Once my software is built, what will it cost me every year to keep it running?
Short answer: Plan for yearly running costs of roughly 15 to 20 percent of the build price as a starting assumption, split between hosting and services (the fixed part) and fixes and changes (the part you control). Custom software is software written for your business alone. For a small internal tool, the hosting part can be a few hundred dollars a year, and the human work is most of the bill.
You are about to pay for a custom tool, and you want to know what the bill looks like in year two and beyond.
I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.
The direct answer
Custom software has two kinds of yearly cost. The first is rent: hosting, database, email sending, error alerts and your domain name. For a small internal tool this is usually small and predictable. The second is work: someone fixing bugs, applying updates, answering "why did it do that" and making small changes. This part is larger and depends on how you buy it.
A common rule of thumb says to budget 15 to 20 percent of the original build cost per year. It is a planning assumption, not a law. I explain below when it fits and when it does not. If you want a monthly view instead of a yearly one, how much app maintenance costs per month goes through what a monthly plan should include.
Why software needs upkeep at all
Software does not wear out, but everything around it changes. Libraries get security fixes. Browsers change. A service you depend on changes its prices or rules. Your own business changes too: a new tax, a new product, a new hire who needs a login. Wikipedia's overview of software maintenance notes that maintenance is the longest phase of a program's life and is often cited as the largest share of its lifetime cost. That is why the first build price is only part of the real price.
Four kinds of work fill a maintenance bill:
- Fixes: something breaks or behaves wrongly.
- Keeping up: updates, security patches and changes in the services it connects to.
- Changes: new fields, new reports, new rules. This is the part owners can turn up or down.
- Watching: alerts, backups and checks so you hear about problems before customers do.
The fixed part: hosting and services
These are the line items for a typical small web app. Prices are the vendors' published starting plans, as of November 2026, in USD.
| Item | What it does | Starting price |
|---|---|---|
| Vercel Pro | Hosts the app on the web | $20 a month |
| Supabase Pro | A database service that stores your data and handles logins, with daily backups kept 7 days | $25 a month, first project included |
| Sentry Team | Error monitoring: tells you when the app throws an error | $26 a month when billed annually |
| Resend Pro | Sends your app's emails, such as password resets | $20 a month for 50,000 emails |
| Domain name | Your web address | Check your registrar, usually renewed yearly |
Many small apps need less. Vercel has a $0 Hobby plan meant for personal projects, Sentry has a free Developer plan and Resend has a free plan with a daily limit of 100 emails. Supabase's free plan pauses projects after a week of inactivity and does not include daily backups, so it is not a fit for a business tool. Read managed databases for a custom business app for how to choose.
Adding up the four paid lines above gives $91 a month, or $1,092 a year. A plain internal tool may use two or three of them and land between $540 and $1,000 a year. Heavy use costs more, because most plans charge for extra traffic, storage or messages.
The work part: fixes and changes
Here are two ways to buy the human work. The figures below are a worked example with made-up numbers, built on a $12,000 internal tool.
| Item | Care plan | Fix when it breaks |
|---|---|---|
| Hosting and services (from the table) | $1,000 | $1,000 |
| Monitoring and updates | Included in plan, assume $100 a month = $1,200 | Not done, until something fails |
| Small changes | Assume 6 hours a year at $120 = $720 | Assume 10 hours at $120 = $1,200 when needed |
| One emergency | Usually covered | Assume 8 hours at $150 = $1,200 |
| Total a year | $2,920 | $3,400, with surprise timing |
The care plan is $2,920 against $3,400, or 24 percent against 28 percent of the build price. The two are close on paper. The real difference is that the care plan is a number you know in advance, and the second path arrives as an invoice on your worst day. For what a plan should cover for a website, see what care plans cover.
When the 15 to 20 percent rule works, and when it does not
The rule is a reasonable first guess for a mid-size build. It breaks in these cases:
- Cheap builds: 15 percent of $3,000 is $450, which does not cover hosting plus any real work. Small builds need a floor, not a percentage.
