Written by Ralph Sun
Robotics marketing has been built on spectacle, and the audience has learned to discount it. The vendors who win the next phase will be the ones willing to publish numbers that could embarrass them.
Marketing in robotics has an unusual problem: the product is so visually compelling that demonstration has substituted for evidence. A humanoid that walks and gestures generates more attention than any performance metric ever will. That has been good for awareness and bad for trust.
Look at how the sector communicates. A Chinese humanoid recently tested traffic support and tourist guidance on rainy streets in Hangzhou without publishing measured performance results. A European industrial trial had a humanoid perform autonomous tote-handling with reported throughput — closer to the right instinct, but a single figure rather than a distribution. And the most-watched humanoid moment of the month involved a leg covering being cut open on stage to prove the machine was not a person in a suit.
That last example is revealing, because it answered a question nobody in procurement was asking. No operations director evaluating a deployment cares whether the robot is convincingly human. They care whether it still works in month nine, how often a person has to rescue it, and what that supervision costs per shift.
There is a useful parallel in an article the industry should read more widely, on evaluating humanoid robots on evidence that matters. The argument is straightforward: capability claims are cheap, operating evidence is not, and the gap between them is where deployments fail.
I want to be honest about why vendors avoid that evidence. Publishing an intervention rate is a competitive risk. If your number is worse than a rival’s, you have handed the market a comparison. If your number is good, you have created a bar you must clear every quarter. Silence is safer.
Silence is also a slow leak. When every vendor declines to publish, buyers discount all claims equally, and the entire category gets priced as unproven. That is the trap the sector is in now: the credible and the incredible are treated the same, so credibility earns nothing.
At Brainlayer we think the fix is to publish the uncomfortable metrics, because they are the ones that demonstrate a real deployment. Intervention rate per thousand attempts. Handoff latency. Recovery success across different operators and shifts. Downtime attributable to task novelty rather than hardware. These numbers are unflattering in isolation and persuasive in aggregate, because they show a system being operated rather than performed.
There is a commercial logic here, not just an ethical one. Buyers who are handed operating data move faster, because they can build a business case. Buyers handed a video stall. A vendor that publishes hard numbers is effectively doing the buyer’s diligence for them, which is the most efficient marketing in a technical market.
So my prediction is simple. The next phase of this industry will not be won by the most spectacular demonstration. It will be won by whichever vendor first makes an intervention rate a headline figure — and discovers that the audience, having learned to discount everything, rewards the one claim that could have gone badly.
