The inventory accuracy formula (and what an acceptable rate actually is)

92.2% — that was the line-level inventory accuracy from a quarterly count at a small enamel-pins store: 14 wrong SKUs out of 180 counted. The same count scored 98.0% by units: just 85 wrong out of 4,300. Both numbers describe one afternoon of counting. Which one you put on the dashboard decides whether 92% feels like a crisis or a rounding error — and the two versions of the formula disagree on purpose, because they answer different questions.

The inventory accuracy formula is a ratio: how much of what you counted matched the books. Define "match" (exact, or within a tolerance) and choose your unit of measure (SKU line, or physical unit), and the rate follows. Here are both versions, worked on the same count data, and how to read the gap between them.

The two versions of the formula

Line-level (SKU-level) accuracy asks: of the SKUs we counted, how many came back exactly right?

Line accuracy = SKUs counted correct ÷ SKUs counted
              = 166 ÷ 180
              = 92.2%

Unit-level accuracy asks: of the physical units on the shelf, how many matched the books?

Unit accuracy = (units counted − variance units) ÷ units counted
              = (4,300 − 85) ÷ 4,300
              = 98.0%

A tolerance variant of either is common: a SKU "matches" if it's within ±2% rather than exact. Tolerance makes the rate look better and the number less useful — more on that below.

The difference is not academic. Unit-level accuracy dilutes: sixty missing units on one bestseller barely dent a 4,300-unit denominator. Line-level counts a SKU as wrong if it's wrong at all, no matter how slightly. So line-level answers the question a small store actually asks every day — can I trust the per-SKU number I'm about to reorder against? — while unit-level answers a different one: how wrong is my inventory value in aggregate? A store can run 98% by units and still be unable to promise stock on a fifth of its catalog.

The same count, worked end to end

The pins store, Pinfleet (illustrative numbers), counted all 180 SKUs in one Saturday session. A few of the misses:

Across the full count: 14 SKUs mismatched, total variance −85 units, 9 of the 14 misses sitting on the store's A list — the roughly 60 fastest sellers it grades with Shopify's native ABC analysis in the admin (verified October 2026).

The arithmetic, all three ways:

• Line accuracy: 180 − 14 = 166 correct → 166 ÷ 180 = 92.2%
• Unit accuracy: 4,300 − 85 = 4,215 correct → 4,215 ÷ 4,300 = 98.0%
• A-item line accuracy: 60 A items, 9 wrong → 51 ÷ 60 = 85.0%

Now add the only layer that ranks severity: value. At an average unit cost of $1.90 across the misses, the −85 units is about $161.50 of recorded inventory value that doesn't physically exist (estimated, illustrative). For enamel pins, 92.2% line accuracy is annoying. Run the identical percentages on a store whose A items are $40 hiking jackets, and the same rate is a five-figure annual problem. The formula measures trust; dollars measure damage.

What an acceptable rate actually is

There is no universal passing grade, and the benchmarks quoted around the industry don't agree with each other — they're computed on different units, different tolerances, and different count methods, so copying one is meaningless. Three signals are worth more than any copied number:

1. A items should count clean almost every time. Your bestsellers are where errors convert directly into unsellable hours.
2. The rate should trend up. A flat 92% for three quarters means your causes are structural, not random.
3. Misses should shrink in dollars, not just in units.

With that framing, here's a response table by band (line-level, exact match):

The band matters less than the direction of travel. And the direction of the variance matters too: overstated records (system says 40, shelf holds 34) sell units you don't have and manufacture oversells, while understated records hide sellable stock and trigger reorders you didn't need. The overstated side has its own name and failure modes — the phantom inventory guide covers it.

Why bother at all? Because records that overstate stock put your bestsellers on the empty-shelf path, and shoppers punish that — a share buys the item at another store, a share substitutes, and a share buys nothing while the phantom units sit in your records looking sellable. Phantom-accurate records are how those empty shelves get scheduled.

Measuring yours without counting everything

You don't need a full stock take to have an accuracy number — you need a count program that covers the right SKUs. Each cycle count produces counted-vs-expected per variant; in Shopify's native count tool, a manager reviews the differences and approving the count applies them, with every resulting change logged in the adjustment history (Shopify Help Center, "Stock counts," accessed November 2026). Roll the slices into a monthly line-level rate and you have a trend within a quarter.

The counting process itself — scopes, cadences, close-of-day rules — is its own discipline, and the Shopify inventory counts guide owns it. The frequency decision (how often each grade should be counted) is covered separately in the cycle count frequency guide. Two boundaries to keep this article honest: when a single SKU's mismatch needs an investigation — receipts, transfers, staff shifts — that workflow lives in the inventory discrepancy guide, not here. And when counted stock needs to be reconciled against book stock period by period, that's the stock reconciliation process, which takes the numbers this formula produces as its input.

One caution on tolerance: setting "within ±2% counts as correct" is legitimate for cheap C items where chasing single units costs more than the units. Never let a tolerance anywhere near your A items — a 2% tolerance on a 500-unit bestseller is ten phantom units you'll reorder against every week.

Common questions

What is a good inventory accuracy rate? There's no industry-standard passing grade — quoted benchmarks use different units, tolerances, and methods, so a copied number means nothing. The useful test: your A items count clean nearly every time, your overall line-level rate trends upward quarter over quarter, and the dollar value of misses shrinks. Judge your rate by those three, not by someone else's.

How do you calculate inventory accuracy in Excel? Put expected quantities in column D and counted quantities in column E, then add a match column: =IF(D2=E2,1,0). Line-level accuracy is SUM(F:F)/COUNTA(F:F). For unit-level, compute total expected and total counted, then use (SUM(D:D)-ABS(SUM(D:D)-SUM(E:E)))/SUM(D:D). Both take about five minutes once your count sheet exists.

What's the difference between inventory accuracy and inventory discrepancy? A discrepancy is one SKU's mismatch plus its cause — the thing you investigate individually. Accuracy is the store-wide rate across an entire count — the thing you measure and trend. You fix discrepancies one at a time; you use the accuracy rate to decide how much to trust your records between counts.

Should I count blind or with expected quantities visible? Blind counting — hiding the expected quantity from the counter — reduces anchoring, where people unconsciously confirm the number on the sheet. A practical setup: counters get a blind sheet, the reconciler compares against expected afterward. Shopify's native count review shows counted versus expected at the review-and-approve stage, which preserves the separation.

How often should I measure inventory accuracy? Every count produces a rate; review the rolled-up number monthly. What matters more is coverage: an accuracy figure computed on a slice you never count is unknown, not good. Make sure your count cadence touches A items monthly, B items quarterly, and C items at least twice a year, then read accuracy on the covered set.

Next step

Run both formulas on your last count sheet tonight: line-level and unit-level, plus the dollar value of the misses at cost. Ten minutes of arithmetic buys you the one number that tells you how much to trust every other number. If the result lands below 97% line-level, the fix isn't more measuring — it's the counting cadence and process rebuild in Shopify inventory counts.