How to stop running out of your best sellers

You avoid stockouts by knowing three things for every SKU that matters: how fast it sells, how long your supplier takes to deliver, and therefore the stock level at which you have to place the order. That level is the reorder point. Everything else in this guide is detail around it.

The trap is that zero is not the deadline. The deadline is the day your remaining stock equals the demand you will see while the next shipment is in transit. Hit that day without a purchase order out, and the stockout is already booked. Nothing you do after it changes the outcome, you just find out later.

So the work splits into five moves:

• Measure sales velocity per SKU over a window you have chosen on purpose.
• Get a real lead time per supplier, in days, including receiving.
• Set a reorder point, and write it down somewhere that is not your head.
• Decide how much safety stock each SKU deserves, knowing it costs cash.
• Get told when stock crosses the reorder point, on the day, not at the end of the month.

The rest of this page works through each one in the order you need it.

It looks unfair, and it is completely predictable.

First, demand concentrates. A minority of your catalogue usually carries the majority of your orders, so a handful of SKUs are burning inventory at a pace the rest of the catalogue never approaches. Forty units is six weeks of cover on a slow mover. On a hero SKU it can be four days.

Second, your own marketing points at the same products. Paid traffic, email, the bundle on the cart page, the product the creator mentioned. You spend money driving demand into the exact SKUs with the least cover, which pulls the reorder date forward while your supplier lead time stays fixed.

Third, attention does not scale. When you have a few hundred SKUs you can hold the risky ones in your head. Somewhere past that, the ones you check become the ones you are already worried about, and the quiet failure is the product that was fine on Friday and gone by Wednesday.

The cost is not only the orders you did not take. You paid for traffic that landed on an out of stock page. The customer who wanted it found somewhere else that had it. Your sales channels reward availability. A stockout on a hero SKU is a real financial hit, and it deserves a process rather than good intentions.

Every reorder decision reduces to four figures per SKU. Get these and the rest is arithmetic.

Sales velocity

Units sold per day, averaged over a window you pick. Divide units sold in the window by the number of days in it. A SKU that shipped 168 units over the last fourteen days is running at 12 a day.

Lead time

Days from the moment you place the order to the moment the stock is sellable on your shelf. Not the factory's production time alone. The whole chain.

Reorder point

Lead time demand plus safety stock. In plain terms: velocity multiplied by lead time, plus the cushion you decided to hold.

reorder point = daily velocity x lead time in days + safety stock

Take the SKU above as an example. Twelve a day, a twenty-one day lead time, and a cushion of sixty units. That is 12 x 21 = 252, plus 60, so the reorder point is 312 units. When stock touches 312, the order goes out. Not when it looks low.

Days of inventory remaining

Current stock divided by daily velocity. It converts a unit count into the only unit that matters, time. Four hundred units at twelve a day is roughly thirty-three days of cover. Against a twenty-one day lead time, you have about twelve days of slack before the decision becomes urgent.

Units tell you how full the shelf is. Days tell you whether you are late.

The window you average over changes the answer, and there is no universally correct choice. A short window reacts fast and overreacts to noise. A long window is stable and slow to notice that something has changed.

The practical guide is to match the window to the length of your lead time and the volatility of the product. If your supplier takes six weeks, a seven day spike should not send you into a panic order. If you are restocking locally in five days, a thirty day average will have you ordering last month's demand.

One caution that catches people out. If a SKU was out of stock during part of the window, its measured velocity is artificially low, because it could not sell what it did not have. That drags the reorder point down, which makes the next stockout more likely, which drags it down again. When you review, exclude or flag the days a SKU was unavailable.

Merchants tend to quote the factory's answer. The factory's answer is one part of the chain.

A realistic lead time counts every day between your decision and sellable stock:

• Time to raise and confirm the purchase order, including waiting for your own approval.
• Supplier production or picking time.
• Freight, including the gap between ready and actually shipped.
• Customs and inland transport if you are importing.
• Receiving, counting, inspection and putaway at your warehouse or 3PL.

That last bucket is invisible in most planning and can add days, especially if your 3PL books receiving slots. Stock in a box at the dock does not fill orders.

Use the worst recent case rather than the average for anything critical. Averages are the right tool for cost. Worst case is the right tool for risk, because you only feel the delay in one direction. Nobody complains about a shipment arriving early.

Also keep lead time per supplier, and per SKU when it differs. A single global figure applied across a catalogue with a domestic printer and an overseas manufacturer will be wrong for both of them.

Safety stock is the cushion that covers the two things you cannot predict: demand running hotter than your average, and the shipment arriving later than promised.

The honest framing is a trade. Every unit of cushion is cash sitting on a shelf, storage you pay for, and a risk of markdown if demand fades. Every unit you decline to hold is exposure on a day when the numbers go against you. There is no setting that removes both.

So size it by consequence, not by uniform rule:

• Hero SKUs, the ones your paid traffic points at, earn a generous cushion. Being out is expensive in ways beyond the missed order.
• Long lead time imports need more cushion than local restocks, because you cannot recover quickly from a miss.
• Volatile sellers need more than steady ones. If weekly sales swing widely, plan against a bad week rather than an average one.
• Slow tail SKUs often deserve almost none. Holding cover on a product that sells twice a month ties up cash for nothing.

A workable starting point is to express the cushion in days of cover, decide how many days of surprise each tier of product is worth protecting against, then convert to units at current velocity. Review it when velocity or lead time changes, not on a fixed calendar.

Most stores that run out of stock have a process. It has just quietly stopped matching reality.

One global low stock number

A single threshold across the catalogue, say ten units, treats a product selling thirty a day the same as one selling three a month. For the fast mover, ten units is a few hours of warning against a three week lead time.

Thresholds set once and never revisited

A reorder point calculated when a SKU sold five a day is dangerous once it sells fifteen. Velocity moves. If the number does not move with it, the number is fiction.

Alerts nobody owns

If a notification goes to a channel everyone can see and nobody is responsible for, it will be read and not acted on. Name the owner.

Alert fatigue

If everything alerts, nothing does. Keep thresholds tight enough that a notification means a purchase order, and keep the noisy tail on the weekly report rather than in real time.

Counting the wrong units as available

Stock allocated to unfulfilled orders, held for a wholesale account, or damaged in receiving is not sellable. If your available figure includes it, your cover is shorter than you think.

Treating the supplier's promise as fact

The lead time you plan with should be the one you have actually experienced, not the one on the pro forma invoice.

Two jobs, in this order: limit the damage, then stop the repeat.

To limit the damage, be straight with customers on the product page. A clear restock date, where you genuinely have one, holds more demand than a silent out of stock badge. Capture back in stock interest so you know what the outage actually cost you in wanted orders. Point paid spend away from the SKU while it is unavailable, because that budget is currently buying disappointment. Then decide honestly whether expedited freight is worth the premium on this specific product, or whether you take the gap and protect margin.

To stop the repeat, do the post mortem while it stings. Write down the date stock hit zero, the date the order was placed, and the date the replacement landed. Compare the lead time you planned with the lead time you got. Compare the velocity you planned with the velocity you saw. One of those two numbers is almost always where the miss came from, and it will keep being wrong until you change it.

And flag the outage in your sales history. Those zero-sales days are not a demand signal, they are an availability gap, and left in the average they will quietly lower the reorder point on the product you can least afford to lose again.

The rhythm to keep

Weekly review of everything at or near its reorder point. Lead times updated after each delivery. Velocity re-checked after promotions and launches. Alerts pointed at one owner. That is the whole discipline. It is not complicated, it is just easy to drop, which is why it belongs in a system rather than a memory.