I have a friend who is a pilot and someone asked him “Aren’t most of the controls on the airplane running on auto mode then why do you get paid so much?”. His answer was “Well, I don’t get paid for the 99% of the times when things go right, I get paid for the 1% when things go wrong and in that case I am the difference between life and death.” Sometimes, I like to view my job as a technology leader through that lens – getting paid to avoid the most expensive mistakes. Although thankfully most of the times, I am not responsible for making life and death decisions. 🙂
Now ideally, you would want to avoid all mistakes. But the matter of fact is, that we cannot avoid all mistakes because we lack necessary context, or the skill set or just bandwidth. So whats the next best thing? I like to categorize mistakes by the cost of fixing them. Here is how I categorize mistakes: multi million dollar, a million dollar or a sub-million dollar mistake. These “buckets” come from the cost of the mistakes I have made personally or experienced from my day job of building software, typically for large, enterprise clients.
Now the question is how do we categorize mistakes?
I haven’t spent enough time thinking about the heuristics for these buckets but I have some examples:
- Multi million dollar mistake. Not arranging your software by business concepts would be a multi million dollar mistake. This is the type of mistake that doesn’t instantly hurt and is hard to detect. It dramatically limits the organization’s ability to evolve over time. I bet everyone is doing this wrong, some to a lesser degree than the others and its really important to be on top of this one. This is the one that the organization’s CTO should be actively looking at.
- Million dollar mistake. Not having the right logic in the right layers of your application/service would be a million dollar mistake. For example, your domain logic is embedded in your database and now your processing cannot scale until you scale your database, which means buying more licenses for your database.
- Sub-million dollar mistake. God classes would be an example of sub-million dollar mistake. This is where there most of your business features depend on one class in your code base. It is very expensive to add new business features or maintain existing ones.
The categorization of the mistakes provides the necessary focus on the ones you should be avoiding – the most expensive ones and the rest you can choose to defer or deal with at the “last reponsible moment”. Avoiding the multi million dollar mistakes takes careful planning.
Going back to our examples,when looking at a particular software, you should be focusing on if the software is arranged by the business concepts (where Domain Driven Design thinking can be a huge help), before you get too deep into the weeds of the code and solving for God classes. If you spend your time wisely avoiding the multi million dollar mistakes, you could still end up with some God classes but at least it won’t be as expensive to fix them.