You ask how many active customers you have. Sales gives you one number. Operations gives you another. Finance gives you a third. None of them got it wrong.
That's the first thing to understand, because it changes entirely what you do about it. If someone had made a mistake, you could correct it. But all three counted correctly: what differs is what counts as an active customer. For sales it's anyone under contract. For operations, anyone who consumed service this month. For finance, anyone who was invoiced. Three reasonable definitions nobody ever wrote down, living inside the same company and producing three simultaneous truths.
The conversation that follows is usually about discipline: align the departments, standardize the report, insist. And it doesn't work, because the problem isn't discipline. It's architecture.
Your reports already decided which questions were worth answering
Every system you have installed — the ERP, the CRM, inventory, payroll — produces reports. Each of those reports was designed at some point by someone who decided, in good faith and with the information available that year, which questions mattered. That person is rarely still at the company. The decision is.
As long as your questions match the ones that design anticipated, everything works. The day a question occurs to you that isn't in any report — and the questions that matter are almost always of that kind — something else begins: someone exports to Excel, someone else cross-references two files by hand, a third person messages the one who knows. Two days later an answer arrives that doesn't match the one from another department, and neither can be audited.
That manual work isn't a symptom of disorder. It's the logical consequence of querying a business through pre-imagined reports.
A case: twelve locations, one number nobody had
A network of preschools in Florida faced exactly this. The question was elementary: how many children will be enrolled next term. Staffing depends on that number, and teacher-to-child ratios are state-regulated — coming up short is a legal problem, going over is a cost paid every month.
Each location kept its own figure. None was wrong. One counted signed enrollments, another counted those who had paid the deposit, another included renewals it took for granted because the family had been there three years. Management didn't have a number: it had twelve, and between the most optimistic and the most conservative lay tens of thousands of dollars a month in hiring decisions.
Nobody was doing their job badly. There simply was no unambiguous version of reality, and the cost of that absence appeared in no financial statement: it appeared as one location overstaffed and another understaffed.
What was done wasn't asking the locations to report better. It was pulling the data straight from the systems where it already lived, consolidating it in one place, writing down for the first time what counts as a confirmed enrollment, and letting management query that figure whenever they needed it — without asking for it, without waiting for Monday, without it passing through anyone's judgment.
Why this is different now
For thirty years, solving this meant a long project: model a data warehouse, define in advance which questions it would answer, build it, and discover six months later that the questions had changed. One set of pre-imagined reports was replaced by another, newer and more expensive.
What changed is that the raw, disaggregated information already sitting in your databases — every transaction, every record, every movement, exactly as it was written — can now be analyzed directly. You no longer have to anticipate the question in order to answer it. The asset was never the report: it was always the data underneath, which until recently was only reachable by translating it into a format someone had decided on in advance.
That's what lets an executive query their own business and get the answer on the spot. It isn't one more dashboard. It's no longer depending on the question having been foreseen.
The goal isn't to remove managers
Worth saying plainly, because it's the reasonable fear in this conversation. What's redundant isn't people: it's the work of manually consolidating what the systems already hold, and the friction of every answer passing through three filters before it reaches the decision.
A manager's judgment — knowing that the big client is unhappy, that one location has a teacher on leave, that an order is about to change — isn't in any database and doesn't get automated. What can be taken out of the way is the mechanical part, so that judgment gets applied to information nobody disputes, instead of being spent arguing over which figure is the right one.
Where it starts
Not with technology. It starts with an uncomfortable map: which decisions get made in your company, what information each one needs, where that information lives today, and how many different definitions it lives under. That map almost always produces the same finding, and it almost always stings: the contradictions don't come from bad data, they come from definitions nobody ever wrote down.
Writing them is cheap. Finding them is the work.
A question to close on: the last time two of your departments gave you different numbers for the same thing, did anyone check which one was wrong — or did anyone check why both were right?
---