Do home robots ship the hardware they promise?
A backed snow robot swapped the module its buyers chose, then slipped delivery to the edge of winter. The hardware a home robot ships is rarely the part a shopper checks.
Someone who backed Yarbo’s modular M-series yard robot for snow clearing spent the summer of 2026 watching the machine change under them. On July 1, a campaign update said the Plow Blade Module they had pledged toward was gone, swapped at no extra cost for a Two-Stage Snow Blower Module, and that the snow bundles would now ship in mid-November rather than August.
Yarbo’s blower is a real machine on the company’s other line, a two-stage head that throws snow up to 40 feet across a 24-inch path. Bolting it to the M-series Core in place of a blade was never a small change.
The heavier module alters how the unit mounts, draws power and balances, so the company re-ran the Core’s structural design, validation and field testing. The delay comes out of that rework.
Framed as an upgrade or not, it is a swap the backer did not choose, on a date with no slack. A snow tool that lands in mid-November has no margin before the weather it exists for.
But it is not the failure most buyers brace for. The worry you carry into a home robot is whether it is clever enough: whether it will find the lawn, dodge the dog, or clean the whole floor. So which failure actually shows up?
Set against the launches tracked here, that worry points the wrong way. Of 1,362 dated entries, 173 record a setback: a regression, recall, withdrawal or removed capability. Just 12 of them turn on autonomy. Thirty-one are hardware. Eighty, the largest share by far, are commercial, a model discontinued, pulled from sale, or slipped the way Yarbo’s was.
When the number on the box is not the number on the floor
Robot vacuums are the most-tracked class on this record and the most setback-prone, with 98 of the 173 attached to them. The MOVA V70 Ultra Complete is a recent flagship, sold near the ceiling of what any robot vacuum advertises, around 40,000 Pa of suction. Independent lab testing could not reproduce it.
Figure
Where a suction figure is measured
- Sealed portNo brush, no floor, nothing in the way. Pressure reads highest here, at the one point on the airpath where no cleaning happens.
- DuctAfter the bends and the filter. Already well short of the quoted number, with the floor still to come.
- Nozzle, on carpetIn contact with the surface being cleaned. This is the pressure that actually lifts grit, and no specification quotes it.
A Pa figure is read inside the airway, before the brush, the bin and the carpet each take their cut. That is how a headline number and a middling clean can both be true at once.
In measured airflow the V70 came in well below its billing, and a hands-on review found it struggling for consistency across floor types at a flagship price. If you are sorting one flagship from another on Pa alone, the figure that separates them is the one least tied to what reaches the bin.
The same distance shows up where the capability is physical rather than numeric. MOVA sells the Rover X10 pool robot on a seven-in-one system whose feature list includes steps. A reviewer found it cleaned the floor well, then drove up to the base of the steps and stopped, in every mode the reviewer tried.
A gallant attempt but all it did was drive up to the base of the steps, not up and over.
What a buyer can actually plan for
Read the three together and a shape repeats. A module changes before it ships. A suction number does not survive a test bench. A climbing feature reaches the first step and quits. None of these is the robot failing to think.
Three machines, what was sold and what the record shows
| Machine | Sold as | What the record shows |
|---|---|---|
| MOVA V70 Ultra Complete | Suction near 40,000 Pa | Independent lab testing measured well below the claim |
| MOVA Rover X10 | Climbs pool steps in a seven-in-one mode | Reached the base of the steps, did not climb in any mode tested |
| Yarbo M-series snow module | A plow blade, shipping August 2026 | Blade swapped for a two-stage blower, snow bundles slip to mid-November 2026 |
The column that stays empty across all three is autonomy. What moved was the hardware a buyer was promised, the number behind it, or the schedule it was promised on. That is where the record says these machines actually break.
The commercial share is the one a spec sheet can never print. Eighty of the setbacks are a company’s decision: a model discontinued, a pool robot replaced by its successor, a mower marked end-of-line while parts support quietly continues. The robot still works. What fails is the business and the support around it, and no amount of on-paper capability guards against that.
The scarce thing to check is not whether a robot can think, but whether the body it ships with matches the one on the page, and whether its maker is still behind it a year on. A spec sheet answers the first and says almost nothing about the other two.
The backer who pledged for snow clearing did not end up with a robot that cannot find the driveway. They got a different module than the one they chose, arriving in mid-November, at the very front of the season it was built to handle.
