AMR and AGV Fleet KPIs: Benchmarks for Utilization, Reliability and Throughput

Use these benchmarks as reference points for spotting a gap worth investigating, not as targets to copy. A fleet's numbers depend on layout, load mix and how long it has been running, so your own trend from the fleet manager logs says more than any single published figure.

Every figure shows its basis. A superscript number links to a public source in the Sources list, and the sentence gives that source's context. A figure marked no cited source is kept exactly as originally published because no public source stating it has been verified, so treat it as illustrative.

At a glance

Fleet utilization

Fleet utilization is the KPI most plant managers watch first, and it separates a justified fleet from an idle one. It is the share of scheduled uptime that robots spend moving loads or repositioning.

Typical brownfield AMR deployments run 45 to 60 percent utilization in the first year no cited source. World-class fleets hold 75 to 85 percent no cited source.

Above 88 percent you have no surge headroom and miss deadlines during demand spikes no cited source. Queuing theory explains the direction: as arrivals approach capacity, waiting time grows slowly at first and then rapidly¹.

How to measure

• Divide productive mission minutes by scheduled available minutes.
• Pull both straight from the fleet manager logs.
• Compare shifts and cells on the same scale.

What moves it

• Fleet availability, which sets the ceiling.
• Congestion and idle repositioning.
• Pickup and dropoff dwell and battery scheduling.

Check how many loads your fleet can move at its measured availability with the AGV Fleet Size Capacity calculator.

Availability and reliability

Availability and reliability set the ceiling on everything else. Target fleet availability of 98 to 99.5 percent for mature AGV lines no cited source and 95 to 98 percent for newer AMR fleets still tuning maps and traffic rules no cited source.

Mean time between failures (MTBF) should land above 400 operating hours per robot for good deployments no cited source, with world-class programs pushing past 700 hours no cited source. Mean time to repair (MTTR) belongs under 30 minutes for software resets and under 4 hours for mechanical swaps no cited source.

Track emergency stops (e-stops) per thousand missions too. Under 2 is healthy no cited source; over 8 signals map, sensor, or traffic-design problems worth a full audit no cited source.

How to measure

• Divide available robot hours by scheduled robot hours.
• Log every failure with the operating hours since the last one.
• Split repair time into software resets and mechanical swaps.
• Count e-stops per thousand missions, by zone.

What moves it

• Map and traffic-rule tuning on newer fleets.
• Map, sensor, and traffic-design fixes found in e-stop audits.
• Fast mechanical swaps instead of repairs in place.

See how availability sets delivered loads with the AGV Fleet Size Capacity calculator.

On-time delivery and mission success

On-time delivery and mission success reflect whether the fleet actually serves the line. Aim for 97 to 99 percent of moves completed within the agreed takt window no cited source.

Aim also for a mission success rate above 99 percent no cited source, where success means no manual intervention, no blocked-path abort, and correct pickup and dropoff. Typical early fleets sit at 92 to 95 percent and bleed hours to exceptions no cited source.

Watch the pickup and dropoff cycle, and benchmark station dwell at 20 to 45 seconds per stop for tote handoffs no cited source. Comparing station dwell with route congestion shows whether misses come from stations or from traffic.

How to measure

• Count moves completed within the agreed takt window.
• Count missions finished with no manual intervention, no blocked-path abort, and correct pickup and dropoff.
• Time station dwell per stop, station by station.

What moves it

• Station dwell at pickup and dropoff.
• Blocked paths and traffic congestion.
• Fewer exceptions that need a person to clear them.

Model delivered loads at your measured mission success rate with the AGV Fleet Size Capacity calculator.

Throughput per robot

Throughput per robot and moves per hour tell you if the fleet is sized right. A single tugger train pulling 4 to 6 carts commonly delivers 25 to 40 line-side deliveries per hour on a fixed milk-run loop no cited source.

A standalone tote AMR handles 12 to 22 moves per hour, depending on travel distance and lift time no cited source.

Distance matters. For example, every extra 30 meters of one-way haul at 1.2 to 1.5 meters per second adds 20 to 25 seconds to the outbound leg, or 40 to 50 seconds to a round trip, before acceleration and braking.

Benchmark payload utilization too, since carrying half-empty totes wastes trips. Target 80 percent volumetric or weight fill per move no cited source.

