The short version
Somebody names what it would take to convince them. You go and get it. Then it turns out that isn’t quite what they meant, and the real requirement is something else — slightly further away, slightly harder, and conveniently just past whatever you’ve now produced.
Nobody has to admit anything for this to work. The old standard is simply never mentioned again.
What it looks like
Show me one customer who actually asked for this.
Sure, but none of those are enterprise customers.
If the benchmark shows a real difference, I’ll drop it.
Benchmarks never reflect production load anyway.
The second one is worth a second look, because notice what happened: the objection that finally arrives is one that would have applied just as well before the benchmark was run. It was available the whole time. It only got deployed once the first standard failed to do its job.
Why it doesn’t work
Because a standard you can revise after seeing the result isn’t a standard. It’s a preference with paperwork.
The honest version of “what would change my mind” is a genuine commitment: you name a condition, and if the condition is met, you move. If the condition can be replaced the moment it’s satisfied, then nothing could ever have met it, and the original question — will any evidence do here? — has quietly been answered no.
That’s the real cost, and it’s mostly borne by the person doing it. Everyone else stops bringing them evidence, because everybody has worked out that evidence isn’t the operative variable. The conversation looks like it’s about data and isn’t.
This is not moving the goalposts
- Learning something from the answer. If your evidence genuinely reveals that the first question was the wrong one, updating the question is what’s supposed to happen. The difference is that an honest update says so: “I asked the wrong thing — that’s on me, here’s the better question.”
- A standard that was always compound. If somebody said at the start “I need a customer, and it has to be enterprise,” meeting half of it isn’t meeting it. Annoying, but fair.
- Multiple independent objections. Some things really do have several separate problems. Answering one doesn’t clear the rest, and being told so isn’t a trick.
- A different person raising a different bar. Two people in a room are allowed to want two different things. That’s a meeting, not a fallacy.
The line isn’t admission. Saying it out loud doesn’t repair it — “fair enough, and I still want more” is the same move with a covering note, and plenty of goalpost-moving is exactly that open: each objection answered, the next one raised, nothing hidden and nothing conceded.
Two other things have to hold. Whatever the met standard actually established has to be conceded — said out loud, with the position moved by that much. And the new bar needs a reason that isn’t “the old one was met”: one that would have applied whatever the evidence had shown, and could have been stated before it came in. If the only thing making the new objection worth raising is that the previous one got answered, it isn’t a standard. It’s a patch.
How to call it out
The real fix is upstream — nail the target down before you go get anything:
- “Before I spend two days on this: if I bring you four enterprise customers, are you in? Or is there something after that?”
- “What’s the full list? I’d rather hear all of it now than one at a time.”
And once it’s already happened, name the sequence rather than the fallacy:
- “That’s the third thing. The first was one customer, then enterprise, now paying. What’s the actual bar?”
- “Is there any answer to this that would change your mind? Genuine question — if there isn’t, that’s fine, but let’s stop pretending it’s about the data.”
That last one ends a lot of conversations, and it ends them accurately. Some positions aren’t held for evidentiary reasons, and it’s better for everyone to find that out on purpose than after two more weeks of research.