I now tell clients the first live test will probably fail
Before a build of ours went through its first proper run with real people using it this week, I said out loud that I expected it to fail. Then I listed the things I thought would go wrong, in order, with a name against each one. I used to do the opposite. I used to say it should be fine, because that is what you want to hear from the person you are paying, and it costs me more than it buys.
Why the sentence matters
A failure nobody predicted turns into a conversation about competence. Whose fault, why was this not caught, what else is broken, should we have gone ahead at all. You spend the next week defending the project instead of fixing it, and the client spends that week quietly deciding whether they trust you.
A failure you predicted out loud, a week before, turns into a list. Here is the thing we said would break, it broke, here is who is closing it out and by when. The same awkward morning, and nobody has to work out what it means about anyone. That is the whole trick, and it is free.
How I say it now
In writing, before the date, not in the meeting on the day. Vague is useless here, so I name the specific things I think will go wrong rather than warning about teething issues. Something like: the first few entries will not match what the finance system expects, somebody will use it in a way we have never seen, and one of the connections between the two systems will time out once more than a couple of people are on it.
I also say what a pass looks like, because without that a test has no end. One whole piece of work, start to finish, with the client watching, and the result showing up correctly in the place they already check. Not a tour of the screens. One real job, all the way through.
What it does to the room
The person who has to sign it off relaxes slightly, which sounds minor and is not. They stop watching to see whether you will be honest with them, and start watching the software instead. They ask better questions too, because a list of expected failures tells them where the soft parts are, and they usually know things about their own business that make the list sharper.
There is a version of this that goes too far, where everything is caveated and nobody can tell what you actually believe. I keep it to one honest paragraph, with dates and names, and then get on with the work. Hope for the best, prepare for the worst, and say which one you are preparing for.
The first run of anything real is where you find out what people actually do, as opposed to what they told you they do. Expecting that to be comfortable was the mistake.
Got a version of this problem? Twenty minutes minimum, seven questions, and a one-page Bottleneck Blueprint within 48 hours. If it's not a fit, we'll say so.
Book a call → Not ready to talk? Take the eight-question AI assessment → Questions about cost, data and ownership? Thirty-five straight answers →