Why percentage-of-spend pricing quietly kills MSP FinOps practices

The incentive problem hiding in plain sight

A percentage-of-spend fee looks simple because it scales automatically with the estate. It also means the provider earns less after a successful optimisation programme and more when client waste grows. That tension is easy to ignore while an MSP is still running isolated, one-off cost reviews. It stops being ignorable the moment that work becomes a durable, recurring service line.

What a successful optimisation cycle does to your own revenue

Take a straightforward illustrative example: an MSP billing 15% of monthly Azure spend as a FinOps fee for a client running $50,000/month. That's $7,500/month in fee. A first optimisation cycle — the kind every FinOps engagement is supposed to deliver — finds and fixes real waste and brings spend down to $38,000/month. The fee drops to $5,700/month, a 24% cut, landing exactly when the MSP's own delivery cost was highest: discovery, validation, and implementation for that first cycle is the most labour-intensive phase of the whole relationship. Do this well enough, often enough, and the model punishes competence.

Operational discipline, not spend, is what's actually being delivered

The ongoing work — continuous detection, triage, evidence gathering, remediation guidance — scales with the complexity of a client's environment: how many resources, how many subscriptions, how fast the estate changes. It does not scale with the size of the bill. A client with a large, stable spend and a tidy estate can need less ongoing attention than one with a smaller, sprawling, fast-changing one. Pricing against the bill charges for the wrong variable.

Three ways to price that don't have this problem

Fixed bands by environment size — price tiers set by resource or subscription count, not dollars. A client's fee doesn't move just because their Azure bill does.

A stable managed-spend band — bucket spend into bands wide enough that one normal optimisation cycle doesn't knock a client into a lower one. Growth and shrinkage both have to be substantial before the price changes.

A capped, time-limited outcome component — a share of validated savings for the first few months of an engagement only, on top of a flat operational fee. It rewards finding the win without turning "spend stays high" into a permanent revenue strategy.

Margin protection is a delivery problem before it's a pricing problem

None of the above holds up if every engagement is bespoke. Discovery, prioritisation, and evidence gathering need to follow the same repeatable process account to account — a standard finder catalogue, a standard validation workflow, a standard evidence format a client's engineers can act on without a custom explanation each time. That repeatability is what lets a fixed or banded fee stay profitable instead of becoming a loss leader on your most complex accounts.

A healthier commercial unit

Price against a stable operating unit — a managed-spend band, a portfolio tier — rather than a direct share of the bill. An MSP can still acknowledge that a larger, more complex estate costs more to service, without its margin depending on the client's cloud bill moving in the wrong direction. It's the same reason Cerberus Cloud is priced in managed-spend bands rather than as a share of the bill.

What clients should be buying

Clients aren't buying a monthly PDF or a promise that every pound of waste disappears. They're buying continuous detection, qualified engineering judgement, safe execution, and evidence that changes produced the expected result. A pricing model should make that operating system easy to understand and renew — not quietly reward the MSP for how much the client wastes.