Robot vacuums: which specs predict a machine you keep using

The short version

The robot vacuums people still run two years later are the ones they can direct precisely and rarely have to touch. That is about navigation and the dock, not about suction. Our comparison schema agrees: navigation carries the top importance weight, with suction, mopping and a self-emptying dock a step below, then obstacle avoidance, the mop-washing dock and runtime.

Navigation is first because the map is the product: a robot that builds a floor plan can be told to clean the kitchen and kept out of the dog's corner. The dock is second because it decides how often the machine is a chore. Suction, the number on the front of every box, is the least reliable figure here.

What actually matters

Navigation comes in four grades and they are not close. Random-bounce robots have no idea where they are: they cover a room by statistics and cannot be directed. Gyroscope navigation drives in rows but drifts over a larger floor. Camera mapping builds a usable plan but wants light. Lidar maps in complete darkness and is what makes room labels, no-go zones and virtual walls dependable. The jump from gyro to lidar changes what the machine is for, and it is worth more than any amount of extra suction.

Mopping is a spectrum rather than a feature. A flat damp pad dragged behind the robot is a dust mop: fine for daily dust on sealed hard floors, useless on anything dried on. Rotating pads pressed down with force are the first tier that lifts something off the floor. Two details decide whether it works: does the robot lift or detach the pad on carpet, and can it re-wet the pad from the dock mid-run rather than carrying a tank that runs dry?

The docks are where the money goes. A self-emptying dock is the highest-value convenience feature for anyone with pets or long hair; the costs are footprint, a loud few seconds, and a recurring bag. A mop-washing dock rinses the pad between passes and, on better ones, dries it. Without one, a wet dirty pad sits on the robot until you hand-wash it, and the smell within a day is the most common complaint in the category.

Obstacle avoidance is a permission rather than a capability: structured-light and camera systems catch socks, cables and shoes reasonably well, none are perfect, and what the feature buys is the confidence to run the robot when you are out. Runtime matters less than it did, because most mapping robots recharge and resume; the question is whether the model resumes at all.

The suction number, and the rest of the noise

Pascals measure pressure, usually at a sealed intake, which is not the same thing as cleaning. Picking dirt off a floor depends on airflow through the head, the brush agitating the fibres, how well the housing seals, and how many passes the robot makes. A pressure figure captures one input to one of those, and there is no mandated common test behind these numbers, so two manufacturers' figures are not comparable even when both are honest.

The trend tells you what the number is for. Headline pascal figures have climbed by roughly an order of magnitude over the last several product generations without a comparable change in what owners report about deep carpet. That is the signature of a marketing axis, and our configuration treats it as one: suction is the only attribute in this category held to a raised confidence floor, so we leave it blank rather than show a figure we cannot corroborate.

Look at the brush instead. Rubber or silicone brushes handle hair far better than bristles, dual counter-rotating brushes better again, and a tangle-cutting comb matters if anyone has long hair. On carpet, set expectations honestly: a robot keeps a floor from getting bad, and on medium or deep pile nothing in this class replaces an upright.

• AI branding of every kind, including named in-house chips, which says nothing verifiable about behavior around a charging cable.
• Five-in-one and seven-in-one counts, which are counting features that already have names.
• Runtime quoted in the quietest mode on bare hard floor, which is not the mode you will use.
• Sixty-day dock capacity claims, which assume a specific home size and a specific amount of dirt.
• Mopping pressure in newtons or grams with no pad area attached.

The things that decide it and are not in the specs

Five checks that predict satisfaction better than any published attribute.

• Height. A raised lidar turret makes a robot taller than the flattest designs, which decides whether it fits under the sofa, the bed and the toe kick. Measure your lowest clearance against the published height.
• Threshold climbing. Published and easy to skip. A raised bathroom threshold or a stiff rug edge can mean a room simply never gets cleaned.
• Consumables. Bags, filters, side brushes, the main brush and mop pads all wear out, are brand-locked, and over three years add up to a real fraction of the machine. Look up the replacement kit first.
• Software longevity. This category turns models over fast. Ask what the robot still does if the app is retired, and whether the maker supports discontinued models with parts.
• Privacy, if it has a camera. Camera-based avoidance means an internet-connected camera driving around your home. Ask whether recognition runs on the device, whether imagery ever leaves, and how you delete the maps. Those answers are in the privacy policy, not the spec sheet.