How to measure

• Count deliveries or moves per robot per scheduled hour.
• Record the one-way haul distance of each route.
• Log volumetric or weight fill for each move.

What moves it

• Haul distance and route layout.
• Lift time at each station.
• Consolidated loads, so totes travel full.

Size a milk-run loop with the Tugger Route Capacity calculator, and price each move with the Material Move Cost calculator.

Battery and charging health

Battery and charging health quietly caps daily output. Opportunity-charging AMRs should hold 90 percent or better state-of-charge availability across a shift no cited source, spending under 12 percent of scheduled time on chargers when using 20 to 30 kW fast charging no cited source.

If robots idle at chargers more than 18 percent of the shift, you are either undersized on chargers or oversized on battery cycling no cited source.

Watch depth of discharge. Keeping lithium packs between 20 and 80 percent extends useful cycle life toward 3,000 to 4,000 cycles no cited source.

Battery chemistry matters more than the charge window. In Sandia National Laboratories' 2020 tests of commercial cells, lithium iron phosphate (LFP) cells lasted 2,500 to 9,000 equivalent full cycles, against 250 to 1,500 for nickel cobalt aluminum (NCA) and 200 to 2,500 for nickel manganese cobalt (NMC) cells². A 20 to 80 percent window extended life measurably only for NMC².

How to measure

• Log state of charge for each robot through the shift.
• Record time on chargers as a share of scheduled time.
• Track depth of discharge and cycle count for each pack.

What moves it

• Charger count matched to mission demand.
• Charge windows scheduled around mission peaks.
• Battery chemistry and charge window chosen together.

Traffic and congestion

Traffic and congestion metrics are the levers most teams overlook. Benchmark path conflict rate below 3 percent of segments per hour no cited source and average wait at intersections under 5 seconds no cited source.

Once robot density passes roughly 1 robot per 250 to 400 square meters of shared aisle, congestion losses climb fast and can strip 10 to 20 points off utilization no cited source. Flag choke points before they cascade.

How to measure

• Count path conflicts per segment per hour.
• Time the average wait at each intersection.
• Calculate shared aisle area per robot.

What moves it

• Passing bays.
• One-way loops in place of bidirectional aisles.
• Staggered release times.
• High-traffic pickup stations moved off the main artery.

These fixes are cheap relative to buying more robots.

Labor redeployment

Add labor redeployment as a business KPI next to the operational ones. A mature fleet should free 1.5 to 3.0 material-handler full-time equivalents (FTEs) per 8 to 10 robots no cited source.

How to measure

• Count material-handler hours moved to other work.
• Convert them to full-time equivalents for each group of robots.

What moves it

• The share of moves shifted from manual handling to the fleet.
• Exceptions that still need a person.

Quantify the freed labor with the Internal Logistics Labor Savings calculator.

Improve in the right order

The improvement sequence matters as much as the targets. Start by lifting availability, because it caps utilization.

For example, a fleet available 93 percent of scheduled time can never exceed 93 percent utilization, so low availability eats the margin between an 80 percent utilization target and its ceiling.

Next attack congestion and idle repositioning, which are usually worth 8 to 15 utilization points with zero capital no cited source. Then tighten pickup and dropoff dwell and battery scheduling.

Only after those levers are exhausted should you add robots. Teams that follow this order typically raise effective throughput 30 to 50 percent on the installed fleet no cited source, deferring capital spending and improving payback.

The sequence

• Lift availability.
• Cut congestion and idle repositioning.
• Tighten pickup and dropoff dwell and battery scheduling.
• Add robots only after those levers are exhausted.

Model the payback of the installed fleet with the AMR ROI calculator, or of pallet automation with the Autonomous Forklift Payback calculator.

How to build your own baseline

Set a small dashboard and review it weekly, not quarterly. The core seven are utilization, availability, MTBF, on-time delivery, mission success, moves per hour, and congestion score.

• Pull every KPI from fleet manager logs, not vendor spec sheets that assume ideal, congestion-free conditions.
• Write each definition down and keep it fixed.
• Review the dashboard weekly and note what changed whenever a KPI moves.
• Size growth on measured throughput with the AGV Fleet Size Capacity calculator and the Tugger Route Capacity calculator.
• Compare against your own trend first and published ranges second.