Seven products in production: what maintaining them taught us

Most consultancies talk about projects they delivered. We also talk about products we still maintain: Thor IA, Dropium, TaskFlow, Quick Rabbit, Hauzzi, Legnds and Cruxly. Seven, all live today.

The difference is not a matter of pride. Delivering and disappearing teaches you little: you never get to see which decisions age well. Holding something up for years forces you to live with your own mistakes, and that changes how you advise.

The first lesson is the most expensive to learn: cost does not end at launch. A live product consumes attention every month. Dependencies to update, infrastructure to pay for, support to handle, small changes that pile up. Anyone quoting you development only is quoting you half the project.

The second: the simplest architecture that works is almost always the right one. Every extra piece is a piece that can break on a Sunday night. The temptation to over-engineer is strong when the product is new and everything seems possible; you pay for it three years later.

The third concerns technical debt, which is almost never what people think. It is not badly written code: it is a decision you took knowing it would have to be redone. The problem is not taking it — often it is the right call. The problem is not writing down why you took it, because then whoever finds it two years later does not know whether they can touch it.

The fourth: what breaks is almost never the interesting part. It is not the complicated algorithm or the feature that took weeks. It is the integration with a third party that changed its API without warning, the certificate that expired, the queue that filled up because nobody was watching it. Real incidents are boring.

The fifth is the one that changed how we work the most: having our own products changes how we advise. When we recommend a technology, we are recommending something we live with ourselves. And when we advise against something, it is usually because we already suffered it.

That also explains why we insist so much on talking about the aftermath. In a first conversation with a client, a good share of the time goes to questions that are not about building: who is going to maintain this, what happens if it grows tenfold, who answers if it fails on a Saturday. They are not comfortable sales questions, but they save a lot of grief.

If you want to see what each of the seven is about, they are detailed on our real work page. And if you are thinking of building something that has to last, that conversation about the aftermath is exactly where we like to start.