Start by defining what counts as a pick error

Knowing how to audit pick accuracy gives you evidence of what is actually leaving the warehouse, where errors are entering the process, and whether the operation can support your next stage of growth.

A useful audit is not a request for a headline KPI from a 3PL. It is a controlled comparison between the order that should have been picked, the stock that was picked, and the controls that were meant to prevent a mismatch. It should also distinguish warehouse picking errors from upstream failures such as bad product data, stock adjustments and unclear order rules.

Before reviewing samples, agree the measurement rules. Providers can report very different accuracy figures while both are technically correct. One may measure orders picked correctly; another may measure individual order lines. A third may remove cancelled orders, substitutions or customer amendments from its calculation. Those choices matter.

For most ecommerce operations, report both measures:

• Order accuracy: correctly picked orders divided by total audited orders.

• Line accuracy: correctly picked order lines divided by total audited lines.

An order containing five lines is only fully accurate if every required line is correct. Line accuracy is useful for diagnosing complexity, but it can make performance look better than the customer experience when multi-line orders are common.

Set error categories before the audit begins. At a minimum, separate wrong SKU, wrong variant, short pick, overpick, duplicate pick, incorrect bundle component, incorrect batch or expiry date, and missed handling instruction. Include dispatch errors only where they arose during picking. A parcel sent to the wrong address is serious, but it is not evidence that the picker selected the wrong item.

This classification prevents an awkward but common outcome: a warehouse accepts a low accuracy result without learning what needs to change.

How to audit pick accuracy using a representative sample

The sample must reflect the operation rather than the easiest orders to check. Auditing only single-SKU DTC orders picked on a quiet Tuesday will not tell you much about performance during a promotion, a wholesale dispatch run or the late carrier cut-off.

Start with a defined audit window, usually two to four weeks for an initial review. Select orders across channels, shifts, zones and order types. If the brand sells apparel, include size and colour variants. If it sells cosmetics, food or health products, include batch-controlled and expiry-sensitive lines. Where wholesale forms part of the mix, include carton picks, retailer-specific labels and routing-guide requirements.

The right sample size depends on order volume and the risk attached to failure. A brand shipping 500 orders per month may physically inspect a meaningful proportion of completed orders. A brand shipping 50,000 orders will need a statistically sensible sample, supplemented by targeted checks of high-risk workflows. The aim is not perfection in one inspection exercise; it is enough evidence to identify a reliable error rate and repeatable patterns.

Build three sample groups rather than relying on random selection alone. Use a random group to establish the baseline. Add a risk-based group containing complex orders, fast-moving variants, bundles, replenishment arrivals and peak-period orders. Then inspect exception orders, including short-shipped orders, customer complaints, manual stock adjustments and replacements. Exceptions are not representative of average performance, but they are often the quickest route to the root cause.

For each order, compare the original order record, warehouse management system pick record, scan events, packing record and physical contents where possible. If the audit happens after dispatch, use returns, customer-service cases, dispatch photographs and weight data carefully. They can indicate an issue, but they are weaker evidence than a physical verification before the parcel leaves the site.

Follow the order through the warehouse controls

A pick-accuracy percentage tells you the outcome. The audit should establish whether the process is capable of producing that outcome consistently.

First, inspect how the order reaches the warehouse management system. Check channel integrations, SKU mapping, bundle logic, substitutions, hold rules and cut-off handling. A wrong item may be picked flawlessly against a wrong order instruction. That requires a data or integration fix, not a picker performance conversation.

Next, examine location design and item identification. Similar-looking packaging, adjacent variants and duplicate SKUs across client accounts create predictable risk. Good controls may include barcode scanning at pick and pack, location check digits, clear bin labels, image prompts and rules that prevent the wrong batch from being selected. Manual processes can work at lower volume, but the control needs to match the consequences of a miss.

Then review the verification point. Some operations scan every item at pick; others use pack-station scanning as the final control. Neither approach is automatically better. Pick scanning can stop errors earlier, while pack verification can catch problems created during consolidation. The sensible choice depends on SKU count, order profile, warehouse layout and labour model. For high-value, regulated or batch-sensitive goods, a single manual check is rarely enough.

Ask to see what happens when a scan fails. Can staff override the warning? Who authorises it? Is the reason captured? Frequent overrides are a watch point even when the published accuracy metric appears strong. The same applies to manual order edits, negative stock positions and stock moved without a scan.

Analyse errors by pattern, not just by picker

Once results are logged, cut the data into operational segments. Look at error rate by SKU, product family, warehouse zone, shift, picker, order size, channel, day of week and order age. The purpose is not to build a league table of staff. It is to find whether the process fails under specific conditions.

A cluster around one SKU may point to poor master data or confusing packaging. Errors concentrated in a particular aisle may indicate location labelling, replenishment discipline or lighting. A rise late in the day may reflect carrier cut-off pressure, staffing levels or rushed replenishment. If marketplace orders perform poorly while DTC orders do not, inspect channel-specific mapping and service rules before changing warehouse labour processes.

Track the cost alongside the rate. A wrong low-value accessory and a wrong serialised device are both one error in the KPI, but they do not carry the same exposure. Record replacement postage, product cost, customer-service time, chargebacks, retailer penalties and stock write-offs. This gives operations and finance a shared basis for deciding which corrective actions justify investment.

Test inventory accuracy at the same time

Pick accuracy and inventory accuracy are connected but not interchangeable. A picker may scan the correct location and still send the wrong item if the bin contains mixed stock. Equally, a warehouse can have accurate stock records while staff select an adjacent variant under pressure.

During the audit, compare system stock with physical stock in the locations sampled. Investigate unexplained stock adjustments, recurring negative stock and repeated replenishment discrepancies. Cycle-count frequency should reflect product risk and velocity. Fast-moving, high-value or batch-controlled items usually need closer attention than slow-moving standard goods.

If a 3PL manages multiple clients, confirm that your stock is physically and systemically segregated. Similar products across accounts are a particular risk where client identifiers, labels or barcodes are inconsistent.

Turn findings into an accountable improvement plan

A credible audit ends with a short corrective-action register, not a vague commitment to be more careful. For every material error pattern, document the cause, the control being changed, the owner, the deadline and the measure that will show whether the fix worked.

For example, a wrong-variant issue might lead to revised location labels, mandatory barcode validation and a temporary second check for the affected range. A bundle error may require a change to bill-of-materials logic and a test order before the update goes live. If peak-period accuracy falls because replenishment is late, the remedy could be revised replenishment triggers and protected labour rather than another pack-station check.

Re-audit after the agreed change, using the same definitions and comparable sample groups. A one-off improvement after managers are warned of an inspection is not enough. What matters is sustained performance through normal trading, promotions and peak volume.

When assessing a new or replacement 3PL, ask for the underlying method behind its pick-accuracy claim: the denominator used, error exclusions, scan points, override controls, exception reporting and peak-period results. The cheapest quote means very little if the operation behind it cannot protect the order experience your customers expect. A well-run audit gives you a practical way to test that fit before small fulfilment errors become a costly pattern.