Inventory Management in Warehouses: Cycle Counting, Accuracy & SKU Optimization

What 90% Inventory Accuracy Actually Means

Let me tell you what a 90% inventory accuracy rate actually means operationally. It means 10% of your SKU-location pairs are wrong. On a 20,000-location warehouse, that’s 2,000 locations where the system says something different from what’s physically on the shelf.

Every pick wave that touches those locations is a potential mispick, a stockout, or a phantom inventory event. Pickers go to the location, the product isn’t there, and the order either gets delayed while someone hunts for the item or it ships wrong. Customers don’t care why their order is wrong. And your operations manager doesn’t want to hear that the WMS said it was there.

Inventory accuracy is not an IT problem. It’s an operations discipline problem. And the operations discipline that solves it is cycle counting.

Why Cycle Counting Replaced the Annual Physical Inventory

Before cycle counting became the standard, the industry default was the annual physical inventory count: shut the operation down for one to three days, count everything, reconcile against the system, reopen. One snapshot per year, massive disruption, and still imperfect because a count taken in January is obsolete by February in a high-velocity environment.

Cycle counting distributes that work across the year. Small batches, daily, no operational shutdown. The operation keeps running. Accuracy is maintained continuously rather than corrected once annually.

The shift matters for a specific reason: errors caught during cycle counting are cheap to fix. Errors discovered when a picker can’t find a product during a live order are expensive — they create order holds, customer service escalations, and OTIF impacts. The closer to the error source you catch a discrepancy, the lower the correction cost.

Three Cycle Counting Methodologies

ABC Frequency Counting

This is the most common method, and the one I recommend for most operations as the backbone of their accuracy program.

Items are classified A, B, and C by velocity, value, or both. Higher-risk (higher-velocity, higher-value) items get counted more frequently — because they’re picked more often, which means more opportunities for errors to accumulate.

A typical program:

• A-items (top 20% of SKUs, 70–80% of picks): counted 4 times per year
• B-items (next 30% of SKUs, 15–25% of picks): counted twice per year
• C-items (bottom 50% of SKUs, 5–10% of picks): counted once per year

The math for a 50,000-SKU warehouse:

95,000 count events ÷ 250 working days = 380 counts per day. That’s two to three dedicated counters. Manageable. Systematic. And it puts your counting effort exactly where your accuracy risk is highest — A-items, which also happen to be the items whose inaccuracies cause the most damage.

Random Sampling

A statistically random selection of SKUs, regardless of velocity class, counted each period.

Random sampling gives you an unbiased snapshot of overall accuracy — useful when you want a facility-level accuracy score that isn’t influenced by where you chose to count. It’s also valuable when you suspect a specific zone or process is creating errors but haven’t identified the source yet.

When to use it: As a supplemental audit layer on top of ABC frequency counting. As the primary method for large catalogs where ABC classification hasn’t been established yet. For external audit purposes when you need unbiased facility-level accuracy data.

Control Group Counting

Control group counting is diagnostic, not primary. You select a fixed set of locations and count them repeatedly over time — not to maintain accuracy, but to detect patterns.

If the same locations keep showing discrepancies cycle after cycle, you’ve found a process problem, not a random error. Possible sources:

• A pick zone where pickers are consistently pulling from the wrong slot
• A receiving area where product is put away without WMS scan confirmation
• A dock where damage isn’t being captured at receipt
• A replenishment zone where double-putaway events are creating phantom inventory

Control group counting identifies the source of errors, not just the rate. Use it when your ABC frequency counting program shows consistent problems in the same area.

The distinction that matters: Cycle counting tells you your accuracy score. Control group counting tells you why that score is what it is. You need both.

Inventory Accuracy Targets: The Four Tiers

2024 industry average for order-picking accuracy: 99.15% (warehousingandfulfillment.com survey). That’s the average across all warehouse types — not a target but a baseline. Your goal for A-item locations specifically should be at or above that number, because errors on A-items are proportionally more damaging than errors on C-items. The WERC best-in-class threshold for order accuracy is above 99.85%.