- Very large builds: a $200,000 system does not need 15 percent of that to keep a database running, but it will need a standing team.
- Busy periods: the first year after launch has more fixes, because real users find things testing missed.
- Apps with many connections: payments, maps, AI services and email each add rules that change. More connections, more upkeep.
- AI-built apps: code written fast by an AI tool often needs cleanup before it is cheap to maintain. I cover that in how to move an AI-built app to production.
How to keep the bill under control
- Ask for the running costs before you sign. A good quote lists hosting, services and upkeep, not only the build.
- Own the accounts. The hosting, database and domain should be in your name, so you can move later. Ask for that in writing.
- Cap the changes. Set a yearly number of hours for changes. Unlimited changes are the most common way upkeep grows.
- Turn on alerts. Free tiers of monitoring tools catch most failures early. Early fixes are cheap.
- Keep a backup you control. A weekly copy outside your host is cheap insurance. See AI-built app backups.
- Review once a year. Cancel services you do not use and move off tiers you have outgrown or never needed.
When a higher tier or a bigger plan is worth it
Pay for more when a failure costs more than the upgrade. Daily backups are the clearest case: if losing a day of data would cost you more than $25 a month, use a plan that includes them. If losing even an hour would hurt, Supabase sells Point-in-Time Recovery as an add-on, at about $100 a month for 7 days of history as of November 2026. Most small tools do not need that. Daily backups plus your own weekly copy are enough.
Questions to ask before you sign
Whoever builds the tool, put these to them in writing. The answers tell you more about your yearly cost than the quote does.
- What will it cost to run each month, itemized? You want each service named, with its plan and price.
- Whose name are the accounts in? Yours is the right answer for hosting, database, domain and code.
- What is included after launch, and for how long? A fixed number of weeks of bug fixes at no charge is common. Ask what counts as a bug and what counts as a change.
- What is the hourly rate or the monthly plan price after that? Get the number now, not when you need help.
- How fast will you respond if it breaks? "Same day" and "within a week" are different products.
- What happens if you are unavailable? A one-person developer should tell you who can step in and where the notes are kept.
A vendor who cannot answer these is telling you the upkeep is not planned.
How I'd approach this
I tell owners to think of a custom tool like a vehicle, not a purchase. You pay for the vehicle once and then pay to fuel, insure and service it. I would set a yearly budget at 15 to 20 percent of the build, push most of it into a fixed monthly plan and keep a small pot for changes you cannot predict.
If you already have a tool with no one looking after it, start with a check of what is running, where the data lives and what would break first. My fixes and finishes start around $1,500, and a fix-or-rebuild decision is usually the first step. A custom tool that is looked after is an asset. One that is ignored turns into a bill.
Questions people ask
What percent of build cost should I budget for software maintenance each year?
A common planning rule is 15 to 20 percent of the original build cost per year. Use it as a first guess, then replace it with a real quote that lists hosting, services and hours of work. Small builds need a minimum amount, not a percentage.
Can I keep the cost near zero by using free plans?
For a hobby project, yes. For a business tool, free plans have limits that matter. Supabase's free plan pauses projects after a week of inactivity and has no daily backups, so a tool your team relies on belongs on a paid plan.
Is it cheaper to pay a monthly care plan or fix things when they break?
Usually a care plan costs about the same over a year but makes the cost predictable, and it includes monitoring that catches problems early. See what care plans cover.
Who pays for hosting if the developer sets it up?
Ask that the accounts be in your name so you pay the vendor directly, or that the developer passes the cost through with no markup. Either way, get it in writing, along with ownership of the code and data.
What does software maintenance cost for an AI-built app?
The hosting costs are the same. The work cost depends on code quality, which varies a lot. A short code check first tells you whether upkeep will be cheap or costly. My custom software development cost guide covers build prices too.
More in this topic: Problem solving: custom software (22 guides) →