The tradeoff everyone assumes
The intuition is reasonable: a motor pulling harder should empty the battery sooner, so the strongest vacuums ought to run the shortest. Spec sheets lead with a single giant Pa number, which makes suction feel like the axis you trade everything else against.
Plot the actual catalog and the line goes the other way. Suction and runtime climb together (r = 0.72, a strong positive relationship). The vacuums with the biggest suction numbers tend to have the longest runtimes, not the shortest.
Why the numbers move together
Both figures track one hidden variable: whether the machine is a flagship. Top-tier models ship larger battery packs, so they post both a high peak-suction spec and a long runtime. Peak suction is also a burst mode the vacuum rarely holds for a whole clean, so it costs far less runtime than the headline implies. Ecovacs Deebot X8 Pro Omni Robot Vacuum and Mop pairs the longest runtime in the set (291 minutes) with near-top suction, which is exactly the point a real tradeoff would forbid.
The Pa number is not what it looks like
Stated suction runs from 1,000 to 18,500 Pa. Nobody cleans 19 times better with the top unit. Pa is measured at the inlet under ideal conditions and inflates freely because no rule forces a shared method. 3 of the 19 vacuums here, iRobot's whole lineup among them, decline to publish a Pa figure at all and compete on brush design and navigation instead. A spec that a big brand can simply opt out of is not a spec you can rank the market by.
How to actually read this
Stop treating suction as the number you sacrifice runtime for. Within the flagship tier the Pa gaps are mostly marketing; pick on the things that vary independently and that you feel every day: mopping quality, whether the base self-empties, navigation, and how the app behaves. Use the full field below to compare those directly.
Methodology
Figures are manufacturer-stated max suction (Pa) and max battery runtime (minutes) for the 19 robot vacuums in the Slopsheets robot vacuum sheet as of 2026-07-22. The correlation uses the 16 models that publish both figures. We store no scraped or live-priced data; this is a spec-sheet aggregation, refreshed as the catalog updates. Where a maker publishes no value, the cell reads n/a rather than a guess.