Inventory shrinkage: The industry average improved to 1.24% in 2024 from 1.38% in 2023 (warehousingandfulfillment.com). Anything above 2% warrants investigation into receiving discrepancies, internal controls, and cycle count program rigor.

Cycle Count Accuracy: The Formula

The accuracy formula is simple but the measurement has a critical nuance:

Cycle Count Accuracy (%) = (Number of Correct Locations Counted ÷ Total Locations Counted) × 100

Accuracy is measured at the SKU-location level — both identity AND quantity must match the WMS record for a location to count as “correct.” A location where the right SKU is present but the quantity is off by two units is not a correct count. This is stricter than it sounds, and it’s why accuracy percentages in operations that measure correctly are often lower than operations that measure loosely.

SKU Rationalization: The Long Tail Problem

Here’s a concept most warehouse operations understand intellectually but don’t act on aggressively enough: the long tail.

In a typical warehouse, 20% of SKUs drive 80% of picks. The bottom 50% of your SKU catalog — the C-movers — generate only 5 to 10% of activity. But they consume pick locations, cycle count labor, replenishment cycles, and slotting decisions that cost real money every single day.

Every C-item sitting in a dedicated slot is a slot that could be used for a faster-moving B-item, eliminating a high-bay pick. Every C-item counted quarterly is a counter who isn’t counting an A-item that needs more frequent verification.

The SKU Rationalization Process

This is not a technology project. It’s a data analysis and a willingness to make decisions.

Step 1: Pull 12 months of SKU velocity data — picks, revenue, margin, inventory turns — from your WMS or ERP.

Step 2: Rank every SKU. Identify the true C-movers: anything with fewer than one pick per week.

Step 3: Ask three questions for each C-mover:

• Can this SKU be dropped from the catalog?
• Can it be consolidated with a similar or substitute SKU?
• Can it be moved to on-demand ordering — no warehouse stock, only ordered when a customer order comes in?

Step 4: Re-slot the remaining assortment with freed space. The locations formerly occupied by dropped C-items become available for B-item forward positions or expanded A-item forward pick depth.

Step 5: Monitor quarterly. Seasonal SKUs complicate this — a C-mover in Q3 may be an A-mover in Q4. Build that seasonality into your classification logic.

A Real Example

A warehouse applied Pareto-based ABC analysis to 12,000 SKUs, reclassified 22% of them into high-speed zones, and eliminated redundant slow movers. Results:

• Picking labor reduced 18%
• Fill rate improved to 98.7% for top customers
• $48,000 saved annually in replenishment labor alone

That’s not an automation project. That’s a velocity report, a spreadsheet, and a decision-making process that most operations have the data to run today.

Every SKU that shouldn’t be in your warehouse is costing you labor, space, and complexity every single day it sits there. SKU rationalization is one of the highest-ROI projects an operations team can take on — and it requires nothing but a velocity report and the willingness to cut.

The WMS Connection: How Technology Supports Accuracy

The cycle counting workflow in a modern WMS is designed to minimize friction:

1. WMS generates daily count tasks automatically based on the ABC frequency schedule
2. Counter receives tasks on their RF gun or mobile device — location, expected SKU, expected quantity
3. Counter physically counts, enters quantity on device without seeing the WMS expected quantity (blind count — prevents confirmation bias)
4. WMS compares against system record in real time
5. Discrepancies generate immediate exception records
6. Discrepancies above a defined threshold require a second-count verification before adjustment is accepted
7. Adjustments are logged with timestamp, counter ID, and reason code — full audit trail for analysis

The reason code field is critical. Over time, the pattern of reason codes tells you exactly where your process failures are. “Could not locate item” = putaway location problem. “Quantity mismatch — received short” = receiving accuracy problem. “Item in wrong location” = undirected putaway or pick confirmation failure. Reason codes turn your cycle count data from a score into a diagnostic tool.