Reorder point: when and how much to order

Thursday morning, 9:20. The warehouse lead calls: there is not a single unit of SKU-3388 left on the shelf. The marketplace dashboard, meanwhile, shows four units sold the previous evening and two more waiting to be picked. The buyer phones the supplier and hears "ten business days." Then someone recalls that the same thing happened last month, though nobody can name the date. The real problem here is neither the supplier's lead time nor the warehouse team's attention. Nobody has a written, shared answer to a simple question: at what quantity do we place a new order for this item?

This article is about building that answer. We start with what a reorder point actually measures, then work through where each term in the formula comes from, how to measure lead time honestly, and what safety stock is really protecting you from. From there we move to the "how much" side: the constraints that set order quantity, which items deserve a threshold of their own, what should happen the moment a threshold trips, the data errors that quietly break the calculation, and how the picture changes once you sell across several channels.

What does a reorder point actually measure?

A reorder point is not the stock level you want to hold. It is a trigger: when available quantity drops below that number, it is time to raise a purchase order. And what it really measures is not units but time — it is the question "how much will I sell before the replenishment arrives?" converted into units.

Missing that distinction is a common mistake. Teams write down a "minimum stock" figure and keep it fixed regardless of how fast the item moves. Give the same minimum to something that sells forty units a week and something that sells three units a month, and the first will stock out repeatedly while the second ties up cash on a shelf. The same number means two different things for two items, because the trigger depends on sales velocity.

The other half of the idea is that a threshold is a reminder, not a decision: it tells you to order now, and says nothing about how many. When and how much are separate calculations, and blending them is the fastest route to solving stockouts by filling the warehouse with dead inventory.

Where does each term in the formula come from?

The basic form is simple: reorder point = average daily sales × lead time in days + safety stock. The first product tells you what you will consume while the goods are in transit. Safety stock is the allowance you leave for the days when that estimate does not hold. All three terms come from different places, and each of them can be wrong on its own.

Average daily sales is computed over a consumption window you choose — say the actual outbound movements of the last 60 or 90 days. Too short a window and a single promotional week drags the average upward; too long and you notice a seasonal turn late. What you are doing here is forecasting, and measuring how well the forecast held is no different from the discipline sales teams apply when they check how closely their pipeline forecast matched reality: if you never record the variance, you can never correct it.

A worked example — the numbers below are illustrative figures invented to show the arithmetic, not an industry benchmark. Say average daily sales is 8 units, lead time is 12 days, and safety stock is 30 units. The threshold works out to 8 × 12 + 30 = 126 units, and an order goes out the moment available quantity falls under 126. If the same supplier's lead time stretches from 12 days to 18, the threshold rises to 174. Nothing looks different from the outside, but 126 units is now too thin a cushion.

How long is the lead time, really?

The number a supplier quotes on the phone is usually only their own preparation time. The number that belongs in your calculation runs from decision to shelf, not door to door: internal approval of the requisition, sending it to the supplier, their picking and packing, transit, customs clearance, goods receipt and counting, and the moment the item finally becomes sellable. A two-day approval delay sitting in the middle of that chain is part of the lead time and belongs in the threshold.

Lead time should therefore be a measured number, not a remembered one. The Supplier Performance view in Stock Management does exactly that: it keeps order and delivery counts, total spend, and average delivery time per supplier. A vague sense that "this vendor runs late" becomes a day count you can write into a threshold. The logic is the same one a sales team applies when it tracks how long its deals actually take to close — without the average, no plan is grounded.

Look at the spread as well as the average. If nine of ten purchase orders arrived in 10 days and one took 25, the mean on its own is misleading. In practice, keep two numbers: a typical lead time for the multiplication, and a worst-case one for the safety stock calculation.

What is safety stock actually protecting?

Safety stock is not a vague "just in case" cushion. It covers two specific risks. The first is demand variability: selling more than expected during the lead time. The second is lead time variability: the goods not arriving on the promised day. When both land at once a stockout is close to certain, because you have neither spare days nor spare units.

A simple and defensible approach is this: safety stock = (maximum daily sales × maximum lead time) − (average daily sales × average lead time). In other words, you hold the gap between the bad case and the normal case as inventory. It is less precise than a statistical service-level model, but every input is already in your own records.

