Exactly the one you've been using: W(out)/W(in).
That would help, but not totally solve the problem. He measures W(in) using a multimeter, ignoring the
power factor.
He might hit a harmonic at 500Hz and has no chance to truly measure it using his equipment. So reported W(in) might be bogus, the system boundary is very unfortunate/irrelevant and the current source is unrelated to the real world application.
I understand startups are looking for a fortunate testcase for promotion. I even think giving out samples for reviews is ok if the review includes that information. But this efficiency claim is way too far off. He might have a case promoting the form factor (it looks tiny) or the support-local-guys route. IP69 might be a seller, if the resin can really resist a high pressure washer and he does something about the USB port (IP67 USB ports exist. IP67 only while closed, of course...).
You could of course implement the
minimal loader design and put it into a fancy case. I'm pretty sure it would be the most ecofriendly system around from the raw materials point of view. And very, very cheap to produce.