Robot Vacuums: What the Extra Money Actually Buys

01

Navigation is the difference you actually feel

A cheap robot navigates by bouncing. It drives until it hits something, turns at a semi-random angle, and drives again. Given enough time that covers most of an open room, which is why budget machines often perform acceptably in a studio flat and badly in a three-bedroom house. The failure is not the dirt left after one pass, it is that nothing about the run is predictable. There is no way to send it to the kitchen, and no way to know whether it finished the floor or ran flat under the sofa. It will cross the same stretch of hallway six times and never reach the far corner of the bedroom.

Lidar is a small spinning laser turret, usually a raised bump on the lid, taking a few thousand distance readings a second and assembling them into a floor plan. It works in complete darkness, holds its position without landmarks, and is the most dependable navigation sold at consumer prices. What it does not do is see the floor. The laser sweeps at deck height, so walls and table legs land on the map while a phone charger does not, which is why bump sensors are still working away underneath it.

vSLAM means visual simultaneous localisation and mapping, which in practice is a camera watching the ceiling or the room ahead and tracking recognisable features to work out where the robot is. It costs less, allows a slimmer body with no turret, and loses the thread in dim rooms and long blank corridors. Sitting between the two is a gyroscope-and-odometry tier that drives tidy parallel lines but saves nothing between runs. It looks organised on the floor and still cannot be told to clean one room, which is the distinction that matters.

What a map actually buys you is control. Named rooms you can dispatch the robot to, keep-out zones drawn around the pet bowls and the cable nest behind the desk, straight overlapping passes instead of wandering, and the ability to recharge halfway through and resume from the same spot. In anything bigger than a one-bedroom flat that last feature decides whether the machine finishes the floor or quietly stops at seventy percent. This is the first real price step in the category, and it is the one I would always pay.

02

Suction numbers are not a standard

Pascals measure pressure, and the pascal figure on the box is the vacuum a fan can pull against a sealed port. That is not airflow, and airflow is what moves debris. No shared test standard exists, so one brand's number cannot be compared against another's, and the measurement is usually taken at the fan rather than at the floor, where the seal, the ducting and a half-loaded filter have already taken their cut. A machine with a lower headline figure and a better sealed floor head will often pick up more of what is actually on your carpet.

Pickup is decided by brushroll design, how closely the suction opening rides the floor, and whether the air path is short and smooth. Rubber or silicone fin brushrolls deal with hair far better than nylon bristles, which weave long strands into a solid collar around the shaft. Deep pile carpet deserves a blunt warning as well. No robot at any price extracts embedded grit the way a mains-powered upright with a driven beater bar does, because the robot has neither the weight to press the head into the pile nor the mains power to spin the brush against that resistance. Extra pascals do not close the gap.

Two numbers would genuinely help and almost nobody publishes them: airflow measured at the floor head, and how far performance drops as the bin fills. In their absence, use what you can see. A wide intake spanning most of the robot's width, a pleated filter with real surface area rather than a flat scrap of foam, and a side brush that feeds debris inward instead of firing it across the room are all visible in photographs and teardown videos. Carpet boost, where the robot senses pile and raises fan speed, is worth having and costs both battery and quiet, which is why it is often shipped switched off.

03

What a self-emptying base really costs

A self-emptying dock pulls the contents of the robot's small onboard bin into a larger container in the base, normally after every run. The point is not laziness. Robot bins are tiny, emptying is the chore people skip, and a robot running with a half-full bin and a dusty filter cleans measurably worse than the same robot cleaned out. Moving that job to once every month or two keeps the machine near full performance, and sealing the dust into a bag rather than tipping it into a kitchen bin is a real benefit in an allergy household.

The cost is ongoing and most buyers underestimate it. Bagged docks use proprietary bags, and across a few years those bags add up to a noticeable fraction of what the robot cost. Bagless docks avoid the consumable but hand you a compacted pile of dust to tip out, which undoes part of the reason you bought one. Look up the replacement bag price on the same page as the robot, multiply it by how often you plan to run the machine, and treat that as part of the purchase rather than a surprise arriving in month five.

Then there are the costs that are not money. A dock is furniture, often as tall as a kitchen bin, and it needs a socket, clear space either side for the robot to align its charging contacts, and room to reverse in. The transfer tube is a genuine failure point, because a sock or a rope of wound hair jams it while the cycle carries on sounding normal. The emptying blast itself is loud enough to interrupt a conversation, which matters if the dock lives near a bedroom and the robot runs at seven in the morning. And it empties the bin only, so the brushroll, the filter and the sensors are still yours to clean.

04

Mopping is mostly theatre until it is not

Mopping on most robots is a damp microfibre pad clipped underneath and dragged across the floor by a machine that weighs a few kilograms and puts almost none of that weight onto the cloth. It lifts a film of dust and makes a floor look better for an afternoon. Dried juice stays where it is, a greasy patch near the hob is untouched, and anything sticky gets spread across a wider area than it started with. As a dust-damping pass between proper mopping the feature earns its place. Sold as a replacement for mopping, it is theatre.

The expensive implementations change this more than sceptics expect. Spinning discs or oscillating pads supply the agitation a dragged cloth lacks, pads that lift or detach keep water off rugs, and docks that wash and then dry the pad solve the real hygiene problem, which is a wet dirty cloth sitting in a warm room for a week. The catch is that those docks need somewhere sensible to live, introduce a second set of consumables, and want their own periodic scrubbing or they develop a smell of their own. If your floors are mostly carpet, ignore the mopping column and put the money into navigation and brush design.

