How do I use ABC analysis to prioritize inventory?

To run an ABC inventory analysis, choose one consistent measure, calculate it for every item in the defined group, sort the items from largest to smallest, and assign classes using a written boundary rule. Then decide what each class actually changes. A letter without an operating policy is just another spreadsheet column.

For an annual usage analysis, the measure is the cost of the units consumed during the year. That can help you direct attention toward the items responsible for most of that cost flow. It does not tell you which products are most profitable, which stock is missing, or which inexpensive part could stop tomorrow's work.

Choose the question before the ranking

ABC describes a way to group ranked items, not one mandatory definition of importance. Oracle's classification criteria distinguish current on-hand value, individual item cost, historical usage quantity and historical usage value. Changing the measure can change the answer.

A costly machine spare used once may rank high by unit price and low by annual usage value. A cheap consumable used thousands of times can do the reverse. Neither ranking is inherently dishonest. Trouble starts when a report answers one question and the reader thinks it answers another.

Here, the question is: which stocked items account for most of the cost of material used across one complete year? The purpose is to allocate stock-record and usage-review effort. We are not selecting a purchase quantity or valuing the current shelf.

Use a defined location or business unit, a common currency and the same start and end dates. Include the complete item population for that scope, not only the items someone remembers as important. Retain zero-usage items in a separate visible category rather than quietly dropping them from the operating review.

Prepare a clean usage file

Export the item identifier, description, unit of measure, qualifying usage transactions and their cost basis. A case of twelve and an individual unit must not share a quantity column without conversion. A supplier receipt is not automatically consumption, and moving a box between two included storage locations should not count as using it twice.

Decide how supported returns or corrections affect the measure, and keep those adjustments traceable. Missing cost is not zero cost. Negative usage or a duplicate item identifier deserves investigation before a clean-looking rank hides it.

Oracle describes historical usage value as the sum of transaction quantities multiplied by their transaction unit costs over the specified period. If costs vary, calculate the relevant transaction values before adding them. Multiplying a whole year's quantity by today's price produces a different measure, which should be named accordingly.

The original case below deliberately supplies a constant complete unit cost for each item throughout the period. That makes quantity times cost sufficient. There are no returns, transfers, cost changes or omitted usage transactions. Those are simplifying assumptions, not findings about a real warehouse.

Six items, one complete year

A fictional workshop analyzes all six stocked material types used from January 1 through December 31, 2027. The quantities represent actual consumption within this invented case. They are not forecasts, purchases, current balances or sales revenue.

Scroll across to follow annual usage value and its cumulative share. Each row represents one distinct stocked item.

For the motor, 300 × $200 = $60,000. Repeat the multiplication for each row and add the results: $100,000 of annual usage value.

The rows are already sorted in descending value. Each item's share is its value divided by $100,000. Cumulative share adds the current row to all rows above it: the motor and filter together contribute $84,000, or 84%.

Do not sort by the quantity column accidentally. The clip has the largest unit count but the smallest annual usage value. Units of different products are not interchangeable measures of cost or consequence.

Say what happens at the boundary

The workshop chooses this explicit rule: Class A contains the highest-ranked items through the first item that brings cumulative value to at least 80%. Class B contains subsequent items through the first item bringing cumulative value to at least 95%. The remaining positive-usage items are Class C.

Under that rule, the motor and filter are A. The filter crosses the 80% boundary, so it stays in A. Their combined share is 84%.

The belt and seal are B. Cumulative value reaches 94% after the belt and 98% after the seal; the seal crosses the 95% boundary. B therefore contributes 14% of total usage value. The fuse and clip are C, contributing the remaining 2%.

Those percentages are chosen boundaries and calculated results for this exercise. They are not universal inventory rules. Oracle's ABC explanation describes user-defined criteria and classes, including using classifications to guide count frequency. It does not require this workshop's thresholds or schedule.

