The build that worked perfectly and only one person used

Everything shipped. Everything worked. Adoption was one person — the owner, who had paid for it.

That is a failure. It does not matter that the software did what the specification said. Software nobody opens is a very expensive PDF.

How it happened

The discovery was too thin. I spoke to the person paying, who was enthusiastic and full of ideas, and I let enthusiasm stand in for evidence.

Then the scope crept. Someone senior added features mid-build, each individually reasonable. By the end we had something powerful and heavy, aimed at a team who mostly needed to do three things quickly on a phone, in a van, with one bar of signal.

Nobody from that team had been in the room. Not once.

What I actually learned

The buyer is not the user. When those are different people, the user has to be interviewed or you are guessing. A director describing a field engineer's day is a translation, and things get lost in translation.

Scope creep is a delivery risk, not a revenue opportunity. Every mid-build addition was more money, and it was also the thing that sank adoption.

You need to subtract. The recovery was not more features. It was cutting it back until it did less, faster, on a phone.

Why I am writing it down

Because the case studies you read never include this, and it gives people a false picture of how builds go. The difference between agencies is not whether things go wrong. It is whether anyone tells you when they do.