Bench test · 19 Nov 2026
Battery runtime: capacity, chemistry and the claim that shrinks in year two
A robot vacuum's runtime claim is measured on a new cell in a lab condition. Three years later the same machine cleans for half as long, and the reason is chemistry rather than usage.

Capacity, energy and the units that get confused
A cell is rated in milliamp-hours, which is charge, not energy. Energy is charge times voltage, and a 4,000 mAh cell at 3.7 volts holds about 14.8 watt-hours.
A robot's runtime is energy divided by power draw, and the power draw depends on the mode. A machine that draws 15 watts runs about an hour on that cell in theory and less in practice.
The claim on the box is often measured at the lowest power mode with the brush off. The number is real and it describes a mode nobody uses.
A useful claim states the runtime in the standard cleaning mode, on a stated surface, with the battery at a stated temperature.
Chemistry and the calendar
Lithium-ion cells lose capacity with age and with cycles. The loss is fastest in the first hundred cycles and in a warm environment, and it is permanent.
A robot that sits on a charger at 100 per cent in a warm room ages faster than one that is allowed to cycle. Charge management - stopping at 80 per cent or delaying the top-up - is a genuine life extender.
The chemistry matters: nickel-manganese-cobalt cells hold capacity better than lithium iron phosphate at the same volume, and iron phosphate cells last longer in cycles.
A brand that states the cell chemistry and the expected capacity at 500 cycles is giving the buyer a number that predicts year three.
The charger, the dock and the parasitic load
A dock that keeps the robot at full charge draws power continuously, and the standby draw is part of the product's energy use even when it is not cleaning.
The charging circuit's efficiency decides how much of the mains power reaches the cell. A poor circuit turns a watt-hour into heat in the dock.
The battery's management circuit should balance the cells and protect against over-discharge. A cheap pack without balancing loses capacity unevenly and fails early.
The specification should name the standby power and the charge time from empty. Both are numbers a buyer can compare and a retailer can audit.
Runtime in the real house
A real clean includes navigation, turns, carpet transitions and the dock approach. Each of those draws power, and the sum is shorter than a straight-line test.
A house with thresholds and rugs uses more energy than an open-plan flat of the same area. The runtime claim should be tied to an area and a floor type, not to a time alone.
The robot's own software can extend the run by returning to charge and resuming. A resumable clean is a feature that converts a short runtime into a completed job.
The claim should state whether the clean resumes. A machine that stops at 60 per cent and starts over is a machine that never finishes a large floor.
Replacement, and the pack that is not user-serviceable
Most robot batteries are replaceable by the user, and some are not. A sealed pack turns a two-year-old robot into waste, and the buyer should know before purchase.
The replacement pack's availability and price are part of the product's cost of ownership. A pack that costs a third of the robot's price changes the economics of keeping it.
A pack with a stated cycle life and a published part number is a pack the brand expects to support. A pack with neither is a pack the brand expects to sell once.
The specification should say whether the pack is user-replaceable and give the part number and the expected price.
Testing the claim without a lab
A buyer can test a runtime claim at home with a stopwatch and a defined area. Run the robot in the standard mode on a hard floor until it returns to the dock, and record the time and the area covered.
Repeat with a carpeted room, and compare the two. A product that matches its claim on hard floor and falls short on carpet is a product whose claim was written for hard floor.
The test should be run three times, because the first run includes mapping and the later runs do not. The average of the later runs is the number to compare.
A supplier who welcomes that test is a supplier whose claim is honest. A supplier who discourages it is a supplier whose claim is the lowest mode.
Charging in a hot room
A robot that charges in a conservatory in summer charges a battery at forty degrees, and a battery at forty degrees ages faster than one at twenty-five.
The charge circuit's temperature management should reduce the charge rate when the pack is warm. A fast charge in a hot room is a capacity loss in year two.
The dock's placement advice should mention temperature, because the dock is usually placed where the furniture allows rather than where the battery would prefer.
A dock with a temperature sensor and a reduced charge mode is a small feature with a measurable effect on pack life.
The deep discharge that never happens
A robot returns to its dock before the battery is empty, so the pack lives in a narrow band between forty and one hundred per cent.
That band is gentler than a full cycle, and it is why robots often outlast laptops with the same chemistry. The management circuit still needs to balance the cells periodically.
A pack that is never balanced drifts, and the drift shows as a runtime that falls faster than the capacity loss predicts.
The firmware should schedule a balance charge, and the service manual should say so. A pack that is replaced because it was unbalanced is a warranty cost that could have been avoided.
The spare pack and the two-year question
A buyer choosing between two robots should ask what a replacement pack costs and whether it is available. The answer predicts the product's life at year three.
A pack that is priced at a fifth of the robot is a pack that will be bought; a pack priced at a third is a decision the user may not make.
The pack's connector and its form factor should be standard enough that a third-party pack can be fitted. A proprietary pack is a lock-in, and buyers increasingly notice.
The specification should list the pack's part number, its capacity, its chemistry and its price. Four numbers that make the product comparable.
The claim, the mode and the marketing
A runtime claim that names a mode nobody uses is technically true and practically misleading, and regulators in some markets treat that as a problem.
The claim should be the runtime in the default mode on the default surface, with the longer and shorter modes given separately.
A brand that publishes all three numbers is a brand that is confident in its product, and the numbers are also the ones the support team needs.
The test method behind the claim should be written down and available on request. A method that cannot be produced is a claim that cannot be defended.
The software that spends the battery
Navigation, mapping and the radio all draw power, and a robot in a new house spends more of its battery mapping than it does cleaning.
The first clean after a reset is usually shorter, and the specification should say so rather than let the buyer conclude the runtime claim is false.
The robot should complete the map and then clean, and the app should show the difference between a mapping run and a cleaning run.
A product that explains its own first-run behaviour is a product that is trusted, and the explanation costs a sentence in the app.
The short version
A runtime claim without a mode, a surface and a temperature is a marketing sentence. The engineering numbers are the cell chemistry, the charge management and the pack's price at year three.
A pack that can be replaced and whose part number is published is the difference between a product with a life and a product with a term.
A pack that states its chemistry and its capacity at five hundred cycles is the one behind the laser-navigation sweeping robot, and its charge management is the feature that predicts year three.
What to check
- Runtime stated per mode, per surface and at a stated temperature
- Cell chemistry and expected capacity at 500 cycles
- Charge management strategy (80 per cent stop or delayed top-up)
- Standby power and charge time from empty
- Whether the clean resumes after a recharge
- Battery user-replaceable, with part number and price