ABC-XYZ Inventory Analysis: Setting a Different Policy for Different Items

Most organizations manage hundreds or thousands of items, and no planning team can give all of them equal attention. Inventory classification is a way of deciding where limited management attention should go. The two common dimensions are ABC and XYZ: the first measures an item's value importance, the second the predictability of its demand.

In ABC analysis we rank items by share of annual consumption value (price × usage volume) and typically call the items that make up roughly 70 to 80 percent of cumulative value class A, the next band class B, and the rest class C. The exact cut-offs are not hard rules; what matters is that class A is a small set of items that truly carry the most capital or revenue and deserve careful management.

XYZ analysis adds the second dimension: the coefficient of variation of demand (standard deviation divided by mean). Low-variability items (X) have steady, predictable demand, Y items have moderate variability, and Z items are highly variable or intermittent. The thresholds are not universal and should be tuned to your own data distribution; a common starting point is something like 0.5 and 1.0, but check it rather than copying it blindly.

The real value appears when you put the two dimensions into a 3×3 matrix. An AX item (valuable and predictable) can run on automatic replenishment with low safety stock. An AZ item (valuable and highly volatile) is the hardest combination: it ties up a lot of capital and the forecast is unreliable, so instead of a formula think about customer collaboration, make-to-order or shorter lead times. A CX item can be handled with simple batch ordering and no planner time at all.

Each cell of the matrix needs a defined policy: target service level, forecasting method, review frequency and decision owner. Without this step the matrix is just a nice report. A simple example: AX gets high service level and weekly review; AZ gets a balanced service level with manual decisions; CZ gets no safety stock and is bought or made to order; CX and CY run on automatic ordering with minimal oversight.

The first mistake is building the classification from a short period, which presents seasonality or one unusual order as a stable pattern. Use at least 12 months of data and keep new items separate. The second mistake is ignoring strategic importance: a cheap class-C item that can stop production deserves its own policy because it is critical. The third is running the analysis once and never revisiting it; items migrate between cells over time.

A practical starting point is a simple table with four columns: item code, annual consumption value, demand coefficient of variation, and ABC-XYZ class. With that table, within a week or two you can establish what share of inventory capital is locked in slow, volatile items and which class-A items are still managed with an indifferent one-size-fits-all policy. Those two answers usually make the list of first actions obvious.

Ultimately ABC-XYZ is a tool for allocating attention, not a goal in itself. It delivers the most when its output feeds concrete decisions in the S&OP meeting: which items change policy, which are discontinued, and which need a structural solution. As in our DDIT approach, diagnosis comes first and decision second — not tools first and a search for problems afterwards.