Performance Metrics

Deployment data, not lab conditions

All metrics from active warehouse deployments running Automcore in production. Fleet sizes range from 4 to 16 arms. SKU categories span rigid consumer goods through deformable flexible packaging. Ranges reflect real variance across facilities and SKU mixes — not controlled benchmark conditions.

Per-arm pick rate — mixed SKU

380–640 picks/arm/hr

Mixed-SKU depalletizing deployments across 12-arm fleets. Range reflects SKU category variance — rigid consumer goods at the upper end, flexible packaging and irregular shapes at the lower. Average across 90-day deployment windows.

First-session validation pass rate

97.4%

Arms that receive and validate a propagated model without requiring remedial training. Measured at point of model activation.

Model compression ratio

94% smaller

Compressed grasp model vs. raw training data. Enables fleet distribution over standard gigabit facility LAN without congestion.

Fleet propagation window

<8 hrs

End-to-end model distribution to a 12-arm fleet during a maintenance window. Scales sub-linearly with arm count.

Benchmark Data

Pick rate by SKU category

Benchmarks measured over 30-day windows per deployment. SKU category definitions follow standard warehouse classification (ANSI MH1.8).

SKU Category	Pick Rate (fleet avg)	Confidence Score	Typical SKUs/Fleet	Notes
Consumer Boxed Goods	580–640/hr	98.1%	80–120	Best-performing category. Rigid geometry, consistent weight distribution, repeatable pose at depalletization point. Vacuum cup end effector optimized.
Beverage (canned / bottled)	500–560/hr	97.8%	30–60	Cylindrical grasp profiles; multi-cup vacuum configuration required for cans. Bottle neck grasp requires mechanical gripper or hybrid end effector.
Flexible Packaging	360–420/hr	92.4%	40–80	Deformable surface requires force-torque calibration. Initial training session longer (~75 min). Model confidence stabilizes after 3 propagation cycles. UR10e and HC20 preferred for force control.
Irregular Shapes (<2 kg)	310–390/hr	89.3%	20–40	Multi-point grasp plan required; 6-DOF pose estimation critical for orientation variance. Confidence improves as more arms in the fleet contribute grasp-quality feedback over time.
Heavy Items (2–15 kg)	280–360/hr	91.6%	15–25	Payload-appropriate arm required: UR16e (16 kg), FANUC CR-35iA (35 kg), or ABB IRB 4600-beta (60 kg). Lower cycle rate reflects higher inertia and slower Cartesian path planning constraints.

Fleet Propagation

Propagation timing by fleet size

Measured over facility gigabit LAN. Propagation runs staggered to prevent network saturation — total window includes distribution + per-arm validation.

4 arms

~2.2 hrs

8 arms

~4.5 hrs

12 arms

~7.2 hrs

16 arms

~8.4 hrs

20 arms

~10 hrs

Fleet sizes above 12 arms represent projection data from current deployment architecture. Field data from 20-arm cohort expected Q4 2026.

Want full benchmark data for your arm configuration?

We'll prepare a performance projection tailored to your fleet model, SKU variety, and operating schedule.

Request a Demo Read the Benchmark Report