Publicité
ERP IMPLEMENTATION
🇫🇷 Lire en français →

ABC/XYZ Classification and Safety Stock in ERP: Set a Stock Policy That Holds

ABC/XYZ matrix, safety stock formula, rules by segment and ERP setup: cut inventory without breaking your service level, with a quarterly review.

ABC/XYZ Classification and Safety Stock in ERP: Set a Stock Policy That Holds

In most companies, safety stock was set once, when the item record was created, by a cautious storekeeper or a buyer burned by a stockout. Three years later, the ERP faithfully applies that number to items whose demand has doubled or vanished, with supplier lead times that have changed. The supply chain director sees stockouts on the items that matter, the CFO sees inventory inflating working capital, and both are right.

The problem is not the average stock level, it is the lack of a differentiated policy. Treating 3,000 SKUs the same way over-protects marginal items and under-protects critical ones. This guide is for supply chain directors, logistics managers and CFOs: it shows how to segment with the ABC/XYZ method, calculate a defensible safety stock, configure it in the ERP and review it before it drifts.

The real cost of poorly managed inventory

Badly calibrated stock costs you in three ways, and the accounts only show part of it.

Tied-up cash. Every unit on the shelf was paid for before it was sold. Inventory weighs directly on working capital, and financing it has a real cost, even when no line of the P&L displays it. Reducing stock without hurting service is one of the most accessible working capital levers, because it depends on neither customers nor banks.

Obsolescence and write-downs. Overstocked inventory ages. Items that sit idle end up in a provision at closing, as detailed in our guide on inventory write-downs and provision setup. The link is direct: an oversized safety stock today is a write-down tomorrow.

Stockouts. Conversely, insufficient stock on strategic items costs lost sales, line stoppages, expedited shipments and unhappy customers. This cost is more diffuse, and therefore easier to underestimate.

The goal is not to minimise stock but to arbitrate between these three costs, item by item. The target service level then becomes a management decision, not a reflex.

ABC and XYZ: building the 9-box matrix

ABC: ranking by value

ABC classification (the Pareto principle applied to inventory) ranks items by decreasing contribution to a value indicator, usually annual consumption value or revenue. The usual approach is three classes: A for the items that concentrate most of the value, B for the middle, C for the long tail. The exact thresholds (70/20/10, 80/15/5…) are an internal convention, not a standard: choose them based on your portfolio and put them in writing.

XYZ: ranking by demand regularity

The XYZ axis measures predictability. You calculate the coefficient of variation of demand (standard deviation divided by the mean) over a consistent period, often monthly over twelve to twenty-four months. X is stable demand, Y is variable demand with a trend or seasonality, Z is erratic or sporadic demand. Again, the thresholds are conventions to document.

Worked example (fictional)

Take a fictional industrial distributor with 1,000 active SKUs. The figures below illustrate the method; they come from no study.

X (stable)Y (variable)Z (erratic)Total
A (high value)907040200
B (medium value)90110100300
C (low value)60140300500
Total2403204401,000

Three readings emerge. The AX box (90 SKUs) is the easiest to manage and the most rewarding to fine-tune. The AZ box (40 SKUs) concentrates the risk: high value, unpredictable demand. The CZ box (300 SKUs) is a third of the catalogue but very little value: that is where complexity hides.

One policy per segment

The matrix is only useful if each box triggers a different rule. Here is a starting point, to adjust for your industry. The service levels are suggested steering targets for illustration, not market benchmarks.

SegmentReplenishment ruleTarget service levelSafety stock
AXAutomatic MRP calculation, frequent small lotsHighLow, finely calculated
AYMRP with smoothed forecast, monthly reviewHighModerate, factoring in seasonality
AZManual analysis, framework agreement or shared stock with the supplierHigh but negotiatedCalculated cautiously, capped by value
BX / BYReorder point or MRP, quarterly reviewMedium to highCalculated, standard
BZReorder point with economic lot sizeMediumReduced, on rolling forecast
CXReorder point, large lots, few reviewsMediumLow, simple coverage rule
CYSimple min/maxMediumFixed days of coverage
CZMake or buy to order, or minimal stock; consider delistingLow, by choiceOften zero

Two principles guide this table. First, calculation effort follows value: A items deserve a fine calculation and frequent review, C items a simple rule. Second, the Z axis requires a change of logic: on erratic demand, a statistical safety stock becomes huge and expensive. You are better off shortening lead time, negotiating with the supplier or producing to order.

For items with predictable consumption in a production chain, the replenishment policy relies on requirements planning: see our guide to MRP II and production planning (currently in French).

Calculating safety stock

The formula, explained simply

Safety stock absorbs two uncertainties: demand that varies during the replenishment lead time, and the lead time itself. In its most common version, when only demand varies:

Safety stock = z × σ × √L

  • z: the coefficient for the target service level (normal distribution). For example, about 1.65 for 95% and about 2.33 for 99%. Going from 95% to 99% raises safety stock by about 40%, which is why aiming for 99% everywhere is an expensive luxury.
  • σ: the standard deviation of demand per period (day, week or month).
  • L: the replenishment lead time, expressed in the same unit as σ.

