English 
Português
Maintaining software typically costs 15% to 25% of its original build cost per year, according to a widely used market rule of thumb. For a system that cost US$ 2 million to build, that means US$ 300,000 to US$ 500,000 per year (illustrative calculation). The figure rises with technical debt, integrations, poor documentation and demanding availability targets. The largest cost, however, is usually the cost of not maintaining: downtime, rework and a loss of capacity to evolve.
In this guide
- How much does it cost to maintain software?
- What goes into the bill: the 4 types of maintenance
- Why maintenance costs more than it seems
- What pushes costs up (or down)
- The cost of not maintaining: downtime and technical debt
- Contracting models: which one to choose
- AMS or traditional technical support?
- How to build a maintenance budget in 6 steps
- 8 ways to reduce maintenance costs
- Maintain or modernize? Signs it is time
- 10 questions to ask before hiring a support provider
- Frequently asked questions
- Conclusion
Delivering a system does not end the investment in it. After go-live, the work continues: fixes, adaptations to new versions and regulations, improvements requested by the business, risk management. That whole set is maintenance, and it lasts as long as the software exists.
For CTOs, CEOs, COOs and CFOs, the useful question is not only “how much does it cost?” but “how much does it cost to do it right, and how much to do it wrong?”. This guide brings together reference ranges, the factors that move the budget, the contracting models and a roadmap for deciding with more predictability.
How much does it cost to maintain software?
The market rule of thumb places annual maintenance between 15% and 25% of the original development cost. It is a starting point, not a guarantee, and should be adjusted for criticality, code quality and the required service level.
Edana describes this range as a starting point, to be adjusted for factors such as criticality, technologies in use and demanding SLAs. The same article presents the case of a Swiss industrial company that began at 20% per year and had to raise it to 35% the following year, because of undocumented technical debt and a tool with declining support.
Here is how the rule translates into values (illustrative examples, unrelated to NextAge pricing):
| Build cost | Maintenance at 15% per year | Maintenance at 25% per year |
|---|---|---|
| US$ 500,000 | US$ 75,000 | US$ 125,000 |
| US$ 2 million | US$ 300,000 | US$ 500,000 |
| US$ 5 million | US$ 750,000 | US$ 1.25 million |
A simple calculation helps the CFO: at 20% per year, five years of maintenance equal the cost of building the system. Approving a development budget without planning for this cash flow means looking at only part of the investment.
What goes into the bill: the 4 types of maintenance
Direct answer: software maintenance is any change made after delivery. It divides into corrective, adaptive, evolutionary and preventive, and most of the budget usually goes to evolution, not to fixing failures.
| Type | What it is | Example | Estimated share of budget* |
|---|---|---|---|
| Corrective | Fixes defects and vulnerabilities in production | Error in an invoice calculation | 15% to 25% |
| Adaptive | Adjusts the system to changes in environment or rules | New database version, tax or data protection change | 20% to 30% |
| Evolutionary (perfective) | Creates or improves features | New module, banking integration via API | 40% to 60% |
| Preventive | Reduces future risks | Refactoring, dependency updates, testing | Varies by strategy |
*Ranges from Edana (2026), a consultancy estimate rather than academic research. Use as an order of magnitude.
There is also a layer that rarely appears in proposals but weighs on results: monitoring, governance and indicator tracking. It separates a system that merely “receives tickets” from one that is actually managed.
The sustainment model used by Brazil’s federal government offers a good scope definition. According to Annex VI of Inmetro’s procurement documents, which cites Portaria SGD/ME no. 5,651/2022, sustainment covers the activities needed to keep software available, stable and performing in production within service levels. The catalog includes corrective maintenance, small adaptations, technical support and diagnosis, under a fixed monthly payment.
How NextAge approaches this
In Application Managed Services (AMS), the four types of work appear as pillars of a single contract: corrective (fixing failures), adaptive (adjustments), evolutionary (improvements) and governance (management focus). The goal is that the client does not need a new contract for every new need.
Why maintenance costs more than it seems
Because it accumulates over years and is underestimated in the initial budget. Depending on the source, maintenance accounts for more than half of total cost of ownership (TCO), reaching 70% or 80% in more conservative estimates.
Edana states that maintenance can add up to 70% to 80% of total investment over the life of the software. Other sources give different numbers, which shows there is no universal percentage. Treat these ranges as a warning, not as a target.
