Robot reliability numbers: the vendor claims and the one independent study

The two kinds of numbers measure two different things. A vendor MTBF figure describes a single robot arm running under controlled conditions. The 88 percent reliability figure describes a whole robot cell in a real factory, including its tooling, fixtures, sensors, and the parts flowing through it.

That distinction is the whole story. The study found that 80 percent of the time a cell stops, the cause is not the robot. So a robot can be genuinely reliable as a component and still sit inside a cell that is down 12 percent of the time. Vendors quote the component number; the plant lives with the cell number.

The vendor figures also have no published methodology behind them. The source lists advertised MTBF values of 40,000, 60,000, 80,000, and even 100,000 hours, and notes that some claims now imply a robot will run 43 years before it fails. None of those are tied to a dataset anyone can inspect.

For a realistic figure on how often a robot cell is actually running, cite the 88 percent reliability and 87 minute mean time to failure from the study of more than 400 factories, and name the source.

Treat any MTBF figure in the tens of thousands of hours as a vendor claim, not a measurement. If you must use one, attribute it to the manufacturer and make clear it describes the arm, not the cell.

When you talk about uptime, say whether you mean the robot or the cell. The gap between them is most of the argument, and mixing them is how a 99 percent robot ends up inside an 88 percent cell.

Quoting a vendor MTBF as if it were measured uptime. An 80,000 hour MTBF is a controlled-condition claim about the arm, not evidence that a cell runs 80,000 hours between stops.

Reading 88 percent reliability as the robot being broken 12 percent of the time. Most of that downtime is not the robot. It is tooling, parts, and the rest of the cell.

Assuming there is a large body of independent robot-reliability data. There is very little. The figures here come from effectively one published study, which is exactly why the vendor numbers go unchallenged.

Every figure on this page is drawn from a single fetchable primary source: The Robot Report's 2019 article, which reports reliability research published in the International Journal of Performability Engineering. We fetched the article and confirmed each figure appears in its text before publishing.

We went looking for a broader base of independent robot-reliability data and did not find one. The underlying journal article sits behind a paywall, and the vendor MTBF and uptime figures trace only to spec sheets and marketing pages, none of which publish a method. Where a number could not be tied to a free primary source, we labelled it a vendor claim rather than a measurement.

The Citation Confidence rating reflects how safely a figure can be quoted today. The measured study figures rate High. The advertised MTBF and the 43-year claim rate Low, because they are marketing figures with no verifiable basis, included here only to show what the study was responding to.