Retail: how much to order per branch, with the data you already have

≈ 2 min read · updated Jul 2026

Each branch orders by eye: one has too much, another runs out, and every stockout is a sale lost without leaving a record. The study forecasts demand per product and branch, and builds each replenishment order with your budget and your space as the limits.

Retail · forecast + replenishment
One product, three branches: where sales are lost and how much to order
How to read it: on the left, weekly sales of the same product in three branches and the stockout where the shelf ran empty · on the right, the order the model builds for the next replenishment
A · One product, 12 weeks, three branches vertical: units sold per week · horizontal: weeks high low wk 1 wk 6 wk 12 Downtown North East estimated demand → stockout sales were lost here B · The recommended replenishment bars: how much to order · illustrative East branch order more ↑ restocks the gap and covers the rise North branch order less ↓ selling slower and stock covers it Downtown branch do not order warehouse still full after the peak finite budget and shelf space: the model picks what to prioritize
Downtown branch North branch East branch Demand estimated by the model Sales lost in the stockout
  • Forecast per branch · nowcasting
  • Replenishment · integer optimization
  • Data: your sales and stock
In simple terms: the same product moves differently in each branch; when the shelf runs empty the sale is lost without leaving a record, and the order on the right spends the budget where the forecast calls for it. Illustrative curves of the method: the real model is fed with your sales.

In a chain with several branches, each manager builds the replenishment order by eye, with memory and a spreadsheet. The result is in the chart above: the same product piles up in one branch and runs out in another.

The sale you never see

Overstock is easy to spot: money sitting in the warehouse and, in pharmacy or food, product that expires. A stockout goes unnoticed: the customer finds an empty shelf, buys somewhere else and the system records nothing. That lost sale stays out of every report, and what never shows up in a report never gets managed.

Each store has its own rhythm: payday, the weather, the long shift, the construction site across the street. Ordering the same amount for every branch guarantees the imbalance.

From sales history to the order

The study builds two pieces that work together, because forecasting well is half the problem: the other half is deciding what to buy when the money and the space are not enough for everything. The forecast looks at the coming weeks, the horizon a replenishment order needs; the optimization turns that expected demand into the concrete order, with the purchasing budget and the shelf and warehouse space as hard limits.

The data: sales per ticket, stock per branch and the calendar (holidays, paydays, campaigns). It works on top of your current system: to start, exporting us an Excel is enough.

The technique: demand forecasting per product and branch (nowcasting with seasonality and events) and integer optimization of the replenishment.

The decision: how much to order, for which branch and when; and what to stop ordering while the warehouse runs down.

Not every SKU deserves the same effort. The ABC classification separates the few that concentrate the purchase value, and for each A SKU the forecast also sets its safety stock.

Retail · ABC + safety stock
A few SKUs rule: ABC by value and the buffer set by the wait
How to read it: on the left, how much of the purchase value each class concentrates · on the right, the safety stock of an A SKU arriving by ship from China versus a food SKU that expires
A · ABC classification bar height: % of purchase value · illustrative ≈15% of the SKUs ≈75% of the value ≈30% of the SKUs ≈20% of the value ≈55% of the SKUs ≈5% of the value A B C a few SKUs concentrate the purchase value B · The safety stock of two A SKUs same method, different wait · illustrative Home appliance from China more buffer ↑ arrives by ship: lead time of ~60 to 90 days cycle buffer the long wait grows the buffer Food product less buffer ↓ short replenishment and shelf-life expiry cycle small buffer it turns faster and holds less: expiry sets the limit same demand forecast for both SKUs the difference comes from the wait and the expiry
Class A Class B Class C Cycle stock Safety buffer
  • ABC classification · Pareto
  • Forecast + safety stock
  • Data: your sales and lead times
In simple terms: the ABC classification ranks the effort and the buffer of each A SKU comes from the classic formula: z times the standard deviation of demand times the square root of the lead time. Waiting for a ship from China grows that buffer; a product that expires shrinks it. Illustrative example of the method: the proportions do not come from real data.

Who this is for

  • The pharmacy chain that restocks per branch and loses sales in the busiest ones.
  • The supermarket or convenience store with a central warehouse and shelves that never balance out.
  • The distributor or hardware store that orders from suppliers weeks in advance.

How to measure it: two numbers per branch, on a simple dashboard: avoided stockouts and days of stock. If the dashboard shows no improvement, the model is not doing its job.

We put this analysis out to be argued with. If you read the balance between overstock and stockouts differently, or you run a chain where orders are still built by eye, write to us with an Excel of your sales and we will take the conversation to your numbers.

Send us an Excel of your sales →

Initial diagnosis at no cost · we reply within 24 business hours