The most concrete data on budget overruns comes from a survey of 500 IT decision-makers in the United States and the United Kingdom, at companies with annual revenue of at least US$ 500 million. According to Ensono (2025), 49% reported that the actual cost of maintaining legacy systems exceeded plan over the past year, and 86% said budget pressure hurts innovation.
Three reasons explain this gap:
- The initial budget looks at construction. What happens afterward is treated as “support”, with an optimistic estimate.
- Cost grows over time. Between the third and fifth year, requests for evolution increase; later, technical debt and outdated dependencies make every change more expensive.
- Knowledge is invisible. When only two people understand the system, staff turnover becomes a financial risk.
What pushes costs up (or down)
Direct answer: the main factors are complexity, code quality, documentation, test automation, integrations, availability requirements and the scarcity of professionals in the technology used.
| Factor | Effect on cost | How to control it |
|---|---|---|
| High technical debt | Up: Edana suggests an extra provision of 10% to 20% | Code audit, incremental refactoring plan |
| Demanding SLA (24×7, 99.9%) | Up: 10% to 15% of the maintenance budget, per Edana | Define SLA by system criticality, not one standard for all |
| Third-party integrations and APIs | Up: external changes create unplanned work | Dependency inventory and version monitoring |
| Weak documentation | Up: every analysis takes longer | Document during maintenance, not afterward |
| Automated testing | Down: reduces regressions and rework | Minimum coverage on critical flows |
| Stack with few professionals | Up: higher pay and longer lead times | Modernization plan or a partner who masters the technology |
| Growth in users and data | Up: demands performance and scale | Load testing and periodic capacity review |
| Regulatory obligations (GDPR, Brazil’s LGPD) | Up: mandatory adaptations | Regulatory calendar in the planning |
The cost of not maintaining: downtime and technical debt
The cost of not maintaining shows up as hours of outage, emergency rework and accumulated technical debt. For mid-size and large companies, one hour of system downtime can cost hundreds of thousands of dollars, according to global research.
In the ITIC 2024 survey, covering more than a thousand companies worldwide, over 90% of mid-size and large organizations estimated that one hour of downtime costs more than US$ 300,000, and 41% pointed to between US$ 1 million and US$ 5 million or more (self-reported values, excluding fines or litigation). The same survey shows that only 22% of organizations can accurately calculate the cost of an hour of downtime.
That gap is the first number in your budget. To calculate it, add up:
- Lost revenue during the outage.
- Idle team cost (salary and overhead of those who cannot work).
- Recovery work (incident response hours).
- Contractual or regulatory fines and penalties.
Technical debt is the other side of the equation. According to McKinsey (2023), it represents about 40% of IT balance sheets; companies pay 10% to 20% extra on every project to deal with it; and 30% of the CIOs surveyed believe that more than 20% of the budget meant for new products is diverted to resolving technical debt. At national scale, CISQ estimated the cost of poor software quality in the US at US$ 2.41 trillion in 2022, with about US$ 1.52 trillion in accumulated technical debt (an aggregate estimate, useful as an order of magnitude).
How NextAge approaches this
If you do not know what an hour of system downtime costs, the diagnostic conversation in NextAge’s AMS starts there: NextAge maps the operational risks of your systems and delivers an action plan, at no cost and with no commitment.
Contracting models: which one to choose
The most common models are an in-house team, hour bank, fixed monthly package with SLA, per-ticket or function-point pricing, and outcome-based contracts. For critical systems, a fixed package with SLA tends to deliver more predictability.
According to Avasant (analysis of IT and application managed services pricing, second half of 2024), fixed-price and hybrid models remain dominant, while delivery- and outcome-based models are gaining ground.
| Model | When it fits | Risk | How to control it |
|---|---|---|---|
| In-house team | Stable systems and a team with capacity | Dependence on a few people; high fixed cost | Document, train backups, measure utilization |
| Hour bank | Sporadic demand, low risk | Paying for unused hours; unpredictable delivery | Monthly cap and consumption report |
| Fixed monthly package with SLA | Critical systems with continuous demand | Poorly defined scope leads to disputes | Clear scope, SLA and indicators in the contract |
| Per ticket or function point | Predictable, measurable volume | Focus on volume, not on root cause | Recurrence targets, not only response targets |
| Outcome-based | Well-measurable goals (availability, for example) | Poorly measured goals distort the contract | Auditable metrics and periodic reviews |
How NextAge approaches this
A structured hour package, a defined SLA and a clear scope from the start of the contract: this is how AMS at NextAge brings predictability to what is usually a variable cost. One third of the package is dedicated to monitoring, technical governance and tracking of operational indicators.