When supplier lead time also varies, you add a term: the square root of (L × σ² + D² × σL²), where D is average demand per period and σL is the standard deviation of lead time. Many companies discover at this point that lead time variability weighs more than demand variability, especially with distant suppliers.

The reorder point then follows: average demand during lead time + safety stock. When available stock (including open orders) falls below this threshold, the ERP triggers replenishment.

Fictional example: an AX item consumed at an average of 100 units a week, weekly standard deviation of 20, lead time of 4 weeks, 95% service target. Safety stock = 1.65 × 20 × √4 = 66 units. Reorder point = 100 × 4 + 66 = 466 units.

Static or dynamic

A static safety stock is a value entered on the item record, valid until further notice. It is simple but ages badly. A dynamic safety stock is recalculated periodically from consumption history and observed lead times. It tracks reality but requires clean data and control: an automatic recalculation on a history polluted by a stockout or an exceptional order produces a bad number, faster.

The limits of the formula

  • It assumes roughly normal demand, independent from one period to the next, which is false for Z items, intermittent items or promotional items.
  • It uses sales history, not true demand: past stockouts understate demand.
  • It ignores minimum lot sizes, packaging multiples and storage capacity constraints.
  • It treats each item in isolation, ignoring substitutions or correlations between SKUs.

The formula gives a defensible starting point, not a truth. The buyer’s judgement remains useful, especially on AZ items.

Configuration in the ERP

ERPs cover all these needs, but with different labels and levels of detail. Screen names change by version and edition: check them in your vendor’s documentation and on a test environment before building a rollout plan. Here are the functions to look for, with no promise on exact labels.

  • SAP (MM and PP): the material master’s MRP data groups the planning type (reorder point or deterministic MRP planning), safety stock, the lot-sizing procedure and planned delivery time. Safety stock there is a manually entered field or one fed by a forecasting function.
  • Microsoft Dynamics 365 Supply Chain Management: item coverage groups drive the planning logic (period, min/max reorder point, safety margin). The vendor documentation details the available coverage modes.
  • Odoo: reordering rules (minimum quantity, maximum quantity, order multiple) cover the min/max model. The statistical safety stock calculation remains to be done upstream, in a spreadsheet or an add-on module.
  • Sage X3: the item record carries safety quantities and requirements calculation parameters, and planning relies on the lots and lead times entered per site.

Whatever the solution, four parameters matter more than everything else.

  1. The category or management class: store the ABC/XYZ class in a field on the item record (or a classification characteristic) so rules can apply by segment and reports can be filtered.
  2. Supplier lead time: a theoretical lead time that is never updated skews everything else. Regularly reconcile it with the lead time actually observed between order and receipt.
  3. Lot size: economic quantity, packaging multiple, days of coverage. A bad lot size can ruin a good safety stock.
  4. The review calendar: how often you recalculate the classification, the history and the safety stock.

For demand planning and logistics needs, see also our supply chain guide to WMS, TMS and demand planning.

Governance: a quarterly review

A stock policy is not a project, it is a cycle. A quarterly rhythm is a good starting point for most companies, with a monthly review of A items only.

What the review examines:

  • the ABC/XYZ reclassification over the last twelve months,
  • items that changed segment, with their new rule,
  • gaps between theoretical and observed lead time,
  • stockouts and their causes (wrong parameter, supplier delay, unforeseen demand),
  • proposals to delist or run down stock.

The indicators to track:

  • Inventory turnover: annual consumption divided by average stock, by segment.
  • Days of coverage: stock divided by average daily consumption.
  • Stockout rate (or service level): share of order lines not served on the promised date.
  • Value of dormant stock: value of items with no movement for a defined period. This is the indicator that links logistics to the CFO, and feeds the closing provision.

A single owner must answer for the result, with a supply chain and finance duo. The calendar must also account for year-end operations: see the annual inventory and cut-off checklist.

Mistakes that keep coming back

ABC calculated once, never revised. Last year’s classification reflects the old catalogue. A recently launched item is classed C for lack of history, and a declining item stays classed A. Automate the recalculation and handle new items with a specific rule.

Supplier lead times not updated. The lead time entered when the record was created often stays unchanged for years. Yet a lead time error is paid for twice: a miscalibrated safety stock and a wrong requirement date. Tension in supply should show up in the parameters: see our article on supply chain resilience and dual sourcing.

Ignored seasonality. A standard deviation calculated over twelve months blends high and low season. The result: too much stock in the trough, not enough at the peak. Use a seasonal safety stock or a forecast that incorporates seasonality before the calculation.

Sales history mistaken for demand. A past stockout flattens the history. Correct stockout periods and exceptional orders before calculating σ.

The same service level for everyone. Aiming for the same rate everywhere makes stock value explode for a marginal benefit on C items. Service level is a parameter to arbitrate by segment.

Going further

Validate a stock policy on a limited scope before rolling it out: pick a family of 50 to 100 SKUs, apply the segmentation and rules for a quarter, then compare turnover, coverage and stockout rate with the previous quarter. To structure your approach and your trade-offs between stock and service, download our supply chain kit and lead magnet and subscribe to the newsletter to receive upcoming guides.