What is inventory distortion?

Inventory distortion is the combined cost of overstock and out-of-stock conditions, counted as one figure. It exists as a term because the two are usually reported separately — markdown sits in gross margin, lost sales appear nowhere at all — which lets a business reduce one while quietly increasing the other and call it an improvement. Counting them together removes that option. The most widely cited estimate, from IHL Group, projected global inventory distortion at $1.77 trillion for 2023 across all retail segments, split roughly $1.2 trillion out-of-stock and $562 billion overstock.

Why are overstock and stockouts treated as the same problem?

Because they are the same decision, made once, observed from opposite ends. A buy quantity is a bet on demand. If demand comes in under the bet, the surplus becomes markdown; if it comes in over, the shortfall becomes lost sales. Nothing about the underlying process differs between the two outcomes — only which side of the forecast reality landed on. Managing them as separate problems produces the classic oscillation, where a brand overcorrects into stockouts after a heavy markdown season and back into overstock the year after.

Why is inventory distortion hard to measure in apparel?

The overstock half is visible and the stockout half is not. Markdown is recorded on every unit sold below first price. Lost sales are inferred: a size that sold out in week three did not record the demand it would have met in weeks four to twelve. Any stockout number is therefore an estimate built on an assumed demand curve, which is why most apparel businesses manage the half they can see. The practical consequence is a systematic bias toward buying too little of what works.

How do you estimate the cost of a stockout?

Take the item's rate of sale in the weeks before it broke, extend it across the weeks it was unavailable, apply a substitution assumption for demand that transferred to another size or style, and value the remainder at the margin it would have earned at full price. Every step carries judgement, so the output is a range, not a figure. Estimating it imperfectly is still better than the default, which is to treat the cost as zero.

Does a higher service level reduce inventory distortion?

Only up to a point, and in apparel that point arrives sooner than in replenishment retail. Raising service level means buying deeper, which converts stockout risk into markdown risk. On a seasonal item with no reorder, the extra units have one season to sell and no second chance, so the cost of the last few points of service level rises steeply. The right level is the one where the marginal unit's expected markdown cost equals its expected contribution — not the highest achievable number.

Where does inventory distortion actually originate?

Rarely in the replenishment logic, and usually further upstream. A size curve applied from a national average to a store that does not match it strands units in the tails and breaks the middle. An option count set above what the buy budget can support spreads depth too thin across too many choices. A delivery that slips past the floor-set date compresses the selling window. Each of these produces both halves of distortion at once, which is why fixing the symptom at the allocation layer rarely holds.

Inventory Distortion: Why Overstock and Stockouts Are One Problem

Inventory distortion measures overstock and out-of-stock cost as a single number, because they are the same planning failure seen from opposite ends. This guide explains why fixing one alone makes the other worse, how to measure both sides in an apparel assortment, and where the trade-off should actually sit.

What inventory distortion is

Inventory distortion is the combined financial cost of having too much inventory and having too little, measured as a single number. It spans both directions on purpose. Overstock shows up as markdown, carrying cost and eventual disposal; out-of-stock shows up as demand that arrived, found nothing to buy, and left. Distortion is the sum.

The reason the term exists at all is that these two costs are almost never reported together. Markdown lands in gross margin, where finance sees it every month. Lost sales land nowhere — there is no line in any ledger for a customer who wanted a medium and found the rack empty. A business measuring only what its systems record is measuring one half of a two-sided problem, and it will optimise accordingly.

The most widely cited estimate comes from IHL Group, which projected global inventory distortion at $1.77 trillion for 2023 — roughly $1.2 trillion in out-of-stocks against $562 billion in overstocks. Two caveats belong with the number whenever it is quoted: it covers all global retail segments, not apparel specifically, and it is a projection published by a firm that sells the underlying study. Treat it as an indication of scale and of the direction of the split, not as an apparel benchmark.

That split is the more interesting half of the finding. Even in general retail, where replenishment is continuous and stockouts are recoverable, the out-of-stock side is estimated at roughly twice the overstock side. It is the half nobody has a report for.

Why they are one problem, not two

A buy quantity is a single bet placed once, months before the season, on a demand number nobody knows. Everything that happens afterwards is that one bet being marked to reality. If demand lands below the bet, the difference becomes markdown. If it lands above, the difference becomes lost sales. There is no third outcome, and no separate process that produces one rather than the other.

This is why treating them as separate initiatives fails so predictably. A brand comes out of a heavy markdown season, concludes it bought too much, and cuts depth across the board. The following season sells through cleanly, margin looks excellent, and the business congratulates itself — while the bestsellers broke in week five and spent the back half of the season unavailable. That damage does not appear in the margin line. It appears the year after, as a growth number that came in soft for reasons nobody can locate.

Then the correction runs in reverse. The oscillation is not a failure of discipline; it is what happens when a two-sided problem is managed with a one-sided measure.

