Sometimes the right answer is not to build anything

We build custom software for a living, and we regularly finish a first call by recommending that the business doesn't build anything. It's the most useful hour we offer, and this post walks through what happens in it.

Before anyone gets to "should we build or buy," there's an earlier question, which is whether fixing the problem needs new software at all. Often it doesn't, and each of the options below costs less than the project you called about.

A setting in software you already have

This is the most common one. It might be a feature in a product you already pay for that nobody ever set up, or the approval steps in Microsoft 365 that nobody switched on. Business software comes with far more than most customers ever use, and "we need an app for this" sometimes turns out to mean "we need an afternoon in the settings of the tool we already have."

An off-the-shelf product that fits

If the way you work matches what a packaged product expects, buy the product. Custom software pays off where your process is unusual, and plenty of processes aren't. Bookkeeping, payroll and help desks usually work the standard way. We keep straightforward pages about the products we get compared with, and every one of them starts with the reasons to stay where you are, because those reasons are real. There's a fuller way to think it through in custom vs off-the-shelf. Briefly, your way of working has to be both different and valuable before it's worth building software around.

When the real problem is the process

Software automates whatever you point it at. If the process is a mess, you get the same mess faster, with a login screen. When jobs get lost in the handoff between sales and operations, the first question is what that handoff should look like, and only then which app should carry it. Sometimes fixing the process is the whole fix, and the software conversation ends there, several hundred thousand dollars earlier than it might have. A good test is whether anyone can draw the process working properly on a whiteboard. If nobody can, no system can be built to run it.

Not yet

Custom software is a fixed cost that pays for itself through hours saved and mistakes avoided, so below a certain size the numbers don't add up. A three-person office losing two hours a week to re-typing should keep the spreadsheet and look again in a year. The calculator on our site will happily show you a negative number, and that's deliberate. "Not yet" is a real answer, and a calculator that could never give it would just be an advertisement.

A one-off job for a freelancer

Our plans are a 12-month partnership, because the systems we build are ones a business depends on, and someone has to keep them patched and improving after launch. A small job that's finished once it's done, like a calculator, a form or a one-time data cleanup, doesn't need that, and paying for it would be a waste. A good freelancer will do it faster and charge you less. We say so on our own homepage.

Why would a software company tell you this?

We work in year-long engagements with a small number of clients. A project that shouldn't exist leads to a year of regret on both sides, a client who won't recommend us, and a system that gets abandoned with our name on it. Doing this work well means saying no early, which is why "we'll tell you not to build" is written into our description of how we work, and not only into a blog post.

How to use this on any first call

Whoever you talk to, ask them to argue against the project. Something like "what would make this a mistake?" works well. If a firm can't name any situation where they'd turn the work away, they had probably decided to say yes before you called. And bring them the business problem rather than a list of features, because a feature list skips the questions above, and those are the ones that save you money.

There's a companion post for when the answer is yes: the signs you've outgrown the spreadsheet. If you're not sure which side you're on, the hour is free, and "don't build anything" is a real possible outcome of it.