that describes nothing real. The comparison has to be per SKU, because the thing it is compared against is per SKU.

Where does the number mislead?

When demand is about to change and the rate is taken from the past. Days on hand is a projection built from history, so a promotion, a delisting or a new stockist makes it wrong immediately and it gives no warning that it has become wrong.

Two other traps are worth naming. Averaging across SKUs hides the distribution, and the distribution is the whole point — a healthy average with two items at nine months of cover is two write-offs wearing a good number. And counting stock you cannot actually sell inflates it: held batches, damaged units and stock sitting at a customer's site all show up as cover if the calculation does not exclude them.

The fix for both is the same: compute per SKU, on sellable stock only, and read the distribution rather than the mean.

What decision does it drive?

Whether to buy, slow down, or start clearing. Cover comfortably inside the window means carry on. Cover approaching the window means stop ordering and watch. Cover beyond the window means the surplus is already lost and the only question left is how much of it you recover.

Used that way it becomes an early-warning number rather than a reporting one, and its value is entirely in how early it is read. Cover exceeding the window three months out is a discounting decision with options; the same fact discovered three weeks out is a disposal. That is the difference between starting the markdown ladder while rungs still have value and arriving at the bottom of it.