Reference
Planning stock you
haven't sold yet
The arithmetic of inventory planning is one short formula, standard practice in operations for decades. You can do it on paper for a single line in about a minute.
The inputs are the hard part. Every one of them is an estimate, and each leans in a known direction. Feed a planning system comfortable inputs and it produces confident numbers that are wrong in exactly the direction the person who supplied them was hoping.
The reorder point is a moment, not a stock level
Planning answers when to place the order. The answer is the amount you'll sell while you're waiting for it, plus a cushion for the times either of those estimates is wrong.
The standard formula
Reorder point = (average daily demand × lead time in days) + safety stock
That's the whole of it. When stock on hand plus stock already incoming falls to that number, the order goes. Everything a planning module does beyond this is either working out the three inputs more carefully, or doing it for ten thousand lines instead of one.
A worked example — invented figures, for illustration only
Suppose a line sells 8 a day on average, the supplier takes 21 days, and you hold 60 as a cushion. The reorder point is (8 × 21) + 60 = 228. At 228 units of stock on hand and on order, you buy: not at 60, and not when the shelf looks empty.
Now let the supplier slip to 30 days, which suppliers do. The reorder point is (8 × 30) + 60 = 300. Still buying at 228, you now stock out on your best line every time the supplier is late. Nothing in the storefront reports it, and the first symptom is a gap.
Three inputs, and how each one lies
The formula is only as good as these, and each fails in a known direction. Knowing which way each one leans is most of the skill.
Average daily demand
Usually taken from sales history, which records what you sold, not what customers wanted. Every day the line was out of stock reads as a day of zero demand, so the average is biased downward by exactly the periods when demand was highest.
The fix is to exclude out-of-stock days from the average rather than count them as zeros, which means recording when you were out.
Lead time
Usually the number the supplier quoted, which is their best case. The number you want is your own measured average from order placed to goods available to sell, including the customs step, the delivery slot you missed, and the two days it sat in goods-in before anyone booked it.
The gap between quoted and measured can be the largest error in the whole calculation.
Safety stock
The cushion, and the one people set by feel. It's a decision about how often you're willing to run out: a higher cushion buys a lower stockout rate and costs cash sitting on a shelf.
Set it per line rather than as one global number: a signature product and a slow accessory don't deserve the same protection, and treating them alike is how cash ends up in the wrong stock.
Four situations where the formula stops working
It assumes demand is roughly steady and history is a guide. Where that doesn't hold, more arithmetic won't rescue it, and the answer is judgement, with the numbers as one input rather than the decision.
Seasonality
A trailing average walks into a peak underweight and out of it overweight, every year, because it's always describing the season that just ended. The usable version compares the same weeks last year rather than the last few weeks.
A line with no history
New products have no demand data by definition, so the first buy is a judgement and should be named as one. The useful discipline is deciding in advance what result would change the second order.
Lumpy demand
A line that sells nothing for three weeks and then forty to one trade customer has an average of about two a day and never sells two in a day. The average is arithmetically correct and operationally meaningless.
An unreliable supplier
When lead time swings between 14 and 60 days, no single figure represents it, and safety stock is being asked to absorb a commercial problem. That's a conversation with the supplier, not a bigger cushion.
All of this depends on the incoming figure being real, which is one of the four numbers a stock system has to hold. stock control software covers which system holds them. If the buying record itself is the constraint, what Shopify already covers is the narrower read.
When a spreadsheet is the right answer
For a few dozen lines, the formula in a spreadsheet with properly measured inputs will outperform a planning module fed the supplier's quoted lead time and an average that counts stockouts as zeros.
What software buys you is doing it for thousands of lines, keeping the inputs current without anyone remembering to, and noticing the ones that moved. At that scale, it earns its cost.
The cheapest improvement you can make: measure real lead times for a quarter, and replace every quoted figure with the measured one.
Start the systems reviewWhen to add the statistical layer
Service-level targets, demand variability and safety stock derived from them are established practice, and useful above a certain catalogue size.
They pay off once the measured inputs above are real, so that's the order we put them in: measure first, then model, then choose the planning tool that fits the catalogue.
Tell us what you need
What the systems are doing now, and what you need them to do. We'll tell you whether it's a platform problem or something cheaper.