Managed Services

What exactly is a managed service?#

A managed service is a continuing arrangement in which a provider accepts responsibility for operating, monitoring, maintaining, or improving a defined customer function. Delivery usually combines software, automation, standardized procedures, and human judgment. Payment is commonly recurring, per managed unit, usage-linked, or a hybrid.

A 2022 joint advisory from CISA and international cybersecurity agencies defines managed service providers as entities that deliver, operate, or manage ICT services for customers under a contractual arrangement such as a service-level agreement. That definition is cybersecurity-specific, but its structure generalizes: an ongoing function, delegated operating responsibility, and a contract that defines performance. (CISA joint advisory)

Managed services appear in IT operations, cybersecurity, cloud infrastructure, finance, compliance, data engineering, AI operations, revenue operations, customer support, and industry-specific workflows. The label matters less than the operating boundary: what will the provider run, what remains with the customer, and how will both parties know the service was delivered?

Managed services versus adjacent models#

The categories can coexist. A vertical SaaS company may charge for software, onboarding, and an optional managed workflow. The accounting and operating model should keep those components separate even when the buyer receives one proposal.

The responsibility boundary is the product#

Founders often describe the service by listing tasks: monitor dashboards, respond to tickets, prepare reports. Buyers care about the responsibility boundary — which systems, accounts, data sets, or business units are covered; which events trigger provider action; what the provider can change without approval; which decisions remain with the customer; what response, restoration, or quality target applies; what evidence proves performance; and what happens when volume, complexity, or risk exceeds the agreed baseline. If these answers are ambiguous, every sale creates a new product.

Why do managed services matter to founders?#

They can turn expertise into a repeatable offer#

Early customers may value an outcome but lack the people, process, or confidence to operate new software. A managed layer can close that adoption gap and let a startup observe the real workflow, learn where exceptions occur, and build reusable runbooks before automating the stable portions. This can be a strong go-to-market wedge in vertical software and AI. The risk is becoming a custom agency whose software merely supports bespoke delivery. The test is whether each cohort can be served with fewer exception hours, shorter onboarding, and more consistent quality.

Recurrence improves visibility, but does not guarantee quality#

Ongoing contracts can make revenue more predictable than one-off projects and may improve retention. Neither property makes the revenue SaaS-like. Accenture's Q3 fiscal 2026 results illustrate both the scale and the caveats of the category: managed services generated $9.48 billion, about 51% of the quarter's $18.70 billion in revenue, and the company notes managed-services bookings are typically multi-year and convert to revenue over a longer period than consulting. It also warns that bookings are not a substitute for revenue analysis and that many contracts can be terminated on short notice. A startup should therefore separate signed bookings, non-cancelable contract value, recurring billings, recognized revenue, and cash collected. (Accenture Q3 FY2026 results; Form 10-Q, quarter ended May 31, 2026)

Managed services expose the real product gap#

Human delivery produces high-resolution evidence about missing integrations, confusing interfaces, weak alerts, incomplete data, and exceptions software does not handle. A disciplined founder converts that evidence into product priorities; an undisciplined one lets the team patch every gap manually and mistakes customer heroics for product-market fit. Track whether the service is becoming more productized:

Human delivery hours per comparable managed unit

Exception rate = cases requiring nonstandard intervention / total eligible cases

Straight-through rate = cases completed within the quality threshold without human intervention / total eligible cases

The last metric should include a quality threshold. Automation that silently creates rework is not operating leverage.

Buyers transfer work, not accountability#

A customer can delegate operation without eliminating its regulatory, security, or business-continuity exposure. NIST's supply-chain guidance recommends defining supplier requirements and matching their rigor to supplier criticality. The FTC's Safeguards Rule is a concrete regulated example: covered financial institutions must select capable service providers, put safeguards in contracts, monitor provider work, and reassess providers periodically. These rules do not apply to every customer, but they show why buyers request security evidence, audit rights, incident procedures, subcontractor disclosure, and exit support. (NIST SP 1305; FTC Safeguards Rule guide) For a founder, compliance is part of the commercial design: a high-privilege service can shorten the path to value and simultaneously lengthen procurement, increase insurance needs, and enlarge breach impact.

The operating framework: scope, load, service level, and control#

1. Scope: define the managed unit#

Choose a unit that can be counted and whose workload is reasonably predictable: endpoint, cloud account, location, data pipeline, invoice, campaign, user cohort, support queue, or monitored asset. For each unit, publish a service catalog covering included routine actions, included incident types, supported systems, operating window, standard reports, volume allowance, explicit exclusions, and change-order triggers. The unit is useful only if it explains cost — one "customer" may have ten times another's assets, data volume, or exception load.

2. Load: estimate demand and capacity#

Measure delivery demand from observed workload rather than sales intuition:

Required hours = managed units x events per unit x handling hours per event + fixed account hours

Productive capacity = delivery FTEs x paid hours per FTE x productive-service fraction

The productive-service fraction excludes leave, training, internal meetings, and administration. U.S. Bureau of Labor Statistics data make the cost point from another angle: in March 2026, private-industry compensation averaged $32.60 per hour in wages and $14.01 in benefits, or $46.60 total. This economy-wide figure is not a staffing quote for specialists, but it shows why a wage-only delivery model materially understates labor cost. (BLS Employer Costs for Employee Compensation, March 2026)

3. Service level: promise what operations can control#

A service-level agreement should connect a measurable event to a measurable commitment — availability, monitoring coverage, acknowledgement time, response time, restoration time, completion time, defect rate, or reporting timeliness. Distinguish acknowledgement (receipt confirmed), response (qualified work begins), restoration (service resumes temporarily), resolution (underlying issue corrected), and outcome (business result improves). Promising a 15-minute response is not promising a 15-minute resolution. Define the measurement clock, severity criteria, exclusions, maintenance windows, and remedy. A percentage without a denominator and clock is not an operable commitment.

