Managed services pricing: what application support should cost and how it is structured

01The four pricing models

Most managed services agreements use one of four structures, and each creates a different incentive. Read the incentive, not just the number - that is what determines whether your incident count goes down over three years or quietly up.

02Why retainer usually wins

A retainer priced against a service level aligns both sides: the vendor keeps the same fee whether you have forty incidents or four, so reducing incidents is directly in their interest. That is the only model where the supplier profits from your system becoming more reliable.

The objection is that you pay the same in a quiet month. That is true, and it is the point - you are buying availability of expertise and a service level, not a quantity of tickets. The comparison that matters is total cost of ownership over a year, including the incidents that a stability-motivated vendor prevented and a volume-motivated one would have billed for.

03What the price actually depends on

Coverage hours are the single biggest driver. Business-hours support in one timezone is a fraction of the cost of genuine 24x7, because 24x7 requires a rota with enough people to sustain it. Many organisations buy 24x7 by reflex and then discover that almost nothing they support genuinely needs a 3am response.

Beyond that: the number and variety of applications, how well documented they are, how much observability already exists, and whether the vendor is taking on code they did not write. Inheriting an undocumented system is real work before the first ticket, and vendors who do not price for it either quote low and change later, or quote high to cover the unknown.

• Coverage hours - business hours versus true 24x7 is the biggest lever.
• Application count and variety, more than user count.
• Documentation and observability quality at handover.
• Whether an improvement allocation is included or billed separately.

04Typical ranges

The bands below reflect mid-market application support with senior engineers on blended or offshore delivery. They assume a reasonably documented estate; inheriting an undocumented system typically adds a one-off onboarding cost of four to six weeks of effort.

05The clause that matters most

Ask whether the retainer includes an improvement allocation - a defined share of each month spent reducing the causes of recurring incidents rather than only responding to them. Without it, you are buying restoration in perpetuity, and the same three problems will still be happening in year three.

The second clause worth reading closely is exit. What happens to runbooks, monitoring configuration and knowledge if you leave? A contract that makes exit painful is a contract you will renew for the wrong reasons, and the time to negotiate that is before signing rather than during a transition.

06What good looks like in reporting

Monthly reporting should include SLA misses, not only hits. A supplier who reports 100% attainment every month for two years is either exceptional or measuring something convenient - and it is usually worth asking which incidents were excluded from the calculation and why.

Also ask for incident trend data by root cause. If the same category keeps recurring, either the improvement work is not happening or it is not targeting the right thing. That single chart tells you whether the relationship is working better than any satisfaction survey.

• Require SLA misses in the report, not just attainment.
• Track incident trends by root cause, not just volume.
• Agree severity definitions in business terms before go-live.
• Confirm what happens to knowledge and tooling at exit.