The decision itself is a trade between service level and tied-up cash. Raising safety stock lowers the odds of a stockout and locks money onto a shelf. So the cushion should scale with how critical the item is rather than being applied uniformly. The reasoning matches what sales planners do when they work out how much pipeline is needed to hit a quota: too little buffer and hitting the target is luck, too much and you have simply wasted the resource.

How much to order: the constraints that set quantity

When a threshold trips, many teams order just enough to climb back above it, which produces a stream of tiny purchase orders. A healthier approach is to require the order to cover two periods at once: the incoming goods should carry you through the lead time and through the interval until the next review. Otherwise you drop below the threshold again almost as soon as the delivery lands.

Then the practical constraints arrive: supplier minimum order quantity, case and pallet multiples, free-freight thresholds, shelf life and warehouse space. Unit of Measure Conversion helps precisely here — stock is held in a single base unit while you can purchase in cases or packs, so a "we need 126 units" calculation rounds cleanly to whole cases and nobody converts between units by hand.

Quantity is a cost decision too. Spread freight, duties and handling across the units and the per-unit cost of a larger order drops; Landed Cost performs that allocation and shows your true unit cost and margin. How smaller, less formal purchase-side costs get recorded is a separate discipline, and knowing when an expense voucher is issued keeps those costs from disappearing from the books.

Which items deserve a threshold of their own?

Measuring lead time and computing safety stock individually for several hundred SKUs is not realistic. The practical route is to classify inventory by value and point the effort where it pays. ABC Analysis splits your stock into A/B/C classes by value and shows where your capital is locked up; the priority order for threshold work falls straight out of that ranking.

The table below is an example working rhythm rather than an industry norm; adapt it to your SKU count, your supplier count and the capacity of the team.

Class	Typical position	Threshold review	Safety stock approach
A	Most of the value, few of the SKUs	Monthly	Calculated per item, lead time measured per supplier
B	Mid value, mid SKU count	Quarterly	Category average, corrected by hand where it clearly drifts
C	Low value, many SKUs	Twice a year	Fixed threshold, larger orders placed less often
New item	No consumption history	Weekly for the first quarter	Start from a comparable item, update once sales settle

Classification pays a second dividend at counting time: building a count sheet by warehouse and ABC class lets you count A items often and C items rarely, raising accuracy without shutting the warehouse down. Threshold arithmetic rests on accurate stock data anyway — if the system says 40 and the shelf holds 26, how well the formula is built stops mattering.

What should happen when a threshold trips?

A threshold written in a spreadsheet is not enough; somebody has to see it in time when available quantity crosses the line. Stock Alerts scan for critical stock, stockouts and overstock every fifteen minutes and raise in-app and email notifications. The subtlety is not whether the alert exists but who receives it and how often.

An alert that goes to everyone and repeats several times a day gets ignored within a fortnight. Routing threshold alerts to the buyer, digesting low-value items daily and suppressing repeats for the same SKU all help — the same reasoning teams apply when they retune their CRM alerts, transplanted to the inventory side.

The path from alert to action should be short. The Items at Risk view builds a prioritised queue from demand velocity, available quantity, lead time and safety stock assumptions: which item runs out in how many days, and which one has already fallen below its reorder point. From a row in that queue you can open a draft purchase order directly, and narrow the scope with warehouse and assumption filters. Whether a threshold earns its keep usually comes down to how many steps sit between the alert and the order.

Which data errors quietly break the calculation?

Even a correct formula produces a wrong answer on broken inputs, and these failures tend to be silent. The most common is confusing available stock with physical stock. Units sitting on the shelf but reserved against another order are not sellable, so the threshold comparison must run against available quantity net of reservations. The Stock Health dashboard brings available stock, reserved quantity, consumption and inactivity signals into one table and flags rows where available stock has gone negative.

The second frequent error is ambiguity about which document moves the stock. In a flow where shipping is tied to a delivery note, stock moves when the note is finalised, and the invoice raised afterwards only produces the bookkeeping entry without touching stock a second time. Stock and Accounting Reconciliation lists the places where the two ledgers have diverged, and once invoiced-but-not-deducted records start piling up, thresholds fire at the wrong moment too. This is inseparable from how the wider order management flow is set up.

