Software Procurement · 9 min read · November 11, 2026

Software Maintenance: What the Contract Covers and What It Doesn't

Software maintenance is everything that happens to a system after it goes live: fixing errors, keeping up with changes in the environment, improving what already exists, and adding what the business asks for. A good software maintenance contract separates these types of work and states what the monthly fee covers, what is billed separately, and how quickly someone responds when something stops working. In this article, we explain how we organize this at SKAD and what you should ask before signing with any vendor.

This article is part of the guide: Custom Software Development Cost: What Really Drives the Price

Developer working on a laptop with code on the screen
Photo by Danial Igdery on Unsplash

The four types of software maintenance

The classic breakdown has four types. In practice, it matters because each one has a different pace, a different level of urgency, and a different way of being billed.

When someone talks about corrective and evolutive maintenance as if they were the same thing, they are lumping together two jobs of opposite nature: one puts out fires, the other builds. Mixing both in the same bucket of hours is one of the most common sources of friction between client and vendor.

Adaptive and preventive maintenance are the most overlooked, because nobody complains about them until the day a language version loses support or an app store rejects an update. For us, this is what separates ongoing system support from simple ticket handling.

  • Corrective: fixes defects that show up in use, such as a wrong calculation, a screen that freezes, or an integration that fails. It is reactive and, by nature, unpredictable.
  • Adaptive: adjusts the system to external changes, such as a new version of the programming language or database, a phone operating system update, a change in a partner's API, or a new legal requirement.
  • Evolutive: adds or changes features. A new module, a report, a different approval flow.
  • Preventive: touches what hasn't broken yet. Updates dependencies, fixes known vulnerabilities, tests backup restoration, cleans up fragile pieces of code.

What a software maintenance contract usually includes

There is no universal standard, but there is a set of items that tends to show up. In general, a software maintenance contract covers fixing defects in what was delivered, security updates, answers to technical questions and, when hosting is involved, monitoring and backup.

What almost always stays out is evolution. New features, changes to business rules, and integration with another system are development work, and they go into a quote or a separate package of hours.

The items below also tend to be left out, and it is worth asking about them explicitly:

  • Defects caused by changes that third parties made to the code or the server.
  • Problems in other companies' systems that yours integrates with.
  • Paid third-party services the system uses, such as email delivery, SMS, maps, or payments.
  • Training for new users and correction of incorrectly entered data.
  • Scope changes disguised as fixes, such as behavior that works as it was specified but that the team now wants to be different.

Warranty and maintenance are not the same thing

This is one of the most common mistakes we see. A warranty covers defects in the delivery for an agreed period, usually short. Maintenance is an ongoing service that continues after the warranty ends.

If the contract only mentions a warranty, the day after the term ends nobody is responsible for the system. If there is a monthly fee, confirm whether it also covers what the warranty no longer covers, and starting when.

Hosting and monitoring: who takes care of the server

A system can be perfect and still go offline because the server's disk filled up or the security certificate expired. That's why the hosting question needs to be asked before you close the maintenance deal, not after the first scare.

There are two common arrangements. In the first, the vendor hosts and maintains everything, and is responsible for the whole. In the second, hosting stays with the client or another provider, and the vendor only looks after the code. The second works, but it requires a clear agreement about who has access to what and who is responsible when the cause is the server.

Monitoring is another point that gets confused. Knowing the system went down before a user calls is different from having someone contracted to act on the alert. Ask what is monitored (availability, response time, application errors, disk space), who receives the alert, and during which hours that person is on call.

A backup is only worth something if someone has tried restoring it. Ask that the contract state how often backups are made, where they are stored, and how often restoration is tested.

SLAs and support channels: what needs to be in writing

An SLA is a service level agreement: the deadlines the vendor commits to meeting. The most frequent mistake is writing a single number, like "support within 4 hours," without saying whether that means a response or a resolution. Replying that the ticket was received solves nothing.

A useful SLA defines at least three things: the severity classification, the first-response time, and the target time for a resolution or workaround at each level. It also says during which hours it applies, because business-hours support and 24-hour support are very different services.

As for channels, the rule is simple: one official channel, with a record. Tickets sent by email or through a ticketing system leave a history, and that protects both sides. A direct message to the developer gets things done fast on a good day, but it doesn't scale, it disappears when the person goes on vacation, and it proves nothing in a dispute over deadlines. An extra channel for real emergencies, such as a phone line or an on-call message, is reasonable.