4. Control: assign duties and preserve evidence#

Create a responsibility matrix for the provider, customer, and material subcontractors covering identity and access, data quality, configuration approval, incident declaration, backup, recovery, and offboarding. CISA's guidance emphasizes shared responsibility, least-privilege access, logging, monitoring, and clear contractual roles because providers can become high-leverage targets. The provider should retain only the access needed, use separate privileged identities, preserve audit logs, test recovery, and give customers evidence rather than a generic claim of security. (CISA joint advisory)

Key formulas and metrics#

Revenue (hybrid managed service):

Monthly service revenue = base fees + managed-unit fees + usage or overage fees + earned outcome fees - service credits

Keep one-time setup, migration, hardware resale, and pass-through spend separate; combining them with recurring fees hides the economics of the continuing service.

Direct cost and gross margin:

Direct delivery cost = delivery labor + benefits + contractors + delivery tools + customer-specific infrastructure + on-call and QA + expected credits

Gross margin = (managed-service revenue - direct delivery cost) / managed-service revenue

Classify costs by function, not title: an operator who performs the contracted workflow belongs in direct delivery cost even if the job title says customer success.

Capacity and payback:

Capacity coverage = productive capacity / expected required hours

CAC payback months = customer acquisition cost / (recurring revenue per account x recurring gross margin)

Use the recurring gross margin of the managed service, not a corporate blended margin or an aspirational SaaS margin. Track gross and net revenue retention on recurring managed-service revenue, and watch why accounts expand — growth from more managed units is healthier than growth from emergency custom work that damages margin.

Pricing and packaging choices#

AWS's current managed-operations materials show two real patterns: infrastructure management priced from the managed environment, and an operations-on-demand catalog sold in monthly 20-hour blocks. The point is not to copy AWS pricing but that even a large provider defines the managed resource, catalogs eligible work, and meters exceptional labor. (AWS Managed Services FAQ; AWS Operations on Demand documentation) Avoid "unlimited support" unless the eligible requests, channels, users, response target, fair-use boundary, and abuse remedy are unambiguous. Unlimited ambiguity is an unpriced liability.

Worked example#

Assume an early-stage cloud-operations provider manages 20 customer accounts.

Monthly revenue: 20 accounts x $6,000 = $120,000.

Direct delivery cost:

• 3.0 delivery FTEs x $12,000 fully loaded = $36,000
• 0.5 FTE shared on-call and QA x $12,000 = $6,000
• Monitoring, ticketing, and security tools: 20 x $500 = $10,000
• Customer-specific cloud and data costs = $4,000
• Expected service credits and rework reserve = $2,000
• Total = $58,000

Gross margin = ($120,000 - $58,000) / $120,000 = 51.7%

This is a viable service margin in the example, not a universal benchmark. Capacity is the next question. Each account consumes ~18 delivery hours per month, so Expected required hours = 20 x 18 = 360. The 3.5 delivery FTEs each have 160 paid hours, but only 70% is productive customer service: Productive capacity = 3.5 x 160 x 0.70 = 392 hours, giving Capacity coverage = 392 / 360 = 1.09 — about 9% headroom. If incident load raises demand 15%, required hours rise to 414 and coverage falls to 0.95: the same profitable P&L is now under-capacity.

CAC payback: at $18,000 acquisition cost, Monthly gross profit per account = $6,000 x 51.7% = $3,102, so CAC payback = $18,000 / $3,102 = 5.8 months — credible only if that margin and retention hold.

What should the founder change? Suppose automation cuts average delivery time from 18 to 15 hours without degrading quality: required hours fall to 300 and coverage rises to 1.31. Margin does not improve immediately if the team stays fully employed — the business has created capacity. Margin improves when that capacity prevents hiring, supports more customers, or replaces contractors. Treating saved hours as instant accounting profit overstates the benefit.

Related concepts#

• Subscription Model: distinguish recurring access from recurring operating responsibility.
• Hybrid Pricing: combine a service retainer with managed-unit or overage charges.
• Usage-Based Pricing: meter variable events without hiding customer cost.
• Outcome and Performance-Based Pricing: structure variable fees when outcomes are attributable and auditable.
• Pricing Metric and Value Metric: select a managed unit that tracks value and delivery cost.
• Packaging: define tiers, inclusions, exclusions, and add-ons.
• Cost-Plus Pricing: understand the cost floor without treating cost as willingness to pay.
• Value-Based Pricing: connect continuing responsibility to customer economic value.
• Ideal Customer Profile: target customers with similar systems, workload, and risk.
• Bootstrapping vs. Venture Capital: match the model's labor and working-capital needs to financing strategy.

Sources#

1. CISA and international partners, Protecting Against Cyber Threats to Managed Service Providers and their Customers (2022)
2. NIST, Cybersecurity Framework 2.0 Quick-Start Guide for C-SCRM, SP 1305 (2024)
3. Federal Trade Commission, Safeguards Rule: What Your Business Needs to Know
4. Financial Accounting Standards Board, ASU 2014-09, Revenue from Contracts with Customers (Topic 606)
5. U.S. Bureau of Labor Statistics, Employer Costs for Employee Compensation, March 2026
6. Accenture, Reports Third-Quarter Fiscal 2026 Results (June 2026)
7. Accenture, Form 10-Q for the quarter ended May 31, 2026
8. AWS, Managed Services FAQ
9. AWS, Operations on Demand documentation

---

This page is an educational and operating explanation, not accounting or legal advice. Have qualified advisers review material contracts, revenue recognition, employment classification, privacy, security, and sector-specific obligations.