A third group of errors comes from product structure. For kits and bundles, sellable quantity depends on component stock; if you hold separate thresholds for the component and for the kit without linking them, the same goods get counted twice. Returns need care as well: a return should convert to a stock receipt only for the quantity actually shipped, otherwise goods that never existed push the threshold up. The durable fix is counting: when a legitimate movement hits a line after a count session has opened, the adjustment applied at completion is shown separately, so the gap between the frozen sheet and live stock cannot vanish quietly.

How do thresholds change across channels and in B2B?

Selling from one warehouse into one channel, the threshold is a single number. Open the same stock to several marketplaces and the picture shifts: demand velocity is the sum of the channels, but the cost of a stockout differs by channel. If stock sync lags, the quantity shown on a channel drifts from the real quantity and the threshold trips late, which is why part of the safety stock in a multi-channel setup exists purely to absorb sync delay.

Volatility is harsher in B2B, because a single dealer order can consume a month's average in one line. Looking only at average daily sales is misleading here; known dealer programmes and repeating order patterns need separate handling. When an order portal you open to your dealers makes that demand visible in advance, the threshold works from a plan instead of a surprise.

The ordering channel itself is a data source. A customer may place an order through a form, a portal or a conversation, and running forms and ordering inside the chat is one way to catch demand at the moment it forms. Nor do you have to lose the order outright when stock runs short: Backorder Management records unfulfilled demand and closes it automatically when the goods arrive, so real demand from the stocked-out days still enters the data.

Seeing where demand originates feeds planning too. Tracking which campaign and which touchpoint turns into an order — the work of mapping the customer journey — pays off on the inventory side as much as in marketing: a team that knows demand is coming can order ahead of the threshold instead of waiting for it.

How often should thresholds be reviewed, and where do you start?

A threshold calculated once goes stale within a few months. Lead times move, an item enters its season, a campaign doubles demand. So a threshold is not a setup task but a calendar-driven review: high-value items monthly, the rest less often according to class. Three things get checked at each review — did we stock out recently, has the average lead time shifted, and has more stock accumulated than expected.

Campaigns deserve separate treatment, because they bend demand in ways historical data cannot predict. When a reactivation campaign aimed at cold customers is scheduled, thresholds for the items in the offer should be raised by hand before it launches and returned to their previous level afterwards. Stock Aging and Turnover holds the other end of the cycle: which item has been sitting on the shelf for how many days, and which one turns fast. Catch dead stock early and the threshold comes down, releasing capital.

Do not wait for the whole catalogue to start. Pick the ten items that stock out most often, write down the average daily sales of the last 90 days and the real delivery time of the last five purchase orders for each, compute the threshold with the simple safety stock formula, and route the alert to one named owner. A month later, count the stockouts those ten items had and correct the numbers. If you want to try stock health, the risk queue and the purchasing flow in one place, you can open a free Rocketly account and run a small pilot on your own catalogue.

FAQ

Is a reorder point the same thing as a minimum stock level?

No. A minimum stock level is usually a fixed floor that ignores how fast the item moves. A reorder point contains the consumption expected during the lead time plus safety stock, so it is a number that has to change when the item speeds up or the supplier slows down.

How often should I update the threshold?

Rather than one fixed rule, use a rhythm tied to class: monthly for high-value, fast-moving items, quarterly for the middle group, twice a year for the long tail of low-value SKUs. Events like a supplier change, a seasonal turn or a planned campaign call for a review without waiting for the calendar.

How do I compute average daily sales for seasonal items?

An annual average is misleading for a seasonal item. Before the season, base the threshold on the sales velocity of the same period last year; during the season, use a short window such as the last 14 to 21 days. Remembering to lower the threshold at the end of the season is the cheapest way to avoid dead stock.

With several warehouses, should thresholds be per warehouse?

If you can transfer between locations, look at both the total and the per-warehouse figure: stock that is sufficient in total but sitting in the wrong warehouse does not exist as far as a shipment is concerned. Where transfers are quick, warehouse thresholds can stay low and be fed centrally; where transfers are slow, each warehouse has to carry its own.