The Ladder Check
How far up the ladder do your answers still agree?
Six questions, about three minutes, answerable by one person from memory. No meeting, no data pull, nobody else told. It ends in a named result rather than a score, and one of the results is that you are fine.
What this is
Every question below can be answered by one person, alone, in roughly three minutes. That is deliberate. The honest version of this test is to hand out index cards at a leadership meeting and have everyone write down the same number, but that takes convening power, it is politically visible, and it can embarrass people. Most people will never do it.
You do not need to run the test. You need to admit what the test would show. You already know whether your CFO and your head of sales would produce the same number, and admitting it takes ninety seconds.
There is no score out of 100, because a score invites gaming and comparison. There is no email gate in front of the result, because gating a verdict makes it a lead-capture form and you would be able to tell. And no service is named until after the result, because a test whose questions are reverse-engineered so that only one product passes is not a test.
The six questions
The four results
Four outcomes, ordered by what they cost to fix. They are all written out below whether you answer the questions or not, because you should be able to read the whole instrument before deciding to use it.
You are in better shape than most
Answers agree, retrieval is self-serve, you reach rung 4 or higher, and disputes end at a source.
You are not the bottleneck on this company's AI work, and the honest advice is to go spend the money on the AI itself.
That is the whole result. There is no service on the other side of it, because you do not need one. A self-assessment that never returns this outcome is not a self-assessment.
Retrieval gap: the numbers are fine, the access is not
The count and the definition hold. Getting to the number still needs a human in the loop.
The organization agrees on the answer; it just cannot reach it without going through a person. That is the cheapest of the three failure modes to fix, and it is often weeks rather than months.
It is also the outcome most likely to be misdiagnosed as a data problem and sold to you as one. Before anybody proposes a cleanup, make them show you which records are actually wrong. If the answer is none, you have a plumbing job, not an archeology job.
Definitional drift: you agree on the records, not on what counts
Everyone is looking at the same rows and counting them differently.
This usually happens because a definition moved once and only some of the systems were told. Nothing is corrupt. Every number in the building is correct according to the rule the system that produced it is following, and the rules stopped matching.
It is expensive in a particular way: it does not look like a failure. It looks like two competent teams disagreeing, which reads as a personality problem right up until someone traces both numbers back to their definitions.
Published case: The Unsexy Data Work That Actually Matters.
Identity drift: you do not agree on who the customers are
Two systems hold the same company twice and nothing reconciles them, so every number downstream inherits the ambiguity.
This is the most expensive of the three, and it is the one that quietly voids software you have already paid for. A campaign tool, a CDP or an AI agent pointed at a customer list that is an opinion will produce confident output about people who are not who the system thinks they are.
It also compounds. Every month it goes unfixed is another month of records written against the wrong identity, so the cost of the fix grows while the license keeps being paid.
Published case: How Many Customers Do You Have?.
One more thing, from question 6
You said the last disagreement was settled by seniority, or that it went away on its own. That is worth separating out, because it is not a data problem and no amount of cleanup will touch it. When the senior number wins, the organization has learned that checking is not what settles things, and it will learn that again next quarter. Fixing the records is necessary here and it is not sufficient.
What to do with the result
Whichever one you landed on, the useful next step is 15 minutes to check whether it is actually the one you are in. I have been wrong about which it is often enough to want to ask.
That is the whole ask. No document, no deck, no proposal attached. Email me and say which result you got.
Where this comes from
Both axes of this assessment come out of published work rather than invented positioning. The ladder is from How Many Customers Do You Have?, and the identity-drift case is the engagement described in The Unsexy Data Work That Actually Matters.
One difference between the two is worth naming. In the article the ladder stays deliberately open, because an essay that terminates in a verdict gives a reader false comfort. Here it is a closed set of six, because an instrument that does not terminate is not an instrument. Both endings are right for their own job.