ABC Inventory Analysis: Classify SKUs, Set Policy by Tier
Classify annual usage value, then account for critical items, demand patterns and supplier risk when setting purchasing policies.
Jainul Vaghasia/Published /Updated /7 min read
Use the definition
Turn procurement terms into an operating system.
This reference page should help you understand the concept first. When the term affects purchasing execution, LineNow connects it to live POs, supplier replies, receiving, and accounting handoff.
ABC inventory analysis ranks SKUs by annual usage value: units used or sold in a year multiplied by unit cost. It groups the sorted items into tiers so a team can prioritize review. This is annual consumption value, not the value of stock currently on hand. Common A/B/C percentage splits are conventions; they are not predictions of a catalog's distribution or stockout impact.
The classification formula
For each active SKU:
AUV = units sold (trailing 12 months) × unit cost
Steps:
Compute AUV for every active SKU.
Sort all SKUs descending by AUV.
Compute each SKU's cumulative percentage of the total AUV across the full catalog.
Assign class by cumulative threshold:
Class
Cumulative AUV threshold
Illustrative SKU share
Illustrative value share
A
≤ 80%
~20%
~80%
B
81–95%
~30%
~15%
Read before ordering
A dense operator briefing for teams that need sharper buying, cleaner supplier follow-up, and fewer expensive surprises.
C
> 95%
~50%
~5%
The 80/95 thresholds are a standard starting convention, not a universal rule. If your catalog's natural break falls at 75%/92%, use those. Single-source SKUs, regulated goods, and production-critical components often deserve A-class treatment regardless of AUV rank — a $4 ingredient that stockouts shuts down a kitchen carries the same operational consequence as a $40 one.
Worked example
A specialty retailer with 200 active SKUs and $400,000 in annual inventory spend:
SKU band
AUV total
SKU count
Cumulative AUV %
Class
Top 40 SKUs
$320,000
40 (20%)
80%
A
Next 60 SKUs
$60,000
60 (30%)
95%
B
Bottom 100 SKUs
$20,000
100 (50%)
100%
C
In this constructed example, 40 SKUs account for 80% of annual usage value. That does not imply 80% of lost-sales risk: a low-cost ingredient or spare can stop the sale or production of a much more valuable product. Combine ABC with criticality, substitution options, supply risk and demand behavior.
What ABC class should determine operationally
Policy dimension
A-items
B-items
C-items
Safety-stock policy
Set by shortage cost and supply risk
Set by shortage cost and supply risk
Check critical exceptions before reducing buffers
Review frequency
Weekly or real-time alert
Weekly
Bi-weekly or on-order
Reorder trigger
Automated alert + PO draft
Alert-driven
Review-on-order
Supplier relationship
Named contact, SLA tracked
Shared queue
Catalog or spot order
Physical count cycle
Weekly or perpetual
Monthly
Quarterly
These are sensible defaults. Volatile A-items may need daily review; stable C-items with a single reliable supplier can tolerate even lighter controls. The point is differentiation — applying the same cadence and service target to a $320,000 SKU band and a $20,000 SKU band wastes both money and operator attention. The physical count cadence in the table above maps directly to a cycle counting program: rotating partial counts that verify Inventory Record Accuracy (IRA) without halting operations, scheduled more tightly for the A-tier items where a ghost-stock error is most expensive.
Inventory turnover reflects ABC tier naturally: A-items — high annual usage value, high velocity — typically turn the most frequently; C-items the least. If portfolio turns for A-items are significantly below the category benchmark, the likely cause is MOQ-driven over-ordering or inflated safety stock buffers. Slow turns in the C-tail typically indicate dead stock accumulation — a write-off review question, not a replenishment controls question.
GMROI follows the same tier pattern with the margin dimension added: A-items' combination of velocity and high usage value typically generates the highest gross margin return per dollar of average inventory cost; C-items accumulating as dead stock drag the portfolio denominator without contributing proportional margin. Viewing GMROI by ABC tier makes visible exactly how much the C-tail's idle capital is costing relative to where the catalog's value is concentrated.
Where ABC analysis breaks down: the demand-pattern blind spot
ABC classification ranks by value. An A-item that sells three units every day has fundamentally different replenishment needs than an A-item that sells zero units for six weeks and then thirty in a single large order. Both are A-class by AUV. But applying the same safety-stock formula or the same forecast method to both will systematically over-stock or under-stock one of them.
The missing dimension is demand variability. The coefficient of variation (CV²) measures how volatile non-zero demand is relative to its mean. The Syntetos–Boylan–Croston (SBC) framework combines CV² with a second parameter — ADI (Average Demand Interval, the average number of periods between non-zero demand observations) — to classify each item into four demand regimes:
Pattern
ADI
CV²
Examples
Smooth
≤ 1.32
≤ 0.49
Daily-sold staples; coffee beans, house wine
Intermittent
> 1.32
≤ 0.49
Slow but stable; specialty bitters, niche parts
Erratic
≤ 1.32
> 0.49
Daily but spiky; trending items, weather-driven
Lumpy
> 1.32
> 0.49
Rare and spiky; B2B bulk orders, catering
These four demand regimes require different forecasting methods. Smooth demand: exponential smoothing or a moving average. Intermittent or erratic demand: Syntetos–Boylan Approximation (SBA), which corrects the systematic over-estimation bias in Croston's method. Lumpy demand: SBA plus operator-level review for event-driven signals.
Applying a moving average to a lumpy A-item will over-order on top of spikes and stockout in the gaps — silently, because the math will look internally consistent. ABC class alone cannot prevent this.
The full policy matrix: ABC × demand pattern
The practical replenishment policy for each SKU is the cross-product of its ABC class and its SBC demand pattern:
Most ABC analysis guides stop at the tier thresholds. The matrix is where the replenishment output actually diverges: an A-item with lumpy demand needs a different buffer, a different PO cadence, and a different supplier conversation than an A-item with smooth demand — even though both rank identically in the AUV sort. Collapsing them into a single "A-class policy" builds the error in at the policy level.
Keeping ABC classes current
ABC classification is not a one-time spreadsheet exercise. Demand drifts, catalogs expand, and seasonal items move between tiers. Three practical rules:
Recompute quarterly using trailing-12-month data. A seasonal item that ranks A in Q4 may fall to C by Q2. Static classes drift out of alignment with actual purchasing behavior within two quarters. Trailing 12 months smooths seasonality while still tracking the catalog's real value distribution.
Lock critical items at A by override. Single-source ingredients, regulated goods, and high-COGS production inputs should hold A classification regardless of recent AUV. A supply disruption on a $4 component that blocks $80,000 of weekly production is an A-item risk even if the AUV says otherwise. For perishable A-items, A-class lock should accompany the tightest physical count cadence and mandatory FEFO picking — an expired lot in the A-tier carries the combined cost of write-off and stockout.
Shadow-class new SKUs at A for the first 90 days. A new SKU does not have 12 months of AUV history. Treating it as C-class by default underweights the risk during the period when demand patterns are least understood. After 90 days, enough history exists to compute a defensible AUV and move the SKU to its natural class.
Apply this to a real purchasing record
LineNow's purchasing workflow connects purchase orders, supplier replies, receiving and accounting handoff. In a demonstration, inspect which usage-value inputs and class thresholds are used, and whether a changed class updates a policy or merely informs the buyer.
Use the result to agree the fields, decision owner and exception process. A linked purchasing record supplies evidence for this analysis; it does not by itself prove a particular dashboard, financial outcome or automatic approval policy.
