An organization completes a consolidation programme. Finance, workforce, delivery and customer data now live in one platform. The technical work succeeded. The management questions are no more answerable than before.
This outcome is common enough to be worth explaining precisely, because it is usually attributed to the wrong cause.
Connection is not comparability
The systems are connected. The meanings are not.
"Customer" in the CRM is an account record created by a salesperson. In the ERP it is a billing entity created by finance. One organization may be three billing entities; three subsidiaries may be one account. The join executes. The result is wrong in a way no error message reports.
"Active employee" differs between payroll and HR by contractors and mid-month leavers. "Project" in a delivery system may be a workstream, while in the ERP it is a cost collector, and a single delivery project may map to four cost collectors or none.
Two tables joining successfully tells you the keys matched. It tells you nothing about whether the things behind the keys are the same things.
Why consolidation programmes miss it
The programme was scoped as a movement problem: get the data into one place, on a schedule, reliably. Those are the acceptance criteria, and they are met.
Semantic reconciliation is a different kind of work. It requires deciding — with the business, not for it — which definition wins where two exist, and getting a person to own that decision. It cannot be completed by a data team alone because most of the questions are business-policy questions wearing technical clothing.
So it gets deferred, and then it gets skipped, and the platform goes live with four internally consistent silos in one location.
The diagnostic
Take one important cross-functional question. Trace it, out loud, from the source systems to the answer.
You are looking for the point at which someone says "well, it depends what you mean by…" That point is where the semantic work stopped, and it is almost always earlier than anyone expected.
What actually connects a business
Not more integration. Three narrower things:
- A small number of shared entities — usually person, project, customer, period — defined once, with a stated source of truth
- Join paths that hold across history, including through reorganizations and system migrations
- Certified metrics with named owners, so that a definition is a decision rather than an opinion
Most organizations need agreement on four or five entities to unlock the majority of their cross-functional questions. That is a smaller job than a consolidation programme, and it is the one that produces the explanation.
Practical check
Start with one disputed management question and trace the answer through the consolidated platform. Record every point where the meaning depends on an unstated choice. That list—not the number of connected tables—is the remaining integration backlog, because those choices determine whether the output can support a decision.
A Management Intelligence Build begins by proving what actually connects across the full dataset — not a sample.