There's a conversation I've had more times than I can count, and it always has the same two people in it. One owns the Fabric bill and wants it lower. The other owns governance and wants budget for it. And they treat each other as opponents — cost versus control, the classic tension — when they should be the same person, because they want the same thing and don't realise it. The cheapest capacity you can run is the one you're not wasting, and governance is how you stop wasting it. Cost management and governance aren't in tension. Done properly, governance is cost management wearing a less popular name.

Let me make that concrete, because "governance saves money" sounds like the kind of thing a governance person would say to get their budget approved.

Where the waste actually lives

Fabric charges you for capacity — a pool of Capacity Units that everything draws from. So "wasting money" means "consuming capacity that produces no value." And when you go looking for that wasted consumption, it turns out to live almost entirely in the places governance is supposed to look:

  • The duplicate models nobody retired. Ungoverned estates accumulate semantic models like a garage accumulates boxes — three slightly different versions of the same thing, each refreshing on its own schedule, each burning capacity, because no one owns deciding which is canonical and deleting the rest. Every redundant refresh is money spent to keep a duplicate alive.
  • The reports that refresh far more than anyone needs. A dataset set to refresh eight times a day, feeding a report three people open once a week, is pure waste — and it exists because no one governs the question "how fresh does this actually need to be?" Refresh frequency is a governance decision with a direct line to the invoice.
  • The pipelines reprocessing everything, forever. The "reload all of history every night" pattern, inherited from an era when compute was free, quietly re-buys the same historical data every single night. Nobody decided to do that. Nobody governs it. It just runs, on the meter.
  • The abandoned workspaces still consuming. Projects end; their workspaces and scheduled jobs often don't. Ungoverned, they keep drawing capacity for outputs no one consumes, because retiring things requires someone to own retiring things.

Look at that list. Every entry is a governance failure — no ownership, no lifecycle, no agreed definition of "needed" — and every entry is also a cost. They're the same problems. The waste and the ungovernedness are two descriptions of one condition.

Governance as the cheapest optimisation you have

Here's why this reframing matters practically. When the bill is too high, the instinct is to reach for technical optimisation — tune the queries, resize the capacity, squeeze the workloads. That work is real and it helps. But it's often the hard, low-yield path, because you're optimising the efficiency of work that shouldn't be running at all.

Governance is the higher-yield path, and it's usually cheaper to execute. You don't need to make the duplicate model faster; you need to delete it. You don't need to optimise the eight-times-a-day refresh; you need to change it to daily because that's all anyone needed. The single most effective cost lever in most Fabric estates isn't a technical tuning exercise — it's the governance work of deciding what's canonical, who owns it, how fresh it needs to be, and what gets retired. That work removes whole workloads from the capacity rather than making wasteful ones marginally leaner.

Technical optimisation makes the wrong work cheaper to run. Governance stops the wrong work from running. One of those is a much better deal, and it's the one nobody frames as cost control.

How to actually do it

So if you want to cut your Fabric bill through governance — which is to say, sustainably rather than by panic-resizing — here's the approach I'd take:

Start from the capacity metrics, not from a policy document. Open the Fabric Capacity Metrics app and find what's actually eating the pool — the specific models, refreshes, and pipelines burning the most. That list is your governance priority queue, ordered by cost, which is a far more motivating way to prioritise governance than "we should really catalogue things." Then, for each expensive item, ask the governance questions: Is this a duplicate — can it be consolidated? Who owns it — will they defend its refresh frequency? Does it need to run this often, or at all? Retire the dead, consolidate the duplicated, right-size the over-refreshed, and assign an owner to what remains so it doesn't silently regrow.

Do that and something quietly satisfying happens: your bill drops and your estate gets more governed, because they were never two projects. You went looking for waste and found ungoverned data; you went looking for governance and found the budget to justify it hiding in the invoice.

The bottom line

Stop treating cost and governance as a trade-off, because framing them as opponents is how organisations end up doing neither well — the cost people resize capacity to hide waste they never removed, and the governance people write policies disconnected from any consequence anyone feels. Put them together and each solves the other's problem. Governance gives cost management its highest-yield lever: not running work that shouldn't run. Cost gives governance the thing it always struggles for: a concrete, funded, board-legible reason to happen. The cheapest data platform isn't the one with the smallest capacity. It's the one that isn't wasting the capacity it has — and getting there is a governance project with a price tag attached, in the good way.