When to reorder: turning sales velocity and lead time into an order date

Reorder a SKU when the stock on your shelf will only just cover demand for as long as the next delivery takes to arrive, plus a buffer you choose on purpose.

Two ways to write the same rule:

• In units. Reorder point = average units sold per day × (lead time in days + buffer in days). When stock on hand touches that number, place the order.
• In days. Order when days of stock left falls below lead time plus buffer days.

The second version is the one to run your week on, because it is already a date. If a SKU sells 11 a day, your supplier takes 21 days from purchase order to receiving, and you want a 7 day cushion, your reorder point is 11 × 28, which is 308 units. Hold 300 and you have 27 days of cover against a 28 day requirement. You are one day late, today.

That is the arithmetic. The rest of this guide is about making the three inputs honest, because a reorder point built on a wrong lead time is worse than no reorder point at all: it tells you that you are safe right up to the day the shelf empties.

Most merchants first meet this as a threshold: type 20 into a box, get an email when stock goes under 20. That works until two things change, and they always change. Velocity moves, so 20 units is three weeks of cover in February and four days in November. Lead time moves, so 20 units is fine for a local printer and hopeless for a container from overseas.

A reorder point fixes that by being derived rather than typed. It is not a number you like, it is a number your own sales and your own supplier produce. And once it is derived, you can turn it into something better than an alert: a date.

Take days of stock left, subtract lead time, subtract your buffer days, and what remains is how many days you have before the order has to go out.

Order in (days) = days of stock left − lead time days − buffer days

With 300 units selling 8 a day, that is 37 days of cover, minus 21 days of lead time, minus a 7 day buffer, so the order goes out in about nine days. That is a line in your diary instead of a vague worry. It also tells you the thing a threshold never does, which is whether you are early, on time, or already behind.

The window you average over is the single biggest lever on the answer, and there is no universally correct choice. It is a trade between reacting fast and reacting to noise.

7 days

Sharp and twitchy. It catches a SKU that started moving this week, which is exactly when you want to know. It also overreacts to a good Saturday, a discount, or one wholesale order that will not repeat, so it can push you into ordering more than the trend supports.

14 days

The usual compromise. Long enough to absorb one strange day, short enough to notice a real shift within a week or two. For most stores with steady weekly patterns this is a sensible default.

30 days

Calm and conservative. Good for slow, steady sellers where a few units a week is normal and daily noise would be meaningless. Bad during a ramp, because a month of history drags the average down and hides the fact that last week doubled.

A practical approach: run a longer window as your baseline, and check the short window on your best sellers. When the 7 day number is well above the 30 day number, treat the short one as the truth and order against it, because you are on the way up and the average is behind you.

Lead time is the input merchants get wrong most often, and the error is almost always in the same direction: too short. The supplier says fourteen days, so fourteen days goes in the sheet. The fourteen days was production time, and it started when they acknowledged the order, three days after you sent it.

Count the whole chain:

• Your own approval and the day you actually raise the purchase order
• Supplier acknowledgement and any deposit that has to clear first
• Production or picking time
• Transit, including the customs step if it crosses a border
• Receiving at your end, counting in, and getting the quantity into Shopify

The last one is invisible and often costs two or three days. Stock in a box in the corridor is not stock you can sell.

If your last few orders for a supplier took 19, 24 and 22 days, do not average to 21 and feel clever. Use something close to the slowest realistic case, or keep 21 as lead time and let your buffer absorb the variance. Either is defensible. What is not defensible is using the number on the supplier's website.

Set lead time per SKU, not per store. The same supplier can be two weeks on a stocked item and ten weeks on a custom print run, and one figure across both will have you late on half your catalogue.

Formal safety stock formulas use the standard deviation of demand and of lead time, multiplied by a service level factor. They are correct, and for a store with a few hundred SKUs and patchy history they are usually more precision than the data can support.