The asymmetry that drives bad decisions

The two halves differ in one respect that matters enormously: how visible they are.

Dimension	Overstock	Out-of-stock
How it is recorded	Directly — every unit sold below first price	Not recorded at all; must be inferred
Who sees it	Finance, monthly, in gross margin	Nobody, unless someone builds the estimate
When it appears	During and after the season	Never, directly
Certainty of the number	Exact	An estimate with a wide range
Who gets blamed	The buyer who bought too deep	Usually nobody

The last row is the one that changes behaviour. An error that produces a number with somebody's name on it is managed far more aggressively than an error that produces no number at all. Over enough seasons this asymmetry does not average out — it accumulates as a systematic bias toward buying too little of the things that work, because that error is invisible and the opposite error is not.

Any serious attempt to reduce distortion therefore has to start by building the missing half of the measurement, however roughly.

Estimating the half you cannot see

A stockout estimate will never be precise, and waiting for precision is how it stays at zero. The workable approach is deliberately crude:

Establish the rate of sale before the break. Take the item's average weekly units over the two to three weeks before it went unavailable, at the size and location level where the break actually happened.
Count the weeks lost. From the break to either the end of the season or the arrival of a replenishment, whichever comes first.
Apply a substitution assumption. Some demand transfers to an adjacent size, a similar style, or another colour. Some walks. The transfer rate is a judgement call and should be stated as a range rather than a point.
Value the remainder at full-price contribution, not at retail — this is margin that did not happen, not revenue.

The output of this is a range, and it should be presented as one. A single confident number invites an argument about the number instead of a conversation about the decision. A range that says "somewhere between a quarter and half a million dollars of contribution, concentrated in four styles" is both more honest and more actionable.

The point is not accuracy. It is to make the invisible half of the trade-off visible enough that it enters the conversation at all.

Where the trade-off should sit

Once both halves are on the table, the question stops being "how do we reduce inventory?" and becomes "where is the marginal unit worth buying?" — which is a genuinely different question with a genuinely different answer per item.

The marginal unit is worth buying while its expected contribution if it sells exceeds its expected cost if it does not. Two things drive that comparison, and neither is a service-level target:

The markdown cost of being wrong. A basic in a continuing colour that carries over to next season has a low cost of error — worst case, it sells later at a smaller discount. A trend item in a signature colour has a high one, because it exits at a deep markdown or not at all.
The recoverability of the shortfall. An item that can be chased has a cheap stockout, because the break is temporary. An item on a twelve-week overseas lead time with no reorder window has an expensive one, because the break lasts the rest of the season.

These two factors point in opposite directions for different parts of the assortment, which is why a single service-level target applied across a range is almost always wrong. The replenishable core and the seasonal newness should be bought to different standards on purpose. This is the same reasoning that makes the standard safety-stock formula misbehave on seasonal assortments — see safety stock for seasonal assortments for the arithmetic.

Where distortion actually originates

Most distortion is created well before the allocation engine sees it, which is why allocation-layer fixes tend not to hold. The recurring upstream sources:

The size curve. A national average curve applied to a store whose customers do not match it strands units in the tails while the middle breaks. The store shows both halves of distortion simultaneously, from one decision made centrally. Store-level or cluster-level curves are the fix, and store clustering is the mechanism.

Option count against buy budget. Breadth and depth trade against each other inside a fixed budget. Set the option count too high and every option is bought too thin — the whole range breaks early and none of it was ever deep enough to matter. This is a line planning decision that surfaces as an inventory problem months later.

The delivery calendar. A delivery that lands three weeks late does not simply arrive late; it arrives into a shorter selling window, at a point in the season when full-price demand has already begun to decay. The same units that would have sold through cleanly now need markdown to clear. A delivery date is an inventory decision wearing a logistics costume.

The forecast's shape, not just its total. A plan that gets the season total right but its weekly phasing wrong produces stockouts in the peak and overstock in the tail, netting to a total that looks correct. Reviewing only the season number hides this completely — see how to phase a sales plan.

What to do about it

The sequence that works, in order:

Report both halves together, every season. Even a rough stockout estimate alongside the markdown number changes what gets discussed. A margin review that shows only markdown will always conclude that the answer is to buy less.
Segment the assortment by cost of error rather than applying one service standard. Core, seasonal and trend should be bought to different depths on purpose, with the reasoning written down.
Attribute distortion to the decision that caused it, not to the layer where it surfaced. If the size curve caused it, changing the allocation rule will not fix it.
Check the phasing, not just the total, when reviewing plan versus actual.
Preserve the chase where lead time allows it. The cheapest way to reduce both halves at once is to commit less up front and retain the ability to react — which is a speed-to-market question as much as a planning one.

The last one is the only lever that genuinely improves both sides simultaneously. Everything else trades one against the other, and the work is deciding where that trade should sit for each part of the range.