The most expensive mistake in business software isn't picking the wrong tool — it's building one you never needed. The quieter mistake is the opposite: patching around a tool's limits for three years, paying a "workaround tax" in staff time that would have covered a proper build twice over.
Quick answer: Most businesses never need custom software — an off-the-shelf tool or a small workaround covers the job. You need custom software when at least three of these are true: a tool's licence cost is climbing past what a narrow build would cost, staff are burning real hours on manual workarounds, the data is sensitive enough that you need control over where it lives, the process is how you differentiate from competitors, or you're maintaining fragile integrations between tools that were never built to talk to each other.
Buy, adapt, or build — the three levels
There are three ways to solve a software problem, and they climb in cost and commitment. Try them in order. Only move up when the level below genuinely can't do the job.
| Level | What it is | Rough cost | Best when |
|---|---|---|---|
| Off-the-shelf tool | A ready product you subscribe to | £0–£50/user/month | A tool already covers the job well enough |
| Adapt what you have | Spreadsheets, manual steps, workaround processes | Free-ish, paid in staff time | The job is occasional or still unproven |
| Custom build | Software built for your exact process | £8k–£60k+ | Core job, sensitive data, real cost, no tool fits |
Jumping straight to "let's build something" is where budgets die. Most jobs resolve at level one. A genuine minority ever need level three — and this guide is about telling the two apart.
What are the real signs you need custom software?
Answer: You need custom software when the cost of staying on a tool or a workaround — in licences, staff time, risk, or missed opportunity — clearly exceeds what a narrow build would cost. Five signals do the actual work here.
- Licensing-cost crossover. A niche tool charging £15 per seat per month for twelve people is £2,160 a year, and it climbs every time you hire. Add a per-transaction fee, and a growing team can double that within two years. A narrow tool built for the parts you actually use often earns back its build cost within two or three years, then costs only hosting — typically £20–£40 a month, flat, regardless of headcount.
- The workaround tax. Every manual reconciliation, every double data entry, every spreadsheet someone has to keep in sync by hand is paid time. Add it up honestly over a year and it's usually bigger than anyone admits — because it's spread across small five- and ten-minute interruptions nobody logs, rather than one visible line item. It only grows as the business does, too: double the customers, and most workarounds don't scale linearly, they scale worse.
- Compliance and data control. Customer records, health data, financial or legal information that shouldn't sit in a shared consumer tool, or that has to stay inside your own infrastructure under UK GDPR. A spreadsheet has no audit trail, no access control, and no record of who changed what and when; a proper system does, by default.
- Differentiation. If the process is how you actually beat competitors — your pricing logic, your matching algorithm, your service model — running it on a generic tool means your edge is rented, not owned, and any competitor with the same subscription has access to the same ceiling you do.
- Integration debt. Three tools that were never designed to talk to each other, held together by exports, imports, and someone's memory of which system is the source of truth. Every new integration is a new failure mode, and the debt compounds quietly until one system falls out of sync and nobody notices for weeks.
If three or more of these are genuinely true, the case for a build is strong. If only one is, you're probably still on level one or two — and that's the more common, and cheaper, place to be.
Is there a simple threshold, instead of "it depends"?
Answer: Add up what the problem costs you now — subscription fees plus staff hours on workarounds — projected over two years, and compare it to the cost of a narrow build. If that two-year cost clearly beats the build price, and the job is core enough that you won't have walked away from it in a year, build. If the maths is close, don't — stay on the cheaper level until the problem forces the decision, not until it feels exciting to build something.
The point of naming numbers is so you can sanity-check them yourself before anyone quotes you. Our pricing guide breaks the ladder down by scope, so you can see roughly where a given problem should land before a scoping call.
What does this look like in a real business?
Take a nine-person letting agency managing around 140 rented properties. Where does it land?
The setup: they run a property-viewing booking tool at £29/month, a separate invoicing tool at £19/month, and a shared spreadsheet tracking gas safety and electrical certificate renewal dates for every property. None of the three talk to each other.
Level 1 — does a tool already do this? Partly. The booking and invoicing tools each do their own job well. Nothing on the market ties viewings, invoicing, and statutory certificate tracking together for a portfolio this specific size and shape. Keep going.
