How suggested order quantity works

Suggested order quantity is one of the highest-return things a distribution business can automate, and one of the easiest to implement in a way nobody trusts. This guide covers the calculation, the inputs that actually determine whether it works, and a rollout sequence that keeps buyers on side.

For the definition and a shorter treatment, see what is suggested order quantity.

1. The calculation

Nearly every implementation is a variation on target position minus current position:

suggested quantity = (demand rate × (target cover + lead time)) − stock on hand − incoming − open commitments, rounded to the lot size and clipped to the minimum and maximum order quantity.

Worked through: an item sells 12 units a week. Target cover is two weeks and lead time is one, so the target position is 12 × 3 = 36. There are 9 on hand and 6 on an undelivered order, so the current position is 15. The suggestion is 21, rounded up to a case of 24.

The term that gets dropped most often is incoming. Omit it and the same example suggests 27 instead of 24 — a 12% over-order on one line, repeated across the catalogue every cycle. Most disappointing SOQ implementations are disappointing for exactly this reason.

2. The inputs decide everything

The arithmetic is trivial. The inputs are the project.

Demand rate. A trailing average over a window long enough to smooth noise and short enough to track trend. Four to twelve weeks suits most fast-moving categories.
Target cover. A policy decision, not a calculation — it prices availability against working capital, and it should differ by ABC and FSN class rather than being one number.
Lead time. From raising the order to stock being sellable, including receiving and putaway. Measure it; do not ask the supplier what it is.
Rounding rules. Minimum order quantity, case or lot size, maximum.

If you have to choose where to spend effort, spend it on lead time. It appears inside the multiplication, so an error there scales with demand — a lead time wrong by a week on a fast mover is wrong by a week of sales, on every line, every cycle.

3. Decide what the suggestion is allowed to do

Three postures, in increasing order of ambition. Most businesses should start at the first and stay there longer than they expect to.

Advisory. The number is shown; a person orders. Builds the trust everything else depends on.
Default. The suggestion pre-fills the order and can be overridden. Most of the benefit, most of the time.
Automatic. Orders are raised without review, usually restricted to C-class items where the cost of being wrong is small.

Skipping to automatic before the inputs are trustworthy is the classic failure. One bad automated order run costs more credibility than a year of good suggestions earns.

4. Roll out in this order

Fix the pipeline data first. If in-transit and pending-order quantities are not reliable, stop — the netting cannot work and every suggestion will be over-stated.
Set parameters for the top decile by value. A-class items are a small list and carry most of the money. Get lead time and consumption right there before touching the tail.
Run advisory and compare. For a few cycles, show the suggestion next to what the buyer ordered. The gap is your data: where buyers consistently override, the parameters are usually wrong, not the buyers.
Promote to default for classes that agree. Where suggestions and decisions converge, pre-fill. Where they do not, keep it advisory and investigate.
Review parameters on a schedule. Lead times and consumption drift. A quarterly refresh is the difference between a system that stays useful and one that quietly stops being right.

5. The failure modes, and what to do about them

Promotions. Trailing averages are wrong exactly when volume matters most. Suppress or override suggestions for promoted lines rather than letting the average absorb the spike and mis-order for the following two months.
New items. No history means no calculation. Seed an assumption from a comparable item and mark it for early review.
Seasonality. A trailing window lags a seasonal turn by design. Either widen it with year-on-year context or accept manual intervention at known transitions.
Stale parameters. The quiet one. Arithmetic over a lead time nobody has revisited looks authoritative and is wrong. Show the inputs on the suggestion so a buyer can spot it.
Capital and constraint blindness. Line-by-line suggestions do not know about a total budget or a vehicle's capacity. If a constraint binds, it has to be applied on top.

6. What to measure

Judge the system on four numbers, tracked before and after: stock-out frequency on A-class items, total inventory value, inventory turns, and — most diagnostic of all — the override rate. A very high override rate means buyers do not trust it. A zero override rate usually means nobody is reading it.

How this works in xMatix

xMatix Procurement generates suggestion lines through three engines — minimum stock level, replenish-to-maximum, and consumption-based forecasting — plus template orders, and can run them on a schedule so a dated run exists to review. The netting includes stock on hand, in-transit shipments, pending purchase orders, open requisitions and unfulfilled sales orders, and runs can include or exclude items by ABC class and by fast, slow or non-moving status.

Every suggested line carries its own inputs — projected quantity, available quantity, each pipeline component, lead time and average consumption — written onto the line, which is what makes step 3 above practical: the override conversation can be had against visible assumptions rather than a black box. Separately, when a purchase order line would order stock the pipeline already covers, Sense Assist raises a nudge showing the netting and offering the corrected quantity.

Where demand is worth predicting rather than assuming, statistical and machine-learning forecasting model it from historical sales, inventory movement patterns, seasonal variation and other influencing parameters — a different instrument from the rules-based engines, and the right one for items whose demand actually has a shape.

One thing to plan for on the rules-based side: those engines read average consumption, lead time and ABC/FSN classification as item attributes, so step 5 — the scheduled parameter review — is a process you need to own for the items you replenish that way.