Your Pick Accuracy Is 99.5%. So Why Are Customers Still Getting the Wrong Item?

A customer doesn't care that your pick accuracy is 99.5%.

They care that they ordered a blue shirt and received a red one. Or that an item was missing from the box. Or that it happened twice in six weeks.

Meanwhile, the warehouse dashboard still looks healthy.

That disconnect is more common than it should be, and it usually comes down to a simple question:

What exactly are you counting as a picking error?

Pick accuracy is useful, but only when everyone understands what sits behind the number. Depending on how the operation and WMS are configured, the metric may primarily tell you whether system-directed picking transactions were completed correctly.

That's not always the same thing as measuring what the customer received.

WERC's DC Measures benchmarking study includes pick accuracy as a core warehouse performance measure. Like most warehouse KPIs, though, the value of the number depends heavily on how the operation defines and captures the events behind it.

If your operation reports strong pick accuracy while customer complaints, short picks, and inventory adjustments remain stubbornly high, it's worth looking past the headline KPI.

First, Make Sure You Know What You're Measuring

Pick accuracy sounds like a straightforward calculation until you start comparing how different operations calculate it.

Some measure accuracy by order.

Others measure it by order line.

Others use units or individual pick transactions.

Those aren't interchangeable.

Take an order with ten lines. Nine are picked correctly and one is wrong.

Measured by line, that's 90% accuracy.

Measured by order, that order is simply wrong.

Same picker. Same order. Same mistake. Very different KPI.

This becomes even more important when comparing buildings, clients, or 3PL operations. Two sites can have nearly identical execution on the floor and still report different accuracy rates because they're measuring different things.

Before worrying about whether 99.5% is good enough, make sure you can answer four questions:

• What is the denominator? Orders, lines, units, or pick transactions?
• At what point is a pick considered correct?
• Which exceptions reduce the accuracy rate?
• Which exceptions are tracked somewhere else?

That last question is usually where things get interesting.

Where Pick Accuracy Starts to Break Down

The cleanest warehouse transactions are also the easiest ones for the WMS to measure.

The system sends a picker to a location. The picker scans the location and item. The quantity is confirmed. The transaction closes.

Those events are easy to report.

The messier transactions aren't always handled the same way.

Short picks

The system directs ten units, but only seven are physically in the location.

What happens next depends on the WMS and the process. The picker may short the task, an exception may be created, replenishment may be triggered, or someone may manually resolve the remaining quantity.

The short pick may be captured perfectly as an exception and still never reduce the headline pick-accuracy percentage.

From the customer's perspective, however, it matters quite a bit if the order ultimately ships short.

Substitutions

An operation can't find the requested SKU and someone approves a substitute.

If that substitution is formally transacted through the WMS, the history may be clean.

If it's handled through a manual process, customer-service workflow, or supervisor decision outside the normal picking path, the standard pick report may never see it.

Manual picks

Rush orders, recovery work, damaged inventory, problem orders, and other exceptions occasionally push teams outside the normal workflow.

Someone grabs inventory from overstock or another location and gets the order out the door.

Operationally, the problem is solved.

Systemically, you may now have a transaction that doesn't look anything like a normal pick.

Customer-reported errors

This is the category warehouses tend to underestimate.

In one client operation, wrong-item complaints doubled over two quarters while the pick accuracy report never dropped below 99.4%. The complaints lived in a customer-service ticketing tool nobody in the warehouse opened. A single afternoon of matching tickets to pick history pointed at three SKUs stored in the same zone, all sharing near-identical packaging.

If the wrong item makes it all the way through picking, packing, shipping, and delivery, the first reliable record of the error may be a customer-service ticket.

If customer complaints live in a separate system and nobody routinely reconciles them against WMS activity, you can have two versions of reality.

The WMS says the pick was correct.

The customer says it wasn't.

Both systems are reporting exactly what they were designed to capture.

The Inventory Problem Hiding Behind the Picking Problem

The customer-facing error is only part of the issue.

A bad pick can also create an inventory problem.

Suppose SKU A is physically picked while the system believes SKU B was picked.

The customer receives the wrong item, but now two inventory records may also be wrong.

One SKU is physically missing even though the system thinks it's still there.

The other may have been decremented even though it never left the building.

On one engagement, we traced a single mis-pick through the client's ledger. The system believed the ordered SKU had shipped, so the on-hand count stayed positive for nearly a month. The shelf was empty and replenishment never triggered. Four more orders shorted against that location before a cycle count surfaced the variance.

The shipping error gets attention because a customer complains.

The inventory error can sit quietly until the next picker hits the location, replenishment fails, or a cycle count eventually finds it.

That's one reason pick exceptions, inventory adjustments, shorts, and cycle counts shouldn't be viewed as completely separate problems.

They're often different symptoms of the same process breakdown.

A cycle count program can correct the inventory record. But if nobody traces why the variance happened, the process that created the error stays in place.

The location gets corrected today and goes wrong again next week.

99.5% Can Still Be a Lot of Errors

High percentages can also hide the scale of the problem.

Take an operation shipping 1,500 orders per day.

If pick accuracy is measured at the order level and the operation truly misses 0.5% of those orders, that's about seven or eight incorrect orders every day.

Across a seven-day operation, you're around 50 errors a week.

Suddenly 99.5% doesn't feel quite as perfect.

And each error can carry more than one cost. Return processing, replacement freight, customer-service time, inventory research, rework, and potential credits or concessions.

And, for a 3PL, another conversation with a client who was told warehouse accuracy was performing well.

This doesn't mean 99.5% is a bad result.

It means percentages need context.

Build the Measurement Around the Exceptions

I wouldn't recommend throwing away your existing pick-accuracy metric.

Start by understanding its blind spots.

Pick one consistent unit of measurement, orders, lines, units, or transactions, and keep it stable. Changing the denominator every few months makes the trend almost useless.

Then start tracking the exceptions separately.

At minimum, I would want visibility into four categories.

• Short picks
• Substitutions
• Manual picks or overrides
• Customer-reported wrong-item shipments

Don't force all four into one giant KPI immediately.

Keeping them separate is often more useful because each one points toward a different operational problem.

Short picks clustered in one zone may point toward inventory accuracy or replenishment.

Repeated substitutions on a handful of SKUs may point toward slotting, receiving, or item-master issues.

Manual picks concentrated on one shift may expose a workflow problem.

Customer complaints tied to specific pack stations may tell an entirely different story.

This is where inventory accuracy and exception control becomes much more valuable than simply watching a percentage move up or down.

You're trying to understand what kind of failure is happening, where it's happening, and what process keeps creating it.

You Don't Need a New Dashboard to Start

You can start with a spreadsheet.

Pull 60 to 90 days of pick and exception history.

Add customer-reported errors if you can get them.

Then start segmenting.

Look at SKU, location, zone, shift, picker, day of week, and exception type.

You're trying to find out what the number isn't showing you.

If short picks are the biggest concern, start there.

Track them for a month. Find the locations and SKUs generating the most exceptions. Compare those records with cycle counts and adjustments.

One good exception report can tell you more about what's happening on the floor than another decimal place added to the accuracy KPI.

A Good Metric Should Match the Operation

Pick accuracy is still a useful warehouse KPI.

But no single percentage can describe everything that happens between releasing an order and putting the correct product in the customer's hands.

Know what your number measures.

Know what it leaves out.

Then track the exceptions separately and compare what the WMS reports with what your customers are actually experiencing.

When those two views start telling the same story, the metric becomes much more useful.