Build or buy: find the one piece that's actually missing

Building your own version of software you already pay for feels like saving money. Usually it means you just hired yourself for a new job: building it, fixing it, and updating it every time something changes. The first working version isn't a lifetime maintenance plan.

1. Write down the job, not the tool. One sentence about what needs to happen and who does it. "The reviewer sends a bad appointment back to the agent with a note" is a job. "We need a better chat app" isn't.
2. Try the job in what you already pay for. Actually do it, start to finish, in the tool you have. If it works, you're done. Keep paying and go do something more fun.
3. Name what's missing in one sentence. Not "it's clunky." The specific thing it can't do. If you can't say it in one sentence, you're not ready to build anything yet.
4. Ask who looks after it. When something changes, who fixes it? If the honest answer is "me, at night," put that on the scale too.
5. Build only the missing piece and connect it. That's what we did. Our reviewers needed to kick a bad appointment back to the agent. Slack wasn't built for that. So we built that one piece, and Slack gets to keep being Slack.

And honestly? Left alone, I'd rebuild the whole internet by Friday. This question is how I decide what to buy and what to build.

Paste this into a note and run it before any build:

BUILD OR BUY: ONE QUESTION FIRST
What's actually missing?

The job (one sentence): ______________________
Who does it (job title): _____________________
Can the tool we pay for do it today?  Yes / Partly / No
What's missing (one sentence): _______________
Who fixes it when something changes: _________

Decision
[ ] Keep paying. Nothing's missing.
[ ] Build only the missing piece and connect it.
[ ] Can't name the gap yet. Don't build. Look again.

Coastline Roofing pays for a team chat app and a scheduling tool. The owner, Dana, is tired of two bills and wants one custom app that does everything.

She runs the check. The job: when an inspection gets booked without the roof's age, the office manager sends it back to whoever booked it, with a note. Can the tools do it? The chat app sends messages fine. The scheduler books fine. Neither one sends a booking back with the reason attached, so the note gets lost in a chat thread.

What's missing: a send-back button on the booking that carries the note. Who fixes it: their web developer, a few hours a month. Decision: build the button, keep both subscriptions. Dana still pays for chat. Dana also didn't accidentally start a software company.

• Starting from the bill. Canceling a subscription isn't the goal. Getting the job done is.
• Rebuilding the whole tool for one missing feature. You needed a button, not a new chat app.
• Skipping "who fixes it." Every build comes with updates, breaks and changes. Somebody owns those.
• Deciding the tool can't do it without trying. A few minutes in the settings beats weeks of building.

Because there's no trophy for canceling a subscription and accidentally hiring yourself for a full-time job.