Tradeoffs by segment

The phrase robot vacuum covers machines that are barely the same product. Within one brand and one year, our seed catalogue records a basic model launching under three hundred dollars and a flagship with a washing dock at fourteen hundred. That spread is not five times the cleaning; it is navigation, docks and mopping.

At the entry level you get gyro navigation, a small bin and a charging dock, which is reasonable for a small apartment with hard floors. The several-hundred-dollar band is where lidar mapping and a self-emptying dock arrive, and it is where most households should aim, because those two changes decide whether the robot stays in use. Past a thousand dollars at launch you buy the washing dock, rotating mops with real pressure, better avoidance and extending brushes for edges: worth it on hard floors with pets, much less so in a carpeted home.

When this category actually discounts

The model cycle dominates, and it is faster here than in any category we cover. Flagships are announced at the January trade shows and ship in spring, with a second wave in the fall, and the cadence in our seed catalogue is close to annual: one brand launched flagships in 2023, 2024 and twice in 2025, and a second did the same three years running. The outgoing flagship is usually the best value, because a one-year-old lidar robot with a washing dock is not meaningfully worse at cleaning.

Launch list prices here are also falling generation over generation, which is unusual. In our seed data, one brand's 2024 flagship launched at a higher list price than its 2025 successor, and a second brand's did the same. That is the opposite of how phones and headphones behave, and it makes an older model's original list price a bad reference point for today's offer.

On the calendar, the July events, Amazon's and the competing Walmart, Target and Best Buy events, are strong because much of this category sells through Amazon anyway. Black Friday and Cyber Week, whose window our calendar opens about three weeks early, are the deepest. Post-holiday clearance picks up superseded stock just before the new flagships are announced. Several of these brands also sell directly on their own promotional schedule.

What a real discount looks like here

This is the category with the widest gap between list and street price, and the most creative discount presentation: a struck-through list price with a clip-this-coupon control underneath, so the page shows two reductions stacked on each other, neither anchored to anything you can verify.

Our test is the same one we apply everywhere: an advertised was-price only counts if it was the prevailing price for at least thirty percent of the previous ninety days in our observations, or we call it the retailer's list price and score the offer's authenticity down. Here that test fails often. By hand, make two comparisons: the model against its own recent street price, and against the current generation's. Given how quickly list prices have come down, a superseded flagship at a large notional discount is sometimes still the more expensive machine.

Questions to ask before you buy

Two are measurements; the rest are five-minute checks.

• What is my lowest furniture clearance, and is this robot shorter than it?
• What is the tallest threshold it must cross, and what is the published climbing height?
• Hard floors or carpet? Mopping features and their docks are near-irrelevant on carpet.
• Does it lift or detach the mop pad on carpet, or drag a wet pad across it?
• Does the dock wash and dry the pad, and if not, will I really wash it after every run?
• What does a year of consumables cost for this model, and are they easy to buy?
• If it has a camera: where does recognition run, and how do I delete the maps?
• Is a successor due soon, and is the discount bigger than the gap to the current generation?

How we would check this

There are thirty robot vacuums in this catalogue, from basic gyro machines to flagships with washing docks, and price tracking has started. We do not yet have verified offers to publish, and the pages say so.

The bar is the same as everywhere here: at least thirty observations across at least fourteen days before a verdict stops being an early read, at least two independent sources or one licensed API feed, and an authenticity score that clears the former-price test above. Suction gets extra scrutiny because our configuration holds it to a higher confidence threshold, so a figure we cannot corroborate is left blank. The weights and substantiation rules are on our methodology page.

Questions shoppers ask