AI Prompts for ABC and ABC-XYZ Inventory Analysis

ABC analysis is a Pareto sort: a small share of items accounts for most of the value, and those items deserve most of the attention. On its own it says nothing about how predictable demand is, which is why pairing it with an XYZ classification on variability produces something more useful — a nine-cell matrix where each cell implies a different forecasting, replenishment and counting approach.

These prompts run the classification, build the matrix with policy implications, and turn it into a cycle counting program. The classification is only as good as the value basis you choose, and the model should state which basis it used and what changes if you switch.

Before you use these

Have these ready to replace the highlighted [variables]:

• Item list with annual usage quantity and unit cost (or annual usage value)
• Demand history by period for variability
• Current counting practice and accuracy results
• Class boundaries you want to use (or ask for defaults)

The prompts

• 1. Run the ABC classification
• 2. Build the ABC-XYZ policy matrix
• 3. Design the cycle counting program

1. Run the ABC classification

Act as an inventory analyst performing an ABC classification.

Data: [item, annual usage quantity, unit cost, and optionally on-hand value, margin, criticality]
Value basis to use: [annual usage value / on-hand value / margin contribution]
Class boundaries: [e.g. A = top 80% of value, B = next 15%, C = last 5%; or by item count]

1. Rank items by the chosen value basis, compute cumulative value share, and assign classes at the stated boundaries.
2. Summary: number and share of items per class, value per class, and the Pareto shape (what share of items makes up 80% of value).
3. Boundary items: those within 2% of a class boundary, where a small change moves them. Recommend whether to hold them at the higher class.
4. Overrides: items that should be reclassified regardless of value — critical spares, regulated items, new products, end-of-life — with the rule for each.
5. If a second value basis was provided, show the items whose class changes between bases and what that implies (e.g. high usage value but low margin).

Present the class summary and the top items in each class. State every threshold used. Do not apply the classification to items with less than a full year of history without flagging them.

2. Build the ABC-XYZ policy matrix

You are building an ABC-XYZ matrix for [site / category].

ABC classes: [item, class]
Demand series: [item, period, demand]
XYZ thresholds: [e.g. X: coefficient of variation < 0.5; Y: 0.5–1.0; Z: > 1.0 or intermittent]

1. Compute the coefficient of variation per item and assign X/Y/Z. Flag intermittent items (many zero periods) as Z regardless of CV.
2. Build the 3×3 matrix: item count, share of items, value share per cell.
3. For each cell state the policy implications:
   - Forecasting: statistical (X), statistical with review (Y), judgmental or none (Z)
   - Replenishment: automated reorder point (AX/BX), periodic review, make/buy-to-order, or kanban
   - Safety stock: formula-based, elevated, or replaced by planned buffer / order-on-demand
   - Planner attention and review frequency
4. Identify the cells that hold the most value with the least predictability (AZ, BZ) — these need the most planner time — and the cells that can be fully automated.
5. Recommend three policy changes from the current state with the value affected by each.

Present as the matrix plus a policy table. Note the limitation: CV is sensitive to period granularity — state the granularity used.

3. Design the cycle counting program

Act as a warehouse controls manager designing a cycle counting program.

ABC classes: [item counts per class]
Current practice and results: [annual count / cycle count, accuracy % by class if known, common discrepancy causes]
Capacity: [counts per day available, staff]
Requirements: [audit or financial requirements, tolerance rules]

1. Set counting frequency per class (typical: A monthly or quarterly, B semi-annually, C annually — adjust to our accuracy history and value) and compute the counts per day required. Compare with capacity and adjust.
2. Define accuracy targets and tolerances per class (e.g. A: 99% within 0 tolerance; C: 95% within ±5%). State how a 'hit' is defined.
3. Daily count plan: how items are selected (class cycle, random within class, triggered counts on zero balance or discrepancy), how the schedule handles receiving/picking cutoffs, and blind vs non-blind counts.
4. Discrepancy process: recount rule, adjustment authority thresholds, root-cause categories (receiving error, picking error, unit of measure, location, system timing), and how causes feed process fixes.
5. Reporting: weekly accuracy by class, trend, top discrepancy causes, adjustments value.
6. The first 90 days: how to establish a baseline and which class to start with.

Present as a program document. If the required counts exceed capacity, show the frequency compromise and its risk.

Related prompts

• Inventory Optimization
• Safety Stock Calculation
• Demand Forecasting
• Warehouse Operations Improvement
• Forecast Accuracy Review
• Working Capital Analysis

Logical next step

After this, most operations teams move on to Inventory Optimization.

All Operations & Supply Chain prompts · Search the full library