The version that survives contact with a real week is buffer days: decide how many extra days of cover you want, and let it scale with velocity automatically. Because the buffer is multiplied by sales per day, a SKU that speeds up gets a bigger cushion in units without you touching anything.

How to pick the number:

• Reliable supplier, steady demand, easy to reorder: 5 to 7 days.
• Variable lead time, or a supplier who has slipped on you: 10 to 14 days.
• A hero SKU, where being out is the expensive outcome: more, and accept the extra cash tied up as the price of not disappointing your best customers.
• Slow movers and long tail: less, or none. A buffer on an item selling two a month is just money on a shelf.

Buffer days are a spending decision, not a technical one. Every extra day of cover is cash sitting in stock instead of in the bank, and running out is margin you cannot recover on demand you already had. Merchants who catch the reorder point early tend to avoid the second problem without overpaying for the first.

Knowing when does not tell you how many. The order needs to cover demand through the lead time and through the period until your next order goes in, and it needs to credit anything already on the water.

Suggested quantity = sales per day × (lead time days + cover days) + buffer − stock on hand − stock already inbound

With 11 a day, a 21 day lead time, 30 days of cover between orders, a 7 day buffer and 300 units on hand and nothing inbound: 11 × 51 is 561, plus 77 of buffer is 638, less 300 on hand is 338 units. Round to the supplier's case pack of 24 and you order 14 cases, so 336.

Then adjust for the real world:

• Minimum order quantities. If the MOQ is 500, you are choosing between over-ordering and not ordering. Decide deliberately, and consider stretching the cover period for that supplier instead.
• Consolidation. If freight is priced per shipment, pull forward SKUs from the same supplier that will hit their point in the next couple of weeks. A slightly early order that shares a container often beats two separate ones.
• Cash. The right quantity you cannot pay for is the wrong quantity. Cut cover days before you cut safety on your best sellers.
• Shelf life and seasonality. Never order past the point where the stock stops being sellable.

New SKUs with no history

There is no velocity to average, so the formula returns zero and the SKU looks safe forever. Borrow a velocity from the closest comparable product for the first few weeks, or set a manual floor, and revisit once real sales exist.

A launch or a promotion

A discount pulls demand forward. The velocity you measure during the promotion is not the velocity that follows it, so do not lock a reorder point to a promotional fortnight without adjusting.

Seasonal peaks

A trailing average is always behind a ramp. Going into a known peak, either shorten the window so it reacts faster, or lift buffer days on the SKUs you know will spike, and place the order earlier than the arithmetic says.

Zero sales and slow movers

A SKU with no sales in the window will never cross a calculated point. That is usually correct and occasionally wrong, for example an item that sells four times a year in batches. Keep a short manual list for those.

Stock that is already at zero

The question is no longer when to order, it is how fast, and whether to keep the SKU. Order, then look at how many days it was out, because that is the part worth fixing.

Multiple locations

Units split across locations can look fine in total while one location is empty. If you fulfil from one place, calculate on that place, not on the store total.

Daily checking is how this dies. Nobody opens a spreadsheet of 600 SKUs every morning for long, and the value is not in seeing the numbers, it is in acting before the order date passes.

A weekly rhythm holds up:

1. Pick a fixed day, ideally one that matches when your suppliers like to receive orders.
2. Pull the list of SKUs at or under their reorder point, with days of cover and a suggested quantity next to each.
3. Group by supplier so one review produces one order per supplier.
4. Sanity check the three or four largest lines by eye, and send.
5. Note anything that surprised you, because a SKU crossing its point earlier than expected is usually a velocity change worth knowing about.

Build the buffer to fit that cadence. If you only order on Mondays, your effective lead time includes up to six days of waiting for the next Monday, so add those days rather than pretending the order goes out the moment the point is crossed.

A spreadsheet can do all of this, and plenty of good stores run exactly that way. It breaks in two places: someone has to refresh the sales data, and someone has to remember the day. Both are silent failures, and you find out about them from a customer trying to buy something you do not have.