Suction, filtration and run time: which numbers actually matter

Pascals and air watts are not the same measurement

A pascal figure describes sealed suction: the pressure difference a motor can hold when the inlet is blocked. An air watt figure describes the power carried by moving air: pressure multiplied by flow. A machine can hold a high pressure against a blocked inlet and still move very little air once the inlet is open, which is why the two figures do not convert into one another.

What picks dust off a floor is airflow at the head, and what pulls grit out of carpet is a driven brush roll agitating the fibres while that airflow is applied. Neither of those is the headline pascal number. This is the practical reason a quiet corded upright can outperform a robot quoting several thousand pascals on the same plush carpet.

Robot makers almost always quote pascals; corded makers more often quote air watts or nothing at all. When two machines are quoted in different units there is no honest comparison to make from the sheet alone, and the useful substitute is the mechanism: sealed path, head design, and whether the brush is driven or dragged.

HEPA is a filter medium claim, not a machine claim

HEPA stands for high efficiency particulate air, and the Environmental Protection Agency describes a filter of this type as able to remove at least 99.97 percent of airborne particles of 0.3 microns, with particles both larger and smaller than that trapped at even higher efficiency. That is a statement about a filter medium tested on its own.

A vacuum is only as good as the path around it. If the body seals leak, air carrying fine dust takes the easier route past the cartridge and leaves through the exhaust, and the 99.97 percent figure says nothing about what left the machine. The claim worth looking for on the manufacturer's page is a sealed system, meaning all exhaust air is routed through the filter rather than around it.

The EPA does not certify, register or list acceptable air cleaners or their manufacturers, so there is no federal mark to look for on a vacuum. What you can check yourself is the filter's replacement interval and whether cartridges are sold singly. A sealed machine on a 2 month filter schedule that nobody keeps is an unsealed machine by month 6.

1. Find the phrase describing the air path, not just the filter: sealed system, or all exhaust through the filter
2. Find the stated filter replacement interval and multiply the cartridge price by the number of changes in a year
3. Check whether the filter is washable, and if so how many hours of drying the manufacturer asks for before refitting
4. Confirm the pre-motor and post-motor filters are both listed as parts, because machines often have two
5. Treat any filtration percentage without a particle size next to it as unreadable

Run time belongs to a power mode

Every battery run time is quoted in a specific mode with a specific head fitted. A stick quoted at 60 minutes is usually quoting its lowest setting with the plain floor tool; the motorised carpet head draws its own current, and boost mode drains the same pack far faster. If most of your floor is carpet, the figure that matters is the one for the motorised head at the setting you will actually use.

Robots are quoted the same way, and there the number interacts with recharge time. A robot with a 90 minute stated run time that needs 180 minutes on the dock before it can resume has a duty cycle of roughly one third, so a level that takes 150 minutes of machine time is finished about 5.5 hours after it starts rather than in 2.5 hours.

Capacity also fades. A lithium-ion pack delivers fewer usable minutes after a few hundred cycles than it did new, so a machine bought with just enough run time for the job has no margin by year 2. Buying a stated run time comfortably longer than the measured job is the cheapest insurance in this category.

The numbers that are not on the sheet

Bin capacity decides how often you interrupt the job. A small bin on a shedding-pet household is emptied mid-run, and that converts an unattended machine into an attended one. The figure is published in millilitres or litres and is worth reading next to the floor area you intend to cover.

Noise is published in decibels and is the figure most likely to decide whether a robot runs on a schedule or sits off. A machine loud enough that you stop it when you are home is a machine running at a fraction of its intended duty.

Last is the brush roll geometry. Hair wrapping a roll is the most common performance failure in a pet household, and tangle-resistant rolls and easy-release end caps fix more of that than extra suction ever does. All three of these are findable in the manual before purchase, and none of them appears in the headline specification.

There is also no efficiency label to fall back on. ENERGY STAR certifies dishwashers, clothes washers, refrigerators and electric cooking products, and it has no category for vacuums or robot vacuums at all, so nothing on the box has been verified against a published performance protocol by a third party. That makes the mechanism — sealed path, driven brush, filter interval, bin size — the only evidence available before the machine is on your floor, and all four of them are printed in documents the manufacturer already publishes.

Before you buy

• Note which unit each machine is quoted in, and refuse to compare pascals against air watts
• Find an explicit statement about the sealed air path, not only a filter percentage
• Look up the filter replacement interval and price a year of cartridges
• Find the run time for the mode and head you will actually use, not the headline figure
• Compare the stated recharge time with the run time to get the real duty cycle
• Check the bin capacity against the floor area you want covered in one pass
• Check that the brush roll releases for hair removal without tools

Questions readers ask

Is a higher pascal number always better?

Only against another machine quoted in pascals, measured the same way, with a comparable head. Across types it predicts very little, because pressure against a blocked inlet is not airflow at an open one. On carpet, agitation from a driven brush roll contributes more to pickup than either figure does.

Do I need a HEPA filter if nobody in the house has allergies?

The filter matters less than the seal in that case, but the two travel together: machines that route all exhaust through the filter are also the ones that put fine dust where you want it rather than back into the room. If you buy one on a 2 month replacement schedule, budget the cartridges, because an unchanged filter undoes the specification.

Why does my machine run for less time than the box said?

Because the box figure belongs to the lowest power mode with the plain head fitted. The motorised head has its own motor, boost mode multiplies the draw, and the pack delivers fewer minutes after a few hundred cycles. Read the manual's per-mode table rather than the headline number.

Can I tell whether a machine is sealed from the specification sheet?

Sometimes. Look for an explicit statement that the whole air path is sealed or that all exhaust passes through the filter. A page that quotes a filtration percentage but says nothing about the path has told you about the cartridge only, and that is the gap where fine dust leaves a machine.

How often should filters actually be changed?

On the interval the manufacturer publishes for that model, which is commonly stated in months, and sooner in a house with a shedding pet. The cost is small and the effect is not: a loaded filter reduces airflow, and airflow is the part of the specification that was doing the cleaning.

Sources

1. U.S. Environmental Protection Agency — What is a HEPA filter? (99.97 percent at 0.3 microns; EPA does not certify air cleaners) — checked 2026-11-03
2. U.S. Consumer Product Safety Commission — voluntary standards for batteries, including UL 2595 for battery-powered appliances — checked 2026-11-03
3. ENERGY STAR — certified product categories (no vacuum category, so no efficiency label to compare) — checked 2026-11-03