At a glance
You did not save money. You deferred it, at an interest rate nobody quoted you.
That is the whole argument, and it is not a scare story. It is simply what happens when the work that makes software durable gets skipped rather than done, because skipped work does not disappear. It waits.
The price you were quoted
Something built in a fortnight for a fraction of what an agency would charge. Screens, a database, a login, the core of the thing. Genuinely working, genuinely impressive, genuinely cheap.
We are not going to pretend that is not appealing. If the alternative was a five-figure quote and three months, of course you took it. Most people would, and in plenty of cases it was the right call.
The trouble is that the quote covered building the software. It did not cover owning it, and owning it is where the money lives.
Where the rest of the cost turns up
Downtime. Not the hosting bill, but the hours: yours, your staff's, and the goodwill of customers who could not do the thing they were trying to do. This tends to arrive at the busiest moment, because load is what triggers it.
Support burden. Software with no error handling and no logging generates a steady drip of confused emails that somebody has to answer. That somebody is usually you, and it is usually in the evening.
The security incident. Most people never have one. The ones who do find it is not the fix that costs, it is everything around it: notifying people, the obligations that come with personal data, and the customer who quietly does not renew.
The feature that turns out to be impossible. This is the sneaky one. Six months in you need something perfectly reasonable, and it cannot be built on the existing structure without unpicking a great deal of what is already there. The estimate comes back at four times what it should, and nobody can quite explain why.
The rebuild. Eventually, for some, the accumulated cost of the above exceeds the cost of starting again properly. That is the bill nobody budgeted for, and it usually lands at the least convenient possible moment.
The house with no foundations
The analogy is overused because it is accurate.
You can build a house quickly by skipping the foundations. It will look like a house. It will have walls, a roof, and a functioning front door. For a while it will be a perfectly good house, and anyone viewing it would say so.
The problem is not that it falls down tomorrow. It is that everything you do afterwards is constrained by it. You cannot add an extension. You cannot put anything heavy upstairs. Every repair is more awkward and more expensive than the equivalent repair on a sound building. And the eventual fix is not a repair at all, it is lifting the house up and doing the job that was skipped, at a cost that dwarfs what it would have been at the start.
Software works the same way, with one cruel difference: you cannot see the foundations, so nobody knows they are missing until something heavy goes upstairs.
Why it compounds rather than sits still
This is the part that makes waiting expensive.
A structural problem in software does not stay one problem. Every feature built afterwards is built on top of it, and inherits it. The workaround in month two becomes the assumption in month six, which becomes the thing three other pieces of code depend on by the end of the year.
Fixing it in month two means changing one thing. Fixing it in month eighteen means changing one thing and everything that grew around it. Same fault, considerably larger bill, and none of the intervening work made it better.
How to not have this problem
Nothing here requires spending a fortune upfront.
Know which category you are in. Prototype, or software people depend on? If it is genuinely the first, build it however you like. Most of the trouble comes from things that quietly changed category without anyone noticing.
Get it looked at before it is load-bearing. A review while the thing is small is quick and cheap. The same review after two years of features is neither.
Budget for the last twenty per cent from the start. If you know the unglamorous engineering is coming, it is a planned phase rather than an emergency.
Fix foundations before adding floors. When something structural turns up, resist the urge to build the next feature first. It is always more expensive afterwards.
We have laid out the real numbers in more detail here: the hidden cost of vibe-coded software.
Which brings it back, as usual, to AI being a brilliant typist and a terrible architect. Typing is the cheap part. It always was.
Want to know what you are actually sitting on before it gets expensive? Our AI Code Audit is a fixed-price £495 review that tells you plainly what you are dealing with, or get in touch for a no-pressure conversation.
