Skip to main content

13. What Do You Review? ABC XYZ Segmentation and Other Methods

Teaching rewrite — Part 3, the Review process step: where should limited planner time go? Original wording; illustrative numbers.

Introduction​

A model can forecast every item; planners can’t review every item. Each cycle you must choose a small set — the products that are important to the supply chain and where a human is likely to improve the baseline. Value-weighted KPIs (lesson 11) say which items matter; FVA (lesson 12) says whether reviews help. Neither says which items to open first. That’s this lesson.

This lesson covers:

  • ABC analysis and ABC XYZ matrices
  • Setting them up correctly: forward-looking value and historical forecast error
  • Their limits
  • A multi-criteria star rating that goes beyond two dimensions

13.1 ABC analysis​

ABC (Pareto, 80/20) splits items into three classes along one dimension:

ApproachABC
By share of itemsTop 5% of items (~20–60% of volume)Next 15% (~20–50%)Bottom 80% (~5–20%)
By share of volumeItems making the first 60% of volume (often 5–15% of items)Next 20% (15–30% of items)Last 20% (50–80% of items)
PriorityHigh — the critical fewMediumLow — the trivial many

ABC XYZ​

Add a second dimension (XYZ) → a 3 × 3 matrix with nine classes, e.g. volume (ABC) × profit (XYZ), from high-volume/high-profit (AZ) to low-volume/low-profit (CX).

13.2 Setting ABC XYZ up for forecast review​

Our goal: point planners to items that are important and hard for the model. So:

AxisMeasuresUseAvoid
ABC — importanceForward-looking, value-weightedNext X months’ forecast × unit valueHistorical sales (backward-looking; constrained by shortages; misses seasonality, promotions, trends)
XYZ — forecastabilityHow badly the engine did on its ownHistorical forecast error, e.g. Score% = MAE% + |Bias%|COV of demand (inflated by trend and seasonality; leaks hindsight — lesson 10)
Forecastability ↓ / Importance →A (high value)BC (low value)
Z (high error)AZ — review firstBZCZ — review if time
YAYBYCY
X (low error)AX — spot-checkBXCX — leave to the engine

Review A-Z first (important, poorly forecast); leave C-X (unimportant, well forecast) to the engine.

Worked example: classify five items​

Forecast value next 6 months and historical Score%:

ItemForecast value (6 mo)Score%ABCXYZClass
P1€400k55%AZAZ — review first
P2€350k12%AXAX — spot-check
P3€60k48%BZBZ
P4€8k70%CZCZ
P5€5k10%CXCX — leave to engine

(Thresholds here: A ≥ €200k, C < €20k; X < 20%, Z ≥ 40%.)

Checkpoint: ABC XYZ​

Quiz

Checkpoint — sections 13.1 and 13.2

1 / 4

What should the ABC (importance) axis use for forecast review?

Pick an answer to see if you are right.

13.3 Limits of ABC XYZ​

ABC XYZ was state of the art half a century ago, when planners had no computers to analyse thousands of items. Its limits:

  • Only two dimensions. Lead time, lead-time reliability, shelf life, obsolescence risk, business criticality, and promotions don’t fit.
  • Arbitrary thresholds. Is the best B really different from the worst A? Two cut-offs per axis is coarse.

Use it for simple analyses — but don’t trust it blindly.

13.4 Beyond ABC XYZ: multi-criteria review​

Today you can score every item on many criteria automatically. Review items where:

  • You have information the model lacks — promotions, marketing, customer actions
  • It is a new product introduction
  • The forecast behaves strangely — persistent over/under-forecasting, sudden shifts
  • Errors are expensive — high cost, long lead time, special storage (frozen), short shelf life (food, medicine)
  • End of season is near — risk of dead stock
  • It is strategically critical — service-level agreements, spare parts needed to sell equipment
  • It is sold together with other items — nuts and bolts should move together

A star rating​

CriterionStars
Good historical forecastability−1
Small volume/value−1
End of season+1
Promotion planned+2
Brand-new product+3
Critical product+3

Start every item at a base (say 2 stars), apply the rules, review from the highest.

Worked example: star scores​

ItemSituationCalculationStars
Q1New product with a launch promotion2 + 3 + 27
Q2Critical spare part, forecast well2 + 3 − 14
Q3Seasonal item, end of season, small value2 + 1 − 12
Q4Small, stable item2 − 1 − 10

Review order: Q1, Q2, Q3, Q4 (likely skipped).

Critical ≠ inflate

A critical product deserves attention, not a higher forecast “to be safe”. Protection against shortages is the inventory manager’s job (service levels, safety stock). Inflated forecasts erode trust and lead to double-guessing — “planners always inflate, so I’ll cut my stock target”.

Checkpoint: beyond ABC XYZ​

Quiz

Checkpoint — sections 13.3 and 13.4

1 / 4

Which is a limitation of ABC XYZ?

Pick an answer to see if you are right.

Rank: building a review list​

Quiz

Review list

1 / 2

Rank the steps to build a monthly review list.

OrderDrag the handle to rearrange — top should be first.

  1. 1Add flags: promotions, launches, end-of-season, criticality, odd behaviour.
  2. 2Score items (e.g. stars) and sort.
  3. 3Compute forward-looking value (forecast × unit value) per item.
  4. 4Review from the top until the time budget runs out; track FVA.
  5. 5Compute historical Score% (MAE% + |Bias%|) per item.

Drag items into the correct order (or use ↑↓), then check.

Check: what to review​

Quiz

What do you review?

1 / 4

Why not base the ABC axis on historical sales?

Pick an answer to see if you are right.

Takeaways​

Remember
  • Review items that are important and where you are likely to add value
  • ABC = forward-looking value-weighted forecast; XYZ = historical forecast error — not historical sales, not COV
  • ABC XYZ is simple but limited: two dimensions, arbitrary thresholds
  • Go multi-criteria: promotions, launches, odd behaviour, cost, shelf life, end of season, criticality, related items
  • Critical means review carefully — not inflate the forecast

Further Reading​