How the contract should handle hours, emergencies, and evolution

This is where most disputes begin. The most common models are a fixed monthly fee with a defined scope, a bank of hours, and pay-per-request billing. None is always the best: a monthly fee gives predictability, a bank of hours gives flexibility, and one-off requests suit systems that hardly ever change.

In our experience, it works well to split the work into two blocks. One covers ongoing support (corrective, adaptive, and preventive maintenance, hosting, and monitoring). The other is evolution, with hours or modules estimated and approved before work begins. That way nobody pays for evolution thinking it's a fix, or discovers a fix being billed as evolution.

On hours, put in writing whether unused hours roll over or expire, how hours are logged, and whether you can see the report. On emergencies, define what counts as urgent: a system outage, risk of data loss, and a security breach are examples. Also define whether urgent after-hours support carries an extra charge or is already included in the monthly fee.

For evolution, require a simple flow: you describe what you need, the team estimates, you approve, and only then does work start. If you want to understand how development effort is estimated and what it depends on, our guide on how much a custom system or app costs details the factors without promising prices.

Why well-structured code makes maintenance cheaper

There is a part of maintenance cost that no contract clause can fix: code quality. In a system where everything is mixed together, with business rules inside the screen, database queries scattered around, and no tests, every fix requires untangling a mess. The same task that would take an hour in an organized system takes a day, and the risk of breaking something else is higher.

The signs of code that makes maintenance more expensive are easy to recognize: no documentation, very old dependencies, no automated tests, a single person who knows how everything works, and fear of touching anything because "last time it broke." If your current vendor describes the system this way, the monthly fee will reflect that risk, one way or another.

When the system is already in that state, replacing everything at once is usually overkill. The path we advocate is modernization done gradually, without stopping operations, which is why our legacy system modernization happens incrementally. For the full picture, see the guide on legacy system modernization.

What to ask before hiring software maintenance

Before signing, bring these questions to the conversation. The answers say more about the vendor than the commercial proposal does. If you want a broader checklist, the article on how to choose a software house complements this list.

  • Which of the four types of maintenance are included in the monthly fee, and which are billed separately?
  • Is the SLA deadline for a response or for a resolution? And during which hours does it apply?
  • Who hosts the system, who monitors it, and who acts when the alert fires?
  • How often is the backup made, and when was restoration last tested?
  • Do unused hours roll over? How do I receive the report of what was done?
  • How do I request and approve an evolution before it is billed?
  • Are the code and documentation mine? If I switch vendors, what does the transition look like?
  • How many people know my system? What happens if one of them leaves?

From our experience

At SKAD, some of our client relationships span a long time. The Inspefire client has been served for 10 years, and in that period we developed more than 10 new modules based on the specifications they send us. For The Horus, we maintain and host the system and the apps under a monthly fee. In both cases, maintenance is ongoing work, not a service that begins after the project ends.

We also refactored, with AI support, the SyncStop, 4Car, KaraokeApp, The Bee App, and DEM (equipment inspection) systems. All of them were plain PHP with jQuery and Bootstrap, and the apps were built with Cordova. Today they run on PHP 8.3 with Slim 4 for the API, React with Vite for the web, and React Native with Expo for the app. We kept the PHP language and changed the framework and the architecture. After this refactoring, maintaining these systems became lighter.

Frequently asked questions

What is software maintenance?

It is the set of work that keeps a system running and useful after delivery: fixing defects, adapting it to external changes, preventing future problems, and adding improvements. It is divided into corrective, adaptive, preventive, and evolutive maintenance.

What is the difference between corrective and evolutive maintenance?

Corrective maintenance fixes something that doesn't work as it should. Evolutive maintenance changes or adds something the system never did. They are usually billed in different ways, and it's good for the contract to make the boundary clear.

Does every system need a software maintenance contract?

If the system is used in the day-to-day of the business, yes. Even if you never change a single feature, it depends on servers, libraries, and external systems that keep updating, and someone needs to keep track of that.

Is hosting part of maintenance?

Not always. In many contracts it is a separate service, or it stays with the client. Confirm in writing who hosts, who monitors, and who is responsible when the problem comes from the server.

Can I switch maintenance vendors later?

You can, if the contract guarantees access to the source code, the documentation, and the hosting credentials. Ask about this before signing, because asking afterward is usually harder.

Conclusion

Before closing any maintenance contract, write down on one page what you expect from the system: what can't go down, how quickly you need a response, and which evolutions are already on the radar. With that in hand, any proposal becomes easier to compare. If you are still at the beginning, and maintenance is a criterion for choosing who to build with, the best test is to see how the team works before you sign. You can request the free prototype of your app and evaluate the code, the communication, and how the team handles your questions.

Keep reading