Level 2 — is the workaround tolerable? No, not any more. The office manager spends around six hours a week manually checking the spreadsheet against the other two systems, chasing landlords for overdue renewals, and re-entering every new tenancy three times. That's roughly £4,900 a year in admin time alone — and a missed gas safety certificate isn't just an admin slip, it's a breach of statutory duty.
Level 3 — do the signals stack up? The workaround tax is real and growing with the portfolio. The data is compliance-sensitive, with a genuine audit-trail requirement. And the three tools are held together entirely by one person's memory — classic integration debt. Licensing-cost crossover and differentiation aren't really in play here, and that's fine — you don't need all five signals, just enough of them.
Verdict: a narrow internal tool — one record per property, automatic renewal reminders, a single source of truth — quotes at roughly £9k–£12k. Against a workaround costing close to £10k over two years before counting compliance risk, the build pays for itself inside that window, and removes a category of risk a spreadsheet was never built to manage.
Contrast that with a five-person consultancy tracking client proposals in a shared document and a general-purpose subscription tool costing £40 a month. No compliance exposure, no integration debt worth mentioning, and the team would happily walk away from the current process without much pain. Nothing here scores against the five signals, so the honest answer is: don't build, don't even prototype — the £40 a month is doing its job.
What should a first custom build actually look like?
Answer: Narrow. Solve the one job that's actually costing you money, not every tool you own at once. This is squarely an internal tools job — a single tracker, workflow, or reporting layer, not a platform.
The letting agency above doesn't need a system that also replaces its invoicing tool and its booking calendar on day one. It needs the one piece nothing else covers: property records with certificate tracking and reminders. Everything else keeps working as it already does. A first build that tries to become "the one system for everything" is the most common way these projects run over budget and over time — scope creep dressed up as ambition.
Where should you start, if you're not there yet?
Answer: Start cheap, always. Most businesses reading this haven't hit the threshold yet, and that's the correct outcome, not a failure.
Before considering a build, make sure you've actually exhausted level one. Our roundup of tools worth paying for covers what genuinely earns its subscription versus what doesn't. If you're specifically weighing this decision for an AI-shaped project rather than general software, our guide to choosing your first AI project runs the same ladder logic through that lens. Both exist because the instinct to build is almost always premature — proving a job is real and repetitive on a cheap tool first is what makes a later custom build actually justified, rather than a guess.
The decision tree, in order
Run your problem through these questions top to bottom. Stop at the first "yes."
- Does an off-the-shelf tool already do this well enough? → Buy it. Done.
- Is the workaround tax — staff time plus subscription creep — still smaller than a build would cost? → Keep adapting. Revisit in six months.
- Does the job touch sensitive data, sit core to the business, or cost real, measurable money in licences or workarounds — and does that cost clearly beat the build price? → Consider a narrow custom build.
- None of the above cleanly true? → You're not ready to build. Stay on the tool, or fix the process first.
When should you NOT hire us?
An honest section, because it's the one nobody writes. Don't hire Canarlo — or any studio — when:
- An off-the-shelf tool already does the job. We'll say so on the call, in writing, and mean it. Paying for a build to replace a £20/month product is money set on fire.
- You haven't run the process manually long enough. If you can't yet say how often it happens, how long it takes, or what it costs, you're not ready to scope a build — you're ready to keep a log for a month.
- The workaround tax is smaller than the build cost. A £15k system to save two hours a month never pays back, no matter how satisfying the software would be.
- You want "custom software" as a status symbol. Building for its own sake is the most reliable way to burn a software budget. Start from the cost of the problem, always.
A studio worth commissioning turns work away when a cheaper option would serve you better. We quote in writing within a business day, and every line of code we build belongs to you from the start — no retained platform, no ongoing licence to us. That's the same test either way: are we solving your problem, or building our own pipeline.
How to actually start
Pick the one process that's actually costing you money — not the one that would be most fun to build. Run it down the decision tree. If it stops at a tool, buy the tool this week. If it stops at "keep adapting," write down the hours it costs you and check again in six months. If it genuinely reaches the build level and the maths works, that's when a proper scoping conversation earns its place.
If you want that scoping done honestly, our pricing guide shows the real cost ladder before you speak to anyone, and a 20-minute call tells you which level your problem actually belongs on — including if the answer is "buy the tool, not us."