Last week’s question was whether the work was being delivered. This week’s episode asks the harder one underneath it: “Could you even tell?” If you’ve watched or listened, you have the argument and the night I learned it. I won’t repeat them here.
This issue does two things the episode doesn’t. It takes on the objection I hear most often. And it gives you a test you can run on your own foundations this week without touching a single system.
The objection
The fair challenge goes like this: Data is never clean. If you wait for one perfect source of truth before you launch, you’ll never launch anything. Fix it in flight.
I agree with the first half. If the test were clean data, nothing would ever go live.
But the episode isn’t asking for clean data. It’s asking for an agreed number, and those are two different things.
Clean means right. Agreed means everyone in the room will act on it, flaws and all. You can launch on an imperfect number that everyone trusts, and improve it as you go. You can’t launch on two good numbers that disagree, because the first meeting after go-live stops being about what to do and becomes an argument about whose figure is real.
And it doesn’t have to be every number. Just the handful your launch will be judged on. If those don’t match, launching doesn’t save time. It schedules the argument for later, in public, when it’s more expensive to have.
“Fix it in flight” works for most things. It doesn’t work for the instrument you’re steering by.
The test: three numbers, three people
Pick the three numbers your next big decision will depend on. Revenue at risk. Active users. Cost to serve. Whatever your board will actually act on.
Ask three people, separately, for each number. Use the same wording, ask in writing, and ask where they got it.
Then compare.
Same number, same source. That’s a foundation.
Same number, different sources. That’s luck. It will come apart the first time something changes.
Different numbers. There’s no foundation yet, whatever the dashboard looks like.
Ask the same three people two more questions.
If this number moved the wrong way tomorrow, whose phone would you pick up? Three different names means nobody owns it.
And in your last review, how much of the hour went into reconciling figures instead of deciding? That’s what a missing foundation costs, every week, and it never shows up in a budget.
Then fix one number before the next meeting. Choose the source everyone will accept, not the best one. Agreement doesn’t make a number right. It makes it something you can fix together. Put one name against it, and write its definition in a single sentence nobody can read two ways.
The line to remember: “You can buy the technology. You cannot buy the trust.”
Watch or listen
The full episode is at the top of this post. It opens on the night before a go-live I stopped, because two of my own people pulled the same core number from two systems and it didn’t match.
You can also listen on Apple Podcasts or Spotify, or watch on YouTube.
If you run the test, reply and tell me how many of your three numbers matched.
Built to Absorb. One problem, one fix, every week.
