What it costs to keep software alive after launch

Launch day does not end the spending: it starts it

Almost every software quote describes an ending: a delivery date, a figure and an invoice. What it describes badly is everything after that. Software is not a building you finish and open; it is closer to an installation somebody has to keep switched on. Nobody shows you the launch invoice with the year-two invoice beside it, and that is exactly the missing conversation.

The expensive way to find this out is the common one: a site or an app that runs on its own for two years, until one day it stops and there is nobody to call. Everything saved on maintenance over twenty-four months gets spent in the single week when somebody has to understand, from scratch, a system they did not build.

What maintenance actually is

Maintenance is not “fixing things when they break”. That is the visible part and, with luck, the rarest. Most of the work is letting the world change underneath the software without the software falling over:

• Monitoring. Finding out before your customer does that something has stopped responding. Without it, your alerting system is annoyed users.
• Backups that restore. A backup nobody has ever tested is not a backup: it is a folder with good intentions.
• Dependency updates and security patches. The dullest work there is, and the most expensive to skip: the longer you leave it, the bigger the jump.
• Third-party platforms shifting under you. Apple and Google tighten requirements every year, a payment gateway moves to a new API version, a data source changes format without warning. None of that is your decision.
• Certificates, domains and keys that expire. Trivial to fix and spectacular in public.
• Small improvements. The twenty-minute change that does not justify opening a project and that, with no channel to ask for it, ends up never happening.

Notice that only the last item has anything to do with what you want the software to do. The other five are the price of it continuing to exist.

What it costs each month

These are our published prices. They are not a market range lifted from a report: they are the tiers we operate on.

On the web and platform tiers we respond within one working day. The only thing separating the critical tier from the rest is the written SLA and the out-of-hours rota, and that is precisely what makes it expensive: it is not more work, it is availability.

Set against the cost of building, it looks like this. A custom platform starts at €30,000 and its operation at €450 a month: €5,400 a year, exactly 18% of the project every twelve months. A local business site starts at €3,500 and is maintained from €150 a month: €1,800 a year, more than half the project every year — and the whole project again inside two. The ratio changes sharply with size, and that is the part almost nobody explains.

Why there is a floor

Because monitoring, taking backups, checking they restore and applying patches costs about the same on a small system as on a large one. What scales with size is the failure surface — more integrations, more apps, more countries — not the watching. That is why €150 a month is not a token price: it is what it costs to have somebody looking.

That floor also explains a recurring pattern. The projects that end worst are usually not the big ones but the small ones handed over with no operating agreement at all, because the monthly fee looked expensive next to a four-figure budget. Next to rebuilding them, it was not.

What breaks when nobody maintains it

• The app stops installing. Apple and Google change requirements every year, and an app that is never updated ends up off the store. Getting back on costs more than keeping current would have.
• The update becomes a migration. Skipping two years of a framework’s releases is not twice the work of one: it is a project.
• Data is lost quietly. Nobody discovers the backups have been empty for months until the day one is needed.
• Third-party changes go unnoticed. A source changes format, a screen goes blank, and if nobody is watching it stays blank for weeks.
• The knowledge evaporates. Whoever built it has moved on, and the cost of understanding the system gets paid all over again.

How to buy it without being tied in

• Is there a lock-in period? On ours, no: one month’s notice on either side. If someone needs a long contract to keep you, the product is the contract, not the service.
• What is the response time, and where is it written down? “As soon as possible” is not a commitment.
• Are the backups tested? Ask when the last real restore was, not whether backups exist. The answer tells you a lot.
• Are small improvements included or billed separately? If every twenty-minute change opens a quote, you will never ask for one.
• Do they take on software somebody else built? The answer should be yes, after an audit. And if the audit ends with them telling you it cannot be kept running safely, that answer is worth more than the fee.

How we know

Because maintenance is not a filler product here. We run SomosTorre, the civic platform for Torrelavega, with its native iOS and Android apps and its live data sources. And we run our own store, which takes payment across 235 countries and territories, 112 currencies and four languages, with the ERP, the warehouse and the carriers connected behind it. When a source changes format or an app store changes its requirements, the call comes to us.

It is the same rota. Buying maintenance from us means joining the on-call we already run for our own systems, not asking us to invent a new one.

Watching a small system costs almost the same as watching a large one. That is why the percentage looks alarming at the small end, not the large one.