Measurement

What belongs in the sell-through denominator?

Sell-through looks like one division: units sold, over units made available. The argument is about what “made available” means. Two readings genuinely compete — the units you received, and the whole pool you had available — and on one SKU selling 540 units in a 30-day month they read 90.0% and 54.0%. Two further ratios get filed under the same name and are answering a different question altogether. The numbers are an invented worked example, built so the arithmetic can be checked.

Turnover tells you how many times over you replaced the stock you held. Days of stock tells you how long what is left will last. Neither answers the question a buyer asks once a delivery has landed and a month has gone by — did that go?

The formula, and the word doing all the work

Sell-through rate = units sold in the period ÷ units made available in the period, × 100.

Nothing in that is contestable except “made available” and “the period”, which is to say that everything in it is contestable. Sell-through looks like one simple division, but retailers and reporting tools do not all use the same inventory denominator: two people can take one SKU, one month and one ledger, apply the formula honestly, and come back with 90.0% and 54.0%. That is the formula under-specifying itself, not sloppiness.

One clearing-up first. Turnover divides throughput by average inventory and asks how many times the stock you carried was replaced; sell-through divides sales by a fixed pool and asks what fraction of that pool left. Different denominators, different questions, neither converting into the other. Hold on to that distinction, because later in this article it is what separates a sell-through rate from a ratio that is only wearing the name. Days of stock is a third thing, covered in its own post.

Step 1 — Fix the window, and match it to the decision

The window silently sets the denominator, so it comes first, and the right length depends on what you are judging. For a specific buy, the buying cycle is usually the most useful window. For a standing report that ranks a whole catalogue, a fixed recent window is easier to read across SKUs bought on different cycles. What you cannot do is judge a buy on a window shorter than the cycle that buy sits in. Here is the SKU the rest of the article runs on, over one ordinary 30-day trading month in one warehouse.

Now suppose the SKU is bought every 60 days — 1,200 units in one delivery — but a receipt-based report runs 30-day windows. The window holding the delivery divides 540 sales by a 1,200-unit receipt and returns 45.0%. The next window holds no delivery, so its denominator is zero and there is no answer at all: a depressed figure, then a blank.

A receipt-based sell-through measured on a window shorter than the buying cycle is measuring your calendar, not your selling.

Step 2 — Two common denominator families that are both sell-through

Back to the 30 days and the 540 units sold. Two pools have a real claim to the name, and they ask two different but related questions: did this buy sell, and did the stock I had to sell from sell.

Denominator 1 — units received in the period

The pool is the delivery: 600 units. 540 ÷ 600 = 90.0%, which judges the purchase order. It carries a matching problem. The 600 landed on day 10, and the nine days before it sold 9 × 18 = 162 units off the opening stock, so 378 of the month’s sales occurred after the delivery landed. That is a fact about timing, not about identity: unless you track lots, identical units are fungible and nobody can say which physical units left. Matched on timing rather than identity, the delivery reads 378 ÷ 600 = 63.0%.

Denominator 2 — opening stock plus receipts

In the simplified flow here, opening stock plus receipts represents the pool available during the period: 400 + 600 = 1,000 units, so 540 ÷ 1,000 = 54.0%. This is closest to the plain-English question, and in a clean inventory-flow model it is naturally bounded at 100% — you cannot sell more than went through. Real ledgers carry returns, transfers, negative adjustments and oversells, so a figure above 100% on this basis points at the data rather than at a strong month. Its cost is blurring the buy into the carry-over: a small delivery on a large old pile and a large delivery on an empty shelf can land on the same figure.

The two side by side

54.0% and 90.0% on one SKU, one month and one ledger, and neither is wrong. Publish either as “our sell-through” without naming the denominator and the number is unreadable. That is where the honest disagreement about sell-through sits. The next section covers two ratios that are reported under the same word and are answering a different question — not a third and fourth denominator, but a different measure. Both sell-through denominators keep appearing after it, because the rest of this article is about choosing between them.

Two ratios that borrow the name

Divide the same 540 sales by an average balance or by an opening balance and you get numbers that are often printed as sell-through. Both are worth having. Neither is the same metric, and the giveaway is the question each one answers.

Average units on hand — an inventory productivity ratio

Divide by the stock the period ran on and you are asking how hard the stock you carried worked. That is turnover’s question in different clothing: turnover is throughput over average inventory, and sales over average on hand is the same shape of ratio on a shorter clock. Kept under its own name it is a good measure. Filed under sell-through, it quietly swaps one metric for another, and the reader has no way to tell.

It also hides a sub-choice. The quick way averages the endpoints: (400 + 460) ÷ 2 = 430 units, so 540 ÷ 430 = 125.6%. The honest way averages the balance across all 30 days, which differs because the 600 landed on day 10 rather than evenly. Sales run at 18 a day:

16,230 unit-days ÷ 30 days = 541 units average on hand, and 540 ÷ 541 = 99.8%. Two numbers, one label, and the gap is entirely the day the delivery landed: the endpoint average cannot see a receipt that arrived and was partly sold down inside the window. That is a second reason to keep this ratio in its own box. Before it can be compared with anything you have to state not only the denominator but how the average was taken.

