This is a series derived from my popular talk “The Seven Righteous Fights”. It’s not an exact transcript, but I think having a written form to refer people to greatly increases accessibility and gives me room to expand my thoughts in a way that is not compatible with speaking.
There are seven fights that I have over and over again, whenever I start at a company. The more software you build, the more it’s obvious that there are seven fights that it will alway pay to have.
My background as a writer means that I am trained to perform a lot of analysis and synthesis. If I don’t understand the product from value proposition through implementation, it means that I am not the writer I want to be. I call it being a full-stack writer. The problem with having all this training in analysis is that you can’t exactly turn it off. I don’t stay in my lane well, I can’t notice problem that only fall into the narrow category of “documentation”. They’re all documentation problems, they’re all software problems, they’re all design and usability problems.
I’ve worked for roughly 15 companies and my conclusion is that when we create something new, we tend to make similar mistakes every time. This is probably because we all tend to settle in a comfortable stage of product development. Some people prefer mature products, some people like the very beginning and hate the mature product maintenance. But that preference means we don’t understand pain points that happen at other stages.
I’m an early-stage person. I love writing the first version of documentation. I love that green field and that new product smell. But I have done enough mid-stage stuff to know the emergent consequences of early-stage decisions. And like everything you’ve ever read about compound interest, early mistakes compound until they are enormous.
I was working with a startup that was pushing super hard to get their product ready to demo at a big industry conference. They were stripping off every feature they didn’t absolutely need in order to make sure they made their deadline, and I was helping. We made the deadline, there was documentation, and they got a POC contract… with a company in Brazil.
In the process of rushing to get ready, we had stripped out all of the plans to make the interface labels link to a file we could localize, and we had hand-coded all the labels. Now instead of handing off a single file to be translated into Portuguese (and having standard labels that could have been done with Google Translate, in a pinch), we either had to branch and hand-code Portuguese labels or spend twice as long digging into the original code and fixing what we had hacked. I’d love to tell you we decided to do it right, but you all know too much about software to believe that happened, don’t you?
The compounded interest from not using a best practice of referred labels when we built the user interface cost us five days of developer time to fix the wrong way, and two weeks to fix and rebuild the right way. That’s a relatively common example of things people skip and pay for later.
Leaving important considerations “until we get a working prototype” means that it’s going to be harder and more expensive to retrofit. Having these fights early prevents you from doing the software equivalent of poking chocolate chips into already-baked cookies.
If you’d prefer to watch the talk as I delivered it:
Shortest recorded version: At The Lead Developer 2016
Longest recorded version: At SpringOne Platform 2016
O'Reilly Software Architecture/Velocity
The Lead Developer London