AMS or traditional technical support?
AMS (Application Management Services) is a model of continuous application management, with SLA, governance and planned evolution. Traditional support is reactive: it waits for the problem to happen before fixing it.
Among AMS providers in the market, the service usually includes tiered support (N1 to N3), monitoring, incident, problem and change management, and progress reporting. What changes from one vendor to another is the depth of governance and the ability to act on root causes.
| Criterion | Traditional support | One-off project | AMS |
|---|---|---|---|
| Approach | Reactive, per ticket | Per specific demand | Proactive and continuous |
| Scope | Incidents | A single deliverable | Correction, adaptation, evolution, governance |
| Cost predictability | Low | Defined per project | High (package and SLA) |
| Visibility for management | Little | Per deliverable | Reports, indicators and periodic meetings |
| System knowledge | Scattered | Concentrated in the project | Accumulated by a dedicated team |
How NextAge approaches this
NextAge’s AMS combines root-cause-oriented corrective support, continuous evolution and technical governance with executive follow-up: periodic meetings, monthly performance reports and access to real-time operational indicators. NextAge works across stacks such as Java, Node.js, .NET, React and Angular, with relational and NoSQL databases, in Azure, AWS and GCP environments.The service is designed for mid-size and large companies whose operations depend on proprietary or heavily customized systems, with a lean IT team. In the interest of transparency: it is not a fit for micro-companies with simple systems, for those who only need a one-off developer, or for those who decide purely on lowest price.
How to build a maintenance budget in 6 steps
List your systems, classify them by criticality, calculate the cost of downtime, use your ticket history, apply the reference percentage and reserve a margin for risk.
- Take inventory. List applications, owners, technologies, integrations and current support contracts.
- Classify by criticality. Systems that generate or protect revenue deserve a stricter SLA; the others can have leaner coverage.
- Calculate the cost of downtime. Use the formula from section 5 for each critical system.
- Survey your ticket base. Volume, type, resolution time and recurrence show where the hidden cost lies.
- Apply the reference percentage. Start at 15% to 25% of build cost, then adjust for technical debt, SLA and integrations.
- Reserve a risk margin. Edana recommends a reserve of 15% to 25% and an annual review with low, nominal and high scenarios. On technical debt, McKinsey warns that simply setting aside 15% to 20% of the IT budget is not enough: the most effective approach is to build debt reduction into every budget, with targets and indicators.
8 ways to reduce maintenance costs
Direct answer: costs fall when problems are solved at the source, the system is tested and documented, and management tracks indicators. Cutting support hours without this tends to raise costs the following year.
- Act on the root cause. Fixing the symptom costs twice: the incident returns and consumes another ticket.
- Automate tests. Fewer regressions, less rework in every release.
- Keep documentation alive. It shortens analysis time and reduces dependence on key people.
- Monitor proactively. Detecting degradation before the incident is cheaper than handling the emergency.
- Refactor incrementally. Small continuous improvements avoid a full rewrite.
- Consolidate vendors. Fewer points of contact and concentrated knowledge reduce transition losses.
- Track indicators. Resolution time, recurrence, deployment failures and the ratio of evolutionary hours to build hours (Edana treats a ratio above 1:1 as a critical point).
- Use automation and AI with judgment. They help with analysis and testing, but according to Ensono, AI speeds up modernization for only 39% of teams and brings new challenges for 33%; it works best when the foundation is already organized.
How NextAge approaches this
Fixing the symptom costs twice. NextAge’s corrective support is root-cause oriented and backed by technical governance, so the same incident does not generate another ticket. Learn more about AMS at NextAge.
Maintain or modernize? Signs it is time
Modernizing makes sense when the cost and risk of maintaining exceed those of evolving. The safest path is usually incremental modernization, in parts and with validation at every stage.
Signs that modernization needs to enter your planning:
- Technical debt is high, and every simple change takes weeks.
- The technology faces a growing shortage of professionals.
- The architecture limits business scale or integration with new platforms.
- Maintenance costs overrun the budget year after year (the case for 49% of companies in the Ensono survey).
- There are security vulnerabilities that the current version does not allow you to fix.
Rewriting everything from scratch is usually the riskiest option. McKinsey also notes that some technical debt is not worth paying off, which reinforces the need to measure before deciding.
