Software maintenance starts with what must not stop
You ask for a small change to the customer portal and expect an hour of work. The answer that comes back: days of investigation first, because nobody knows exactly which parts will be affected. That is the moment you see how much software maintenance has been left undone.
Other problems choose the worst possible moment. You notice an order that does not come through or a slow portal exactly when customers are ordering and staff are logging in. And after launch, a lot around your software keeps changing: browsers, iOS and Android, external APIs, security standards and your own processes.
This is the order I would work in: first map out what stops when a component fails and who has access to what. Then a fixed maintenance rhythm. I only talk about a rebuild once targeted clean-up falls short.
Maintenance beyond fixing bugs
Besides fixing the bugs users report, there is preventive work: replacing outdated libraries, improving logging, checking capacity. And there is work that follows from changes outside your organisation, such as a payment provider changing something or a new iOS version.
Then there is the process that made sense at the start and has become a daily drag two years later. Data is copied between systems by hand. A report has to be corrected every time before anyone can use it. The customer portal does not keep up with the number of users.
The rule of thumb I use: if a task comes back every week, it belongs on the maintenance list.
You pay for delay in emergency work
Every temporary fix, unpatched dependency or unclear integration makes the next change harder. That is how technical debt builds up, and you pay the interest in hours. The one-hour change that needs days of investigation is a direct result.
With security, putting things off becomes an operational risk. Outdated components can contain vulnerabilities. Without updates, monitoring and a clear owner, such a risk sometimes goes unnoticed until someone exploits it. For an online shop or an organisation holding customer data, that can hit revenue and reputation directly.
Delay often ends in emergency work, which costs more than planned maintenance. You fix an outage under time pressure, with no room to deal with the cause properly, which makes another temporary fix more likely. So work out what it costs you when your most important system is not reliably available for a day, and put that figure next to the price of planned maintenance.
Five signals to check today
Go through these points and count how many apply to you:
- Updates only happen after something breaks, or when a supplier threatens to stop.
- Staff have fixed manual workarounds for tasks that should be automated.
- Nobody can quickly tell you which integrations, servers and software versions are business-critical.
- Small changes take remarkably long, because of unclear code, missing documentation or dependence on one person.
- Performance drops at peak times and the cause is only investigated once customers complain.
If you recognise one or two, targeted clean-up can be enough: better monitoring, or replacing that one fragile integration. With three or more, I would first decide how long the system still needs to last, before any money goes into updates.
Where you draw the line depends on the business impact and the age of the system. An internal tool for twenty users makes different demands from a platform your revenue runs on.
Software maintenance needs an overview first
Without an overview, maintenance is just firefighting. The first thing I would make is a list of all your systems: which applications there are, where they run, which external services are connected and who has access. Next to each component, note which process stops when it fails. Then you set priorities on business impact, not on who shouts loudest.
Make that list technically useful too. Write down where the source code lives, how the deployment process works and who can get into the hosting, the backups, the domains, the API keys and the documentation. If those pieces are spread across different parties without a clear handover, an outage quickly turns into a search. One party points at the server, another at the code and a third at the external integration. Meanwhile your customer waits.
That is why a single point of contact for development, hosting and management is often faster. Responsibility sits in one place. When performance drops, application code, database, infrastructure and traffic can be looked at straight away. That shortens the analysis, and an incident spends less time stuck between suppliers.
A rhythm that fits your risk
The rhythm scales with what is at stake: from frequent monitoring and quick follow-up for an e-commerce environment or SaaS service to a fixed monthly check for a stable internal application. One exception: a security update for an acute risk never waits for the next round.
Use those fixed moments for more than updates. Also review the error messages, the performance, the backup results and capacity, the open improvement points and what is changing at external suppliers. After a few rounds you see which small issues keep coming back, and those are the ones worth fixing properly.
Testing is the last place I would cut costs. Under pressure it is tempting to put an update straight onto production, and then the risk lands on your users and customers. An acceptance environment, automated tests and a clear rollback plan take time and make changes manageable. The more business-critical the application, the less sensible it is to release and hope for the best.
What belongs in your maintenance agreement
A contract that only says 'support is available' is of little use when something goes down. Spell out:
- which incidents get priority;
- within what time you get a first response;
- whether there is monitoring outside office hours;
- who assesses security updates;
- what counts as regular maintenance and what counts as new development.
Ask about capacity as well. An agency that only responds when it has time to spare can leave a small but urgent change waiting longer than needed. If you currently work with a prepaid block of hours, check when hours were last taken off it. If that was after an outage every time, you are only paying for repairs after the fact. For a system your business runs on, I would go for a fixed maintenance agreement. It gives room to get to know your environment and spot recurring problems, so problems get dealt with sooner.
At LJPc we combine development, management and hosting under one roof. That lets us fix a fault and also look at why it happened and whether the environment itself plays a part. For us, fast and personal support means direct contact and clear responsibility.
Which process must keep running tomorrow?
An application with up-to-date components, clear documentation and controlled releases is easier to extend as your organisation grows. A new integration or an extra user group then does not have to start as a rescue project.
Open a blank document today and write down the digital process that must keep running tomorrow. Next to it, note which system carries it and who can get into it. Wherever you cannot give an answer, that is where your maintenance starts.