Somebody in a leadership meeting says expansion doesn't work here.
Nobody argues. There's evidence. They tried it eighteen months ago, it went nowhere, and the number has been a rounding error ever since. The meeting moves on, because what's there to discuss? You ran the experiment and you got the result.
Look at what the experiment actually was.
Somebody got handed a number under a deadline. The only playbook on the shelf was sales pressure, so they used it, on customers picked by the renewal calendar because no other list existed. Nobody owned the outcome past the quarter. Nothing triggered anything. No forecast contained it.
It didn't work. Correct. And it tells you exactly nothing about whether expansion works, because that's not what got tested.
What got tested
Whether sales pressure works on existing customers.
It doesn't, and there are good reasons why, and none of them are what that leadership team concluded.
This is a diet you went on while eating pizza. You didn't lose weight. That's true, it's measured, and you have the scale readings to prove it. What you don't have is any information about whether diets work, because you didn't do one.
Nobody would accept that reasoning about their own body. Everybody accepts it about their revenue.
Then the verdict does work
And this is the part that makes it permanent rather than merely wrong.
The conclusion is useful. Once expansion doesn't work here, you don't need to instrument readiness. You don't need to hold anything back out of the initial sale. You don't need to name an owner or defend a number in a forecast review. A whole category of uncomfortable work becomes unnecessary, on the strength of evidence everybody saw.
And the belief protects itself from the only thing that could disprove it. A real test needs four conditions the first attempt didn't have: delivery that's actually landing, a defined readiness condition, something held back to sell, and an owner who isn't measured only on the quarter. None of those get built, because why would you build them for something that doesn't work here.
So the experiment that would overturn the verdict is the one the verdict prevents.
The version with no incident
At least the failed campaign leaves a body. There's a date, a spreadsheet, somebody who ran it, and a result everybody can point at. You can argue with it.
The harder version never had an incident at all.
Somewhere early on, a premise got adopted. We're a subscription business, not a services business. Our customers buy once and renew. Growth comes from the pipeline. Nobody announced it, nobody voted on it, and it wasn't wrong when it was adopted.
Then it quietly set policy on everything downstream, and it removed the reason to test whether the other thing would work. There's no meeting to go back to. No decision to revisit. No bad quarter that traces to it. Just a premise doing structural work for years while presenting as common sense.
That one is harder to catch precisely because nothing broke.
What would have to be true for you to find out you were wrong
That's the question, and it's worth asking out loud in the meeting where somebody says it doesn't work here.
If the answer is that nothing would surface it, you don't have a conclusion. You have a loop that consumed its own evidence.
The way out isn't argument. Arguing with a self-fulfilling prophecy just gets absorbed, because it already ate the evidence once. The way out is a measurement the loop doesn't control, taken before anybody needs it: go count what your existing customers are ready to buy and haven't been asked for. Not a campaign. A count.
Run that and one of two things happens. The number comes back small, in which case you were right and now you know it for a reason instead of by assumption.
Or it comes back large, and you find out you've been on a diet this whole time, holding a slice.
Lincoln Murphy formally named and popularized Customer Success starting in 2010 and has spent 15 years connecting it to expansion revenue and commercial outcomes. Read The Premise.