Inventory accuracy guide

Inventory Record Accuracy: How to Calculate It (and Why Your Number Is Probably Wrong)

How to calculate inventory record accuracy from cycle counts, the mistakes that inflate or distort the number, and how to turn count variances into fixes instead of a report.

If your warehouse reports 99% inventory accuracy, I’d bet money the real number is lower. I’ve seen the same few mistakes inflate it in building after building, and almost none of them are about bad counting. They’re about how the percentage gets calculated.

Inventory accuracy is one of the easiest KPIs to get wrong, because there’s no single universal formula everyone uses. Change the rules and the number changes with them.

In one operation I worked in, the system showed material in the correct location because the transaction had already been completed, but the physical pallet had never actually moved. On paper, the record looked right until someone counted the rack. Once we started separating transaction errors from physical-movement errors, the accuracy number became more useful because we could finally see what process was actually failing.

What inventory record accuracy actually measures

Inventory record accuracy (IRA) answers one question: when the system says you have it, do you?

IRA = lines counted accurately ÷ total lines counted

A “line” is one SKU at one location (or one SKU, depending on how you count). A line is accurate when the physical count is within your tolerance of the system quantity (the book quantity):

|count − book| ≤ book × tolerance %

Say you counted 200 lines this week and 176 were within tolerance. That’s 176 ÷ 200 = 88%. It’s a hit-or-miss measure: each line either passed or it didn’t.

Simple enough. Here’s where it goes wrong.

Six mistakes that make your number wrong

1. You never defined “accurate”

A tolerance of 0% means the count has to match exactly. A tolerance of 5% means 95 counted against 100 on the books passes. Run the same counts under those two rules and you’ll get two very different percentages.

Neither is automatically wrong. What’s wrong is not knowing which one you’re using, or having each department use a different one. Pick a tolerance, write it down, and don’t change it quietly. If you change it, say so on the report.

2. You’re counting units when you should be counting lines

If a bin holds 1,000 units and the count says 998, you can say “99.8% accurate.” But as an accuracy line, that’s a miss under a 0% tolerance. Record accuracy counts lines, not units.

Both views are useful. Record accuracy (hit rate) tells you how often the system is right. Quantity accuracy tells you how far off the misses are, calculated as 1 − |count − book| ÷ book for each line, then averaged. A warehouse with lots of tiny misses and one with a few giant misses can show the same record accuracy and need totally different fixes.

3. You’re measuring after the cleanup

If someone recounts a bad line, fixes it, and then you report the result, you’re reporting your recount process, not your inventory. Measure the first count, before recounts and adjustments. Track recounts separately.

4. You’re blending A, B, and C items

A 98% overall number can hide A-class items at 90%. A-class is where the money and the customer pain live. Report accuracy by ABC class, not just one overall number. Count A items more often than C items, and set your own frequencies. Mine won’t match yours.

5. You’re averaging weekly percentages

Same trap as any other KPI. Week 1: 10 of 10 lines accurate (100%). Week 2: 50 of 100 accurate (50%). The average of those percentages is 75%. The real number is 60 of 110, which is 54.5%.

Add the accurate lines, add the counted lines, then divide. Never average the weekly rates.

6. You’re only looking at one dollar number

There are two ways to total variance dollars, and they tell different stories:

  • Net $ variance adds overs and shorts together. It’s the P&L view. A $5,000 overage and a $5,000 shortage net to $0.
  • Absolute $ variance ignores direction. Those same two misses are $10,000 of absolute variance.

A net number near zero can look clean while the absolute number says you have a real problem. Track both.

Turn variances into fixes

An accuracy percentage tells you how you’re doing. It doesn’t tell you what to do. For that you need a reason for every miss.

Capture a root cause for each variance: mispick, receiving error, unit-of-measure mix-up, unposted transaction, damage, shrink, and so on. Then rank the causes and the SKUs by dollar impact. In my experience a small number of causes and items drive most of the loss. That’s my experience, not a universal statistic. Fix the top few and the number moves. Chase every miss equally and nothing does.

Keep the reason codes few and specific. “Other” shouldn’t be your biggest category.

A simple process that works

  1. Set the rules once: tolerance, accuracy target, and count frequency by ABC class.
  2. Count the first time, record the result: book quantity, count quantity, date, location, and a reason for any miss.
  3. Calculate from totals: sum the accurate lines and the counted lines for the period.
  4. Report three things: overall record accuracy, accuracy by ABC class, and the top variances by absolute dollars.
  5. Assign fixes to the top causes, then check next period whether those causes shrank.

Setting a target

Don’t grab a number off the internet. Targets quoted online depend on how the source defined accuracy and tolerance, so they usually aren’t comparable to yours. Run a few weeks of counts under your own rules, see where you actually are, and set a target you want to sustain. Then raise it when you’ve earned it.

Put it into practice

If you already have a cycle-count process, test it against the six mistakes above. Most operations I’ve seen fail at least two.

If you’d rather not build the workbook yourself, the LogisticsDeck Cycle Count & Inventory Accuracy Tracker does the calculation for you. Paste your cycle-count export from SAP, a WMS, or any other system, tell it once which column is which, and it works out record accuracy, quantity accuracy, absolute and net dollar variance, accuracy by ABC class, the top variances, root causes, and a weekly trend against your target. It’s a plain .xlsx with no macros.

View the Cycle Count & Inventory Accuracy Tracker →

Need help setting it up? See the setup guide →

The takeaway

Inventory accuracy isn’t a number you report. It’s a number you defend. Define “accurate,” count lines instead of units, measure the first count, break it out by class, calculate from totals, and look at dollars both ways.

Then ask the only question that matters: what are we going to fix because of this number?