There is a line item in your data budget that nobody approved, nobody can cancel, and nobody knows the size of. It funds work that produces no decisions. It grows quietly every quarter. It is paid out of capacity, not cash, which is why it never triggers a CFO review. I call it the Invisible Tax, and in every growth-stage data portfolio I have audited, it sits between 25% and 40% of total spend.
What the tax actually pays for
The Invisible Tax is the recurring cost of working around problems that should have been designed out. It has four components, in roughly the order of size:
- Reconciliation work. Two dashboards, one number, three people in a meeting trying to figure out which is right. Multiply by ten metrics, twenty teams, fifty meetings a quarter. The cost is in person-hours that produce no new information.
- Translation work. Every team has its own definition of "active user," "revenue," "churn." The data team becomes a translator instead of an analyst. Every new question requires re-stating the definitions before it can be answered.
- Re-answering. The same question, asked by a different person, gets re-built from scratch because the previous answer is buried in a Slack thread or a deprecated dashboard.
- Waiting. Permissions, broken pipelines, missing context, the analyst who knows is on vacation. Capacity that was paid for sits idle.
Why it stays invisible
Three reasons:
It is denominated in time, not money. A €40k Snowflake overrun triggers an emergency. Forty hours of reconciliation work per week across ten people, which costs the same, does not. The salary line absorbs it.
It looks like normal work. Every individual instance of friction is rational. Of course you reconcile two numbers before showing them to the CEO. Of course you re-build the dashboard nobody could find. Of course you wait for permissions. Each instance is correct. The aggregate is the problem.
Naming it is socially expensive. The Invisible Tax is not anyone's fault, it is a property of how the portfolio grew. But naming it feels like blaming individuals, which is why nobody does.
How to size it
You can estimate the tax in three weeks with a simple time-bucket exercise, the same one I describe in the FTE Debt essay. The four buckets are:
- Decision work
- Plumbing (work the company would do anyway)
- Friction (reconciling, translating, re-answering)
- Waiting
Friction + Waiting is your Invisible Tax. Annualize it. Multiply by fully-loaded cost. That number is what you are paying, every year, to keep the friction in place.
The compounding problem
The Invisible Tax has a feature that makes it especially dangerous: it compounds. Each new tool adds a small reconciliation tax against every existing tool. Each new team adds a translation tax against every existing team. Each new dashboard adds a maintenance tax. The math is not linear, it is closer to n².
Which is why companies that doubled their data team in two years often find their decision throughput did not double, or even rise. The capacity went into paying the tax.
Why hiring does not fix it
The default response to data slowness is to hire more analysts. But adding a person to a system with 35% friction means 35% of the new person also goes to friction. You bought more capacity, you wasted a larger fraction of it, and the friction% itself often rises because the new person needs to be onboarded into the friction.
The fix is not more capacity. The fix is to recover the capacity you already paid for.
Three places to start cutting the tax
Recovery work is unglamorous. It is also where most of the ROI lives:
1. Kill the duplicate dashboards. In every portfolio I audit, 40–60% of dashboards have not been opened in 90 days. They still cost compute, they still cost trust ("which one is right?"), and they still cost maintenance. Inventory, audit, retire.
2. Centralize metric definitions. One number, one definition, one owner. The technical implementation matters less than the social one, somebody owns "active user" and gets pinged whenever a new team needs the number.
3. Document the decision, not the data. When a decision gets made, document which numbers it was based on and where to find them. Most re-answering happens because the prior decision left no trail.
What this is worth
For a 30-person data-touching organization with €85k average cost and 35% friction, the Invisible Tax is roughly €900k a year. Cut the friction in half, not to zero, just in half, and you recover €450k of annual capacity without hiring anyone. The calculator does the math.
That is the budget for two senior hires, three new initiatives, or a 30% reduction in the data spend, depending on what the company needs more.
The CEO question
The single best question a CEO can ask the data leadership team this quarter is: What share of our team's time pays the Invisible Tax, and what would it take to halve it?
If the answer is "we don't measure that," the next conversation is the one worth having.