Water logistics decide whether you actually use any of it. A small onboard tank covers a modest hard floor once, so a large ground floor means mid-run refills unless the dock holds clean and dirty water for you, and those tanks then need emptying, rinsing and occasionally descaling. Carpet handling matters just as much: some robots sense drag or listen to the pile, others use an ultrasonic sensor, and the answer that actually works on mixed flooring is a robot that lifts the pads or lets you leave them at the dock. Treat heated pad washing as cleaning rather than sanitising, because dock water is warm, not anywhere near boiling.

05

Obstacle avoidance, judged by how it fails

The cheapest robots detect obstacles by hitting them, using a sprung bumper plus infrared cliff sensors that stop them falling downstairs. That is fine for furniture legs and useless for cables, socks and whatever a pet has left on the floor. Above that sits a forward camera running object recognition, or a structured light projector that throws an infrared line ahead and reads its distortion to judge the shape and distance of what is there. Both approaches work in daylight. Neither is dependable in the dark unless the robot carries its own lamp, and a surprising number do not.

Judge this feature by how it fails rather than by how it demonstrates. Aggressive avoidance means the robot gives a wide berth to anything it cannot identify, so a cluttered room ends up with unswept islands ringing every unfamiliar object, and you have traded a tangled brushroll for dirt you can see. Waste detection is the one obstacle feature worth paying real money for if you have a dog that has accidents, since the alternative outcome is memorably bad. In a tidy home without pets, sixty seconds of picking things up beats any sensor on the market.

The privacy question is worth settling before purchase rather than after. Ask where images are processed, whether the app includes a remote-viewing mode and whether you can switch it off, and what the brand says happens to footage used to improve its recognition models. Then calibrate what the sensing can do. Thin dark cables on dark flooring, transparent and mirrored surfaces, and objects under roughly three centimetres tall are the common misses across every system regardless of price, and a flat sock can slip beneath the notice of a structured light projector because it never distorts the line.

06

The maintenance nobody puts on the box

Owning one of these is a standing commitment of about ten minutes a month, which no product page mentions. Filters clog and want tapping out weekly and replacing a few times a year. Hair wraps the brushroll and the side brush axle, and you will be cutting it free with scissors unless the brush has genuine tangle-cutting geometry. Cliff sensors gather dust and start refusing to cross dark rugs. Wheel bearings collect hair and squeak. None of this is difficult, but a neglected robot degrades quickly and people conclude the machine was poor when it was simply full.

Battery packs fade after a few hundred charge cycles, which for a daily user means noticeably shorter runs inside about two years. Before buying, check that batteries, brushes, filters and wheel modules are sold separately and that the brand has a history of keeping them in stock. Software is the other long-term risk. Mapping and scheduling usually depend on a phone app talking to a manufacturer's server, so a brand that folds or drops support can reduce an expensive mapping robot to an expensive bumper robot overnight.

Budget the consumables before you commit rather than after the first filter dies. Across three years a typical machine wants filters several times over, a replacement main brushroll and side brushes, mop pads if it mops, dock bags if it self-empties, and quite possibly a battery. Added up, a cheaper robot with widely stocked spares can cost less to own than a mid-priced one whose parts vanish from sale. It is also worth knowing where clogs happen: the short duct between the brush chamber and the bin, and the bin inlet itself, are where a sock stops the airflow while the fan carries on sounding perfectly healthy.

• Weekly: tap out the filter, clear hair from the main brushroll and side brush
• Monthly: wipe the cliff sensors and the charging contacts, check the front caster for wound hair
• Every few months: replace the filter, inspect the brushroll fins for wear
• Yearly: check whether the runtime has dropped enough to justify a new battery
• Before you buy: confirm the spare parts are actually listed for sale, not just promised

07

How to decide what tier you need

The right amount to spend is set by your floor plan more than by anything printed on the box. A single open floor of hard surfaces with no pets is the one situation where a basic random navigator honestly works. Several rooms, thresholds, rugs, or anyone in the house who sheds hair pushes you straight to lidar mapping and a rubber brushroll. Stairs mean you either carry the robot between levels or buy one that stores multiple floor maps. Very low furniture clearance is the strongest argument for a camera-based robot, since the turret is what usually stops a machine sliding under a bed frame.

Ranked by how much they change daily life, mapping navigation comes first by a wide margin, then a hair-resistant brush system, then obstacle avoidance if you have a pet or a cluttered home, then a self-emptying dock, and mopping last. Most of the money spent above the middle of the market goes into the dock rather than into the robot itself. That is worth paying if you want the machine to disappear into the background of the week. It is money wasted if you never particularly minded emptying a bin.

Think in capability tiers rather than in prices, because prices move and capability sets do not. The bottom tier bumps and drives and suits one open room. The next adds a saved map, named rooms and no-go zones, which is the step that turns a gadget into an appliance. Above that come hair-resistant brushes, obstacle recognition and a self-emptying base, all of which reduce how often you touch the thing. The top tier is mostly dock: pad washing, hot air drying, water refilling, occasionally a plumbed connection. Decide how often you will actually run it too, because a robot scheduled daily repays its price in a way a weekend robot never does.