Opening stock alone — a coverage check

The pool is what was on the shelf when the window opened: 400 units. 540 ÷ 400 = 135.0%. The arithmetic is right and the reading is useful — could day one’s shelf have carried the month? No, by 140 units. But a ratio that runs past 100% whenever the receipts arrived later is answering a coverage question, not measuring what fraction of a pool sold. Call it sales against opening stock. Calling it sell-through teaches the reader that any inventory denominator is fair game, which is how a 135% and a 54% end up in the same column of the same report.

The related views

None of the arithmetic is wrong; the ratios answer different questions. The spread from 54.0% to 135.0% on one SKU in one month is not a disagreement about sell-through. It is several different measurements, two of which are sell-through and the rest of which are wearing its name.

So which one

Step 3 — Read it per SKU and per channel, not for the business

Per SKU, because the ranking flips

A second SKU over the same 30 days sells exactly the same 540 units, from an opening 900 and a 200-unit delivery, closing at 560.

Identical unit sales, and the two sell-through denominators disagree about which SKU did better. That reversal sits inside the metric, not between it and a cousin of it: the second SKU sold nearly three times its delivery, which says the delivery was small, and moved 49.1% of everything it had available against the first SKU’s 54.0%, which says it is sitting on old stock. The coverage row is a third statement again, and it is the one most likely to be mislabelled on the way into a report.

Per channel, where the pool is genuinely separate

Of the first SKU’s 1,000 available units, say 350 went into a marketplace’s own fulfilment network, a physically separate pool, while 650 stayed in your warehouse serving your own store and another marketplace.

The blended 54.0% describes neither pool, yet it is the number that gets reported. On it the SKU is moving moderately, when one pool is close to clearing and the other has barely started. Those two pools deserve different replenishment decisions, and the split tells you where to look rather than what to ship: lead time, days of cover and whether you can transfer between the pools decide that. The 650-unit pool serves two channels off one shelf, so report sell-through at the pool.

There is no good sell-through rate

This is where articles on this topic reach for a round number, and there is no defensible one to reach for. A figure that is healthy for a SKU bought twice a year is alarming for one bought weekly, and two businesses quoting the same percentage may not be computing the same thing.

Refusing the benchmark leaves a hole, so here is what goes in it. Fix the denominator, the window and the product. Compare the SKU against its own comparable periods on the same basis — comparable periods with similar known context, the same stage of its life, not merely the month before. Then read the direction of travel and ask what it requires: a larger buy, a smaller buy, a pause, a transfer between pools, or an exit. That is a decision rule. A borrowed percentage is not.

Hold the denominator, the window and the SKU fixed, and the movement between windows carries the information. Same first SKU, three consecutive 30-day windows, standing order of 600 in each.

Nobody needs a benchmark to read that. Sell-through on receipts falls ten points a window while closing stock climbs from 460 to 760, because sales fall by 60 units a window and the order does not move. Across the 90 days 1,800 arrived and 1,440 sold: 360 units accumulated, exactly the rise from 400 to 760.

What the number actually decides

Sell-through earns its keep in merchandising, clearance and supplier reviews as well. But the replenishment decision is the one this trend answers, and it answers it in three directions.

Reorder less

End of window 3: 760 units on hand selling 420 a window, which is 14 a day, so 760 ÷ 14 = 54.3 days of cover. End of window 1: 460 units against 18 a day, or 25.6 days. Cover more than doubled while the order stayed at 600. Sales averaged 1,440 ÷ 3 = 480 a window, which is the backward-looking break-even receipt rate: the order that would have held stock flat across the three windows just gone, not a recommendation for the next one. The pace is falling by 60 a window, and if that straight line continued, window 4 would be 420 − 60 = 360. Order 600 again on that scenario and closing stock goes to 760 + 600 − 360 = 1,000.

Stop reordering

The next decision along, and the one that gets skipped. Cancel window 4’s order and, on the same straight-line scenario, closing stock lands at 760 − 360 = 400 units, which at 360 a window is 400 ÷ 12 = 33.3 days of cover. This SKU needs no delivery next cycle.

Reorder more

The mirror case, a different SKU on the same standing order of 600 a window:

Above 100% on receipts and rising means the shelf is being eaten into: 1,980 sold against 1,800 received, stock down from 300 to 120, sales rising by 40 a window. Run window 4 unchanged and the sum stops working. If the rise held to 740, opening 120 plus a 600 delivery is 720 units against 740 of demand, so the SKU finishes 20 short; at 740 ÷ 30 = 24.67 a day, 720 units last 29.2 days.

For a replenishment decision, a sell-through rate that never changes the quantity is decoration with a percent sign on it.

What to do this week

1. Match the window to the decision. The buying cycle when you are judging a buy, a fixed recent window when you are ranking a catalogue. SKUs read on different windows are not comparable with each other.
2. Write the denominator beside the number. “Sell-through 90%” is not a fact. “Sell-through 90% of units received, 30-day window, one SKU, one warehouse” is.
3. Check what your report is dividing by. If the denominator is an average balance or an opening balance, you are reading a productivity or a coverage ratio. Keep it, rename it, and stop setting it beside a receipt-based figure.
4. Compare multiple consecutive windows rather than one isolated figure. A single figure has nothing to be read against, and invites comparison to a benchmark that does not exist.
5. Split the denominator by physical pool where a channel holds its own stock, and refuse where the stock is shared.
6. For replenishment, finish with a quantity. Project the next window off the trend, say out loud that it is a projection, and set the order so closing stock lands where you want it.

Sell-through is a property of the order you placed and of the shelf you sold it from, read back to you after the fact. Which of those two you divided by is not a detail. It is the metric.