How NextAge approaches this
Modernizing does not have to mean rewriting everything. NextAge works with incremental modernization: it replaces parts of the system in a planned way and validates each stage before moving on, without exposing operations to the risk of a complete rewrite. The assessment can be done within AMS.
10 questions to ask before hiring a support provider
Direct answer: the contract must make clear what is in scope, how the service is measured and how exit works. Use this list in RFPs and decision meetings.
- What is included in the package (corrective, adaptive, evolutionary, governance), and what is left out?
- What is the SLA by incident type, and what penalty applies if it is missed?
- How much of the package is dedicated to monitoring and governance, not just ticket handling?
- How is the root cause addressed, and how is recurrence measured?
- Which reports and indicators will be delivered, and how often?
- Who will be on the team, and how is system knowledge preserved if someone leaves?
- How are onboarding and knowledge transfer handled, and how does the exit work?
- Which technologies and cloud environments does the vendor master?
- How are security, code access and data protection compliance handled?
- How is out-of-scope work approved and billed?
How NextAge approaches this
A practical example: a Brazilian dairy cattle association faced cloud instability, manual invoicing and limits on data collection. With continuous support, NextAge led the migration to a more stable environment and integrated invoicing with the banking system through APIs. See the cases on the AMS page.
Frequently asked questions
How much does software maintenance cost per year?
In general, 15% to 25% of the original development cost per year, according to recurring market estimates. A system that cost US$ 2 million tends to require US$ 300,000 to US$ 500,000 per year (illustrative calculation). The value rises with technical debt, integrations, poor documentation, user volume and availability requirements.
What is the difference between maintenance, support and sustainment of software?
Support serves users and handles incidents. Maintenance is the change to the software: fixing, adapting, improving or preventing. Sustainment is the set of activities that keeps software available, stable and performing in production, as defined in Portaria SGD/ME no. 5,651/2022 (Brazil). AMS adds governance, SLA and planned evolution in a continuous contract.
What are the types of software maintenance?
Corrective (fixes defects), adaptive (adjusts the software to environment changes, such as database versions, operating systems or legislation), evolutionary or perfective (adds or improves features) and preventive (reduces future risks, such as refactoring and dependency updates).
What is AMS (Application Management Services)?
It is a model of continuous management of enterprise applications that combines defect correction, adaptations, evolution and technical governance in a recurring contract, with a defined SLA. It usually includes tiered support (N1 to N3), monitoring and reporting. At NextAge, the service is structured around four pillars: corrective, adaptive, evolutionary and governance.
Hour bank or fixed monthly package: which should you choose?
An hour bank suits sporadic demand and low-risk systems. A fixed monthly package with SLA tends to suit critical systems better, because it gives budget predictability, clear scope and indicator-based follow-up. Outcome-based contracts require well-measured goals.
How can you reduce software maintenance costs?
Act on the root cause of incidents, automate tests, keep documentation current, monitor the system, refactor incrementally, consolidate vendors and track indicators such as resolution time and recurrence. Treating technical debt continuously prevents it from becoming a crisis.
How much does one hour of system downtime cost?
It varies widely by sector and company size. In the ITIC 2024 survey, over 90% of mid-size and large companies estimated the cost at more than US$ 300,000 per hour. Calculate yours by adding lost revenue, idle staff, recovery work and contractual fines.
When should you modernize instead of continuing to maintain?
When the cost and risk of maintaining exceed those of evolving: high technical debt, languages with a shortage of professionals and an architecture that limits scale. Incremental modernization, in parts and with validation at every stage, is the safest path.
Does outsourced support replace the in-house IT team?
Not necessarily. It can act as a strategic extension of the internal area, adding expertise, governance and execution capacity, especially when the team is small and does not cover every front.
Conclusion: maintenance is the cost of protecting revenue
Maintaining software costs, as a rule, 15% to 25% of its build value per year, but the number that matters most to the board is a different one: how much it costs not to maintain. With systems down, technical debt piling up and knowledge concentrated in a few people, the bill arrives anyway, and usually higher.
The safest path combines three decisions: budget for maintenance from the start, choose a contracting model with clear scope and SLA, and track indicators continuously.
Your systems deserve more than an open ticket.
Talk to NextAge to understand the reality of your systems and design a support model that protects your operations before the next incident happens. You leave with a diagnosis of your risks and an action plan, at no cost and with no commitment.
Talk to NextAge
