Simplify, then add Lightness

Before I got into software engineering, my studies were in other engineering disciplines. One of my idols in engineering is Colin Chapman, of Lotus; he would operate on a fraction of a budget of those around him, and at them all through creativity and thinking outside of the box. Even now, 50 years on, his influence and innovations are still visible in Formula 1 car design. The quote “simplify, then add lightness” was one of Chapman’s responses when asked how his cars were so successful – it’s a quote I really like, because it applies very well to other fields of engineering, in particular software.

At the time when Lotus entered Formula 1, the sport was dominated by big heavy powerful cars. The Lotus design philosophy was very different; they took the approach of trying to build the simplest and most nimble cars possible, recognizing that the biggest gains in performance would come from removing weight (even if that meant less power) and improving cornering and acceleration. Another of his quotes on this is “Adding power makes you faster on the straights. Subtracting weight makes you faster everywhere.” In software there are two main ways we can apply these principles; the first is in how we organize ourselves, and the second is in the design of the things we build.

Simplify…

Simple designs for a racing car offer a huge advantage over complicated rivals, because they are easier to maintain. A complicated design has more things that can go wrong, and when they do go wrong it is more difficult to design and fix. This should be painfully familiar to anyone working in software engineering, and yet as a whole we keep producing over-complicated solutions. Perhaps part of the problem is that our industry has a lot of very clever people, and building complicated things can be kind of fun, and a way to show off people’s ability to build these things. Perhaps it stems from a lack of understanding, or poor focus on the underlying need. I often see relatively simple business problems that have a disproportionately complicated software solution, and the results are never pretty. The ability to extend a complicated solution is much lower, and the cost of adding these extensions rises to the point where it becomes no longer viable.

Things like the Clean Code movement go some way to addressing these problems at a low level; the main principle behind Clean Code is to write code for human readability with a strong focus on clarity in each individual module. This is a great start, but it’s still common to see well-applied attention to detail at this level, but the big picture is a complicated overall system design. Approaches like Microservices are one way that systems design can be designed to have smaller and simple components, but where it’s very easy to make big mistakes in the overall system design and create a mess by over-engineering and introducing boundaries that aren’t appropriate. I see many teams who have built a complicated monolith, and believe the solution is to build Microservices. They apply the same design thinking, and end up building a set of complicated “Microservices” that are coupled together, and actually worse than the monolith because all the challenges of distributed computing have been added with none of the benefits realized. The thing that is lacking in this case is often the ability to strip away complexity, and identify the simplest possible solution that could work. As it happens, the simplest possible solution that could work often turns out not just to work, but to be a very good solution.

Equally, quite often one of the most effective ways to build a highly performant solution is to look for simple and elegant designs. The worst performing systems I see are often full of entire components that are not needed, and rather than optimizing them, the solution is to remove them entirely, re-routing the application into a simpler format, which is less prone to bottlenecks.

… Then add lightness

The big benefit that a lightweight car had over its rivals was its rate of acceleration, and ability to change direction. The same is true of a lightweight team structure; the ability to build and change momentum quickly allows a small team to outperform a larger team, where the overheads and weight that come with that size mean that a larger team will often deteriorate in performance as process and bureaucracy – ironically, the things that Agile methodology advocates removing, but that creep back over time unless kept in check – take their toll. The big front-engine cars that the Lotus design philosophy made obsolete are the equivalent of the software team fallacy that is adding more and more people to a project in an attempt to increase delivery speed by adding “power” – this normally has the opposite effect. The concept of a “two-pizza team” is fairly well known; that is to have a team size that can be fed by two pizzas, i.e. 5-6 people. A team with less people that that can still be successful, but as a team grows above that in size there is an argument for breaking the team down into smaller units. For me, a nicely balanced 2-pizza software team could be 3 developers, 1 domain expert/business analyst who works as part of the team, and 1-2 QA/test specialists.

Another way we can improve the health of software delivery by “adding lightness” is in the judicious and ruthlessly prioritized choice of features. There is a trend emerging that simple well-designed tools that do one job but do it very well are winning over bigger more complicated tools that try to do much more – look at things like Slack vs email, mobile apps vs software suites, Trello vs Microsoft Project, and countless other examples. There is a lot of value in delivering a focused set of features that complement each other, and that value sharply decreases as features get away from that original targeted selling point. Software projects that have a multi-year roadmap of features with no clear priority are much more likely to fail by simply delivering the wrong thing, where they might succeed by delivering focused value and responding to customer feedback. It’s also very important to try and add lightness to the end user’s experience. The features that resonate the most with customers are often the ones that simplify some cumbersome workflow, or remove friction from a task they are trying to perform.