Excel Sprawl vs a Governed Semantic Model

The question is not which tool is better. It is which number the board sees when four departments each maintain their own version of revenue — and whether anyone can say which one is right.

Framed as a tool comparison, this discussion never resolves. Excel is genuinely better at some things, the BI platform at others, and both camps produce accurate arguments indefinitely. Framed as a governance question, it resolves in about ten minutes.

From the standard

“Power BI semantic models represent a source of data that's ready for reporting and visualization.”

What Excel is actually excellent at

It is worth saying plainly, because the migration pitch usually pretends otherwise. Excel is the best tool most organizations have for analysis where the structure changes as you think: scenario modelling, reconciliation, anything requiring a column that exists for twenty minutes, and financial work where an analyst needs to override a value and see the consequence immediately.

No BI platform matches that iteration speed, and organizations that try to eliminate spreadsheets entirely find their analysts exporting to CSV within a month. The tool is not the problem.

Where it fails as a system of record

The failure begins when a spreadsheet stops being an analysis and becomes a report other people rely on. At that point the properties that make Excel excellent become liabilities.

The definition is invisible. A measure lives inside a cell formula, in one file, on one person's drive. Nobody outside that file can see how revenue was calculated, and nobody is notified when it changes.

It forks silently. Someone copies the file to adapt it for their region. Both copies are maintained. Both are now correct in their own terms and disagree with each other. Nothing detects this.

Access control is file-level. Row-level security — this regional manager sees their region — is not expressible. In practice the whole file is shared and the sensitive columns are hidden, which is not a control.

There is no audit trail. Six months later, nobody can say who changed the figure, when, or why.

What a governed semantic model buys

The value is not the visuals. It is that a metric has exactly one definition, stored where it can be inspected, versioned, and secured.

Revenue is defined once. Every report referencing it inherits that definition, and changing it changes every consumer at the same moment. Row-level security is a property of the model rather than of who received which file. The definition sits in source control, so a change is a reviewable diff with an author and a date attached.

That is what makes the board meeting a decision rather than a reconciliation exercise.

Why sprawl happens anyway

Organizations that already own a BI platform still accumulate shadow spreadsheets, and the reason is almost never that people prefer Excel. It is queue length.

A business unit needs a view. The BI backlog is eleven weeks. They export to a spreadsheet and build it themselves — not as a governance decision but as a scheduling one. Every week the queue stays long, more definitions escape the model.

Which means governance policy alone does not fix sprawl. Banning spreadsheets while the queue stays at eleven weeks produces the same spreadsheets with worse naming. The durable fix is making the governed path faster than the ungoverned one.

A migration that works

Do not migrate files. Migrate definitions.

Inventory the spreadsheets that circulate and identify the handful of metrics they disagree on — usually revenue, margin, headcount, and one operational measure specific to the business. Implement those in the semantic layer, get the definitions agreed in a room with the people who currently disagree, and publish.

Most of the spreadsheets then collapse into thin views over a governed measure. The ones that survive are genuine modelling work, and those should stay in Excel — now reading a trusted number instead of recomputing their own.