KPIs for warehouse automation

Picks per hour (PPH)

The primary productivity metric for picking automation. Measure at the operator level for GTP stations. Baseline manually before automation, then track weekly post-go-live. Target: 20-50% improvement versus manual, depending on implementation maturity.

Orders shipped per day

Total order output from the automated area per operating day. Track against peak and average targets. Useful for capacity planning and for validating that throughput assumptions in the business case are holding.

Robot task completions per hour

For AMR and GTP systems: the number of tasks (movements, retrievals, deliveries) completed by the system per hour. Useful for understanding system utilisation and identifying bottlenecks in fleet management or WMS integration.

Units per labour hour (UPLH)

Total units processed divided by total labour hours in the area. The composite productivity metric that accounts for both automation performance and remaining manual labour. This is the metric most directly connected to labour cost reduction.

Order accuracy rate

Percentage of orders shipped without error. Baseline before automation. Automation typically improves accuracy in picking operations (directed picking at GTP stations reduces pick errors). Track separately for automated and manual areas.

Inventory accuracy rate

Percentage of inventory locations where the physical count matches the WMS record. Automation is sensitive to inventory inaccuracy. A declining inventory accuracy rate in the automated zone is an early warning of integration or data quality issues.

Pick error rate

Errors per 1,000 picks. Separate from order accuracy: one pick error can affect multiple orders. Track root causes: wrong item, wrong quantity, wrong location scan. Use to distinguish system errors from operator errors.

System uptime

Percentage of scheduled operating time the system was available. Distinguish between planned downtime (maintenance, software updates) and unplanned downtime (failures). Unplanned downtime is the relevant metric for SLA management. Target: 97-99% unplanned uptime after stabilisation.

Robot utilisation rate

Percentage of available robot time spent on productive tasks (not waiting, charging, or in error state). Low utilisation indicates fleet sizing is too large or WMS task flow is creating idle time. High utilisation (above 85%) indicates capacity constraint risk at peak.

Exception rate

Percentage of tasks requiring manual intervention. Track by exception type (navigation, load, integration, operator). An increasing exception rate over time is a diagnostic signal for operational, data, or integration issues. Target: below 2% after first 90 days.

Travel distance eliminated

For AMR and GTP systems: reduction in total picker or operator travel distance versus pre-automation baseline. Useful for validating the labour saving assumption (travel time reduction is the primary driver of labour saving in GTP operations).

Labour cost per order line

Total labour cost in the automated area divided by order lines processed. This is the most direct financial metric for measuring automation return. Track monthly and compare against business case projection and pre-deployment baseline.

Cost per pick

Total operational cost (labour, maintenance, software, facility) divided by total picks. A complete cost-per-pick metric includes all cost components, not just direct labour. Useful for benchmarking against manual operation and vendor business case claims.

Payback tracking

Cumulative net saving (actual savings minus actual costs) tracked against the business case payback projection. Report quarterly to operations management and finance. Update the payback forecast if actual performance differs materially from projections.

Measure in the target area only

Baseline should be specific to the area and tasks that will be automated. Whole-warehouse averages will dilute the measurement. If you are automating the pick area, measure pick-specific KPIs in that zone.

Measure over a representative period

A 4-week baseline during a non-peak period is the minimum. Include at least one peak period in the measurement window if your operation has seasonal variation. Peak period performance is often more important than average performance for automation sizing.

Include all cost components

Labour baseline should include all cost components: direct labour, supervision, agency premium, and overtime. Excluding any component will overstate the savings rate when automation is deployed.

Document methodology for replication

Document how each baseline metric was measured so the same methodology can be applied post-deployment. Inconsistent measurement methodology is the most common reason automation improvements cannot be demonstrated credibly.