Building a restock rule instead of restocking by feel

What restocking by feel actually costs

In most small stores, reordering happens when someone notices stock looks low and has time to place an order. That produces two failure modes that mask each other.

Too late on fast movers, because a product that sells steadily crosses your mental threshold between the moments you happen to look. You notice at the point where it is already going to run out before the delivery lands.

Too early on slow movers, because a product that has not sold in a while is visible every time you look at the shelf, and ordering it feels like being organised. Cash goes into stock that will sit for months.

Net effect: out of stock on the things that make money, overstocked on the things that do not. Both feel like bad luck rather than a system problem, which is why they persist.

The four numbers

You need these per product, and you probably have three already.

1. Daily sales rate. Units sold divided by days, over a window long enough to smooth noise. Thirty days is usually right; ninety if the product is slow or seasonal. Exclude obvious distortions like a one-off bulk order or the week it was in a campaign.

2. Lead time. Days from placing an order to stock being sellable. Measure the whole chain, including the parts that are not your supplier: their dispatch, transit, customs if relevant, and how long it sits in your receiving area before anyone books it in. That last one is routinely omitted and routinely material.

Use your worst recent lead time, not your average. Averages describe what usually happens; you are protecting against what sometimes happens.

3. Safety buffer. Extra days of cover for variability. Start at seven days for reliable suppliers on steady products, fourteen where either the supplier or the demand is erratic.

4. Minimum order quantity. Whatever your supplier will actually accept, which sets the floor on how much you order once triggered.

The rule

Reorder point = daily sales rate x (lead time + safety buffer)

A product selling 4 a day, with a 21-day worst-case lead time and a 7-day buffer, has a reorder point of 4 x 28 = 112 units. When stock hits 112, you order. Not when it looks low, not when you remember, at 112.

The point of the formula is not precision. It is that it converts a judgement made under distraction into a number decided once while thinking clearly. A roughly right threshold applied consistently outperforms a carefully considered decision made at the wrong moment.

How much to order

The reorder point tells you when. Order quantity is a separate decision and the constraints are different, because this one is about cash rather than availability.

A workable default is enough to cover the lead time plus a review period, subject to the minimum order quantity. For most small stores a review period of thirty to sixty days is sensible: long enough to avoid constant small orders, short enough that you are not funding six months of a product that may stop selling.

Adjust for three things. Shelf life, obviously, where it applies. Cash, which is the binding constraint far more often than anyone admits: the theoretically optimal order is irrelevant if it leaves you unable to pay for something else. And discount thresholds, treated sceptically, because a supplier discount for ordering triple is a real saving only if you sell it all, and it is a cash trap if you do not.

Segmenting, because one rule does not fit

Applying identical logic to every product is what makes thresholds feel wrong. Split the catalogue three ways.

Core products that sell steadily and matter to the business. Full calculation, larger buffer, and never let these go out of stock. If you only do this exercise for one group, do it for these.

Long tail that sells occasionally. Do not compute a daily rate on something selling twice a month; it produces meaningless numbers. Set a simple minimum, reorder when it hits, and accept occasional stockouts as the cost of not tying up cash.

Seasonal or promoted products, where historical rate is actively misleading because the future will not resemble the past. These need a planned buy rather than a reorder point, and they are where running a pre-mortem before a sale does more good than any formula.

Where the rule breaks

Be honest about the limits, because a rule trusted past its range is worse than none.

New products. No sales history means no rate. Order a small quantity, watch for a few weeks, then calculate.

After a campaign. The thirty days following a promotion are distorted in both directions. Exclude the promotional period from your rate rather than letting it inflate everything downstream.

Variants. Rates should be calculated per variant, not per product. A jumper selling well overall may be selling entirely in two sizes, and a product-level reorder point will keep you stocked in colours nobody wants.

Supplier changes. A new supplier invalidates your lead time until you have measured theirs. Use a pessimistic figure for the first few orders.

A reorder point is not a forecast. It is a decision you make once, calmly, so that a busy version of you does not have to make it badly.

Making it fire without vigilance

A reorder point nobody sees is arithmetic in a spreadsheet.

Set the alert threshold in whatever system watches your stock, per product, at the number you calculated rather than at one flat figure across the catalogue. Send it to whoever actually places orders, and nobody else, per the routing logic in who should see inventory alerts.

Then review the numbers quarterly. Sales rates drift, suppliers get faster or slower, and a reorder point calculated a year ago is describing a business that has changed. The review takes twenty minutes and is the difference between a system and a spreadsheet somebody made once.

A worked example

The arithmetic is easier to trust once you have seen it run once.

Take a product selling steadily. Over the last thirty days it sold 96 units, which is 3.2 a day. Exclude nothing, because there was no promotion in the window.

Your supplier quotes ten days. Your last three orders actually took 12, 14 and 19 days from placing to sellable, with the 19 including four days sitting in receiving before anyone booked it in. Use 19, not the ten they quoted and not the 15 you would get by averaging.

The supplier is reliable and demand is steady, so a seven-day buffer is enough. Reorder point is 3.2 x (19 + 7) = 83 units.

For quantity, cover the lead time plus a thirty-day review period: 3.2 x (19 + 30) = 157, rounded to whatever the supplier's case size makes convenient. If their minimum is 200, you order 200 and accept slightly more cover than the model wants.

The instructive part is the lead time. Using the quoted ten days would have produced a reorder point of 54, and you would have run out for roughly a week on every cycle while believing your maths was sound. The quoted figure is a sales number; the measured one is the truth.

Where the buffer actually comes from

The safety buffer is not padding. It is covering the fact that both inputs are estimates.

If your sales rate is steady and your lead times are consistent, both estimates are good and a small buffer is fine. If either is erratic, you are carrying more uncertainty and the buffer is what absorbs it.

So the buffer should track variability rather than importance. A product you care about deeply but which sells at a metronomic rate from a reliable supplier does not need a large buffer. An unremarkable product with lumpy demand and a supplier who sometimes disappears for a month does. Teams usually get this backwards, sizing buffers by how much they would mind rather than by how unpredictable the inputs are.

When the rule says order and the cash says no

This is the situation the arithmetic does not cover, and it is common enough to plan for.

The reorder point has triggered on four products at once, and you cannot fund all four. Ordering by whichever alert arrived first is the default and it is close to random.

Rank instead by what running out actually costs. Three questions, quickly. How much margin does this product contribute over the cover period? Would being out of stock affect anything else, such as a promotion already booked or a bundle it appears in? And how long is the lead time, since a long-lead product ordered late stays out much longer.

That last question does most of the sorting. A product with a three-week lead time missed today is unavailable for a month; a short-lead product can wait a fortnight and cost you almost nothing.

Write the ranking down when you do it. The same four products will collide again next quarter, and having last quarter's reasoning to hand turns a stressful decision into a five-minute one.

Common questions