A different rule that included only rows whose cumulative shares remained below 80% would put the filter elsewhere. That is why the boundary instruction belongs beside the report. Also specify how equal-value items are treated; a stable item-code tie-break is reproducible, while keeping tied values together may better fit a particular policy. Document the choice before comparing runs.

Here, A contains two of six items, or one-third of the item types, and 84% of usage value. Do not rewrite the outcome as “twenty percent of our items produce eighty percent of the value.” The calculation did not say that.

A C item can still stop the work

Now add one supplied operating fact: the fuse is required for the workshop's equipment to operate safely, and there is no approved substitute available during a shortage. The annual-dollar ranking has not measured that consequence.

Calling the fuse C is still correct under the chosen measure. Treating C as permission to ignore it would be the mistake. Keep its economic classification visible and add a separate critical-item flag, with the reason, responsible person and review date.

Do not turn the flag into an invented estimate of lost revenue. This case supplies no interruption length, forgone sales or recovery cost. It establishes a reason for closer availability review without pricing a loss it has not measured.

Other overrides may concern expiry, safety, an unusually long replenishment wait or the absence of an acceptable alternative. Check the actual consequence and evidence. A generic “important” flag on everything would merely recreate the undifferentiated list you started with.

Give the letters a feasible job

Suppose the workshop proposes counting each A item monthly, each B item quarterly and each C item twice a year. This is a fictional control proposal, not a recommended frequency for every business.

With two items in each class, that creates 24 A-item counts, eight B-item counts and four C-item counts: 36 item-count sessions per year. Assume each session has a planned twenty-minute allowance for locating, counting and recording that one item.

The baseline workload is 36 × 20 = 720 minutes, or twelve hours. Counting all six items monthly would instead create 72 sessions and twenty-four planned hours. The difference is twelve planned hours, not proven labor savings or reduced payroll.

The critical fuse changes the proposal. Moving it from twice-yearly counting to monthly counting adds ten sessions, or 200 minutes. The revised plan has 46 sessions and 920 minutes: fifteen hours and twenty minutes.

The workshop has allocated sixteen paid hours to these sessions over the year. The revised plan leaves only forty minutes of that allowance. A twelve-hour budget would no longer cover it. The exception therefore needs an actual scheduling decision, not just a colored cell.

The twenty-minute session allowance excludes extended discrepancy investigations, repairs, purchasing work and training. Those need separate capacity when required. Annual totals also do not prove that the work fits each month's staffing. Put the sessions on actual dates, assign an owner and check competing commitments before calling the policy workable.

Do not ask ABC to do another record's job

More frequent counting is not permission to accept unexplained adjustments. The physical-count reconciliation guide handles the item, location, movement cutoff and condition evidence needed when recorded stock differs from what is present.

Nor does an A classification say to buy a larger order. The reorder guide separates the replenishment trigger from order quantity and usable arrival dates. A high-usage-value item may warrant close review without warranting excess stock.

The $100,000 annual usage flow is not current inventory worth $100,000. The case supplies no closing quantities. It also supplies no selling prices, customer receipts or margins. For the relationship between stock, turnover and spendable money, use the working-capital guide rather than treating this ranking as a cash forecast.

Rerun the analysis with a reason

Retain the dated source file, scope, cost method, sorting rule, thresholds, classification results and approved exceptions. When a later run moves an item from B to A, investigate what changed: usage, cost, the included population or the method itself.

A new item with no history should not inherit a reassuring C label simply because its recorded usage is zero. Give it a provisional review based on supported demand and consequences. A discontinued item with no usage may need disposal or return decisions, not a lighter count schedule indefinitely.

Before adopting your first report, ask four questions: Can another person reproduce the numbers? Do the class boundaries handle crossing items and ties? Are critical low-value items protected? Can the team carry out the resulting policy on real dates?

Start with one well-defined stock group and finish with those decisions recorded. A classification earns its place when it changes attention intelligently, without confusing inexpensive with expendable.