And the winner for the best picture is…

In my many years in software development, software architecture and support I have developed a few rules that I abide by, one of which is “if the solution is complex, it is probably wrong”. When working for previous employers, I was often frustrated by the convoluted approach to customer problems, so much so I had a slogan (that I chose not to print on my tee-shirt): Making the difficult things possible and the simple things difficult.

My rule of simplicity recognises that most people are like me: average. Violating this principle has caused several nuclear accidents, or planes to crash. It contributed to Warren Beatty and Faye Dunaway incorrectly announcing La La Land as the winner of the 2017 best picture Oscar. It also reminded me of a wonderfully entertaining presentation I received last year in Dubai.

The presenter recognised, that in our world of Revenue Assurance and Fraud, current AI solutions occasionally (understated) run awry. I’ve seen this myself, where a clever SIM Box solution incorrectly learned normal behaviour as interconnect bypass, resulting in most subscriptions being identified as fraud. The presenter proposed that we solve this problem by developing a standard for a monitoring layer that watches the AI layer to identify these errors. Presumably, if this comes to fruition, it too will need monitoring.

With every layer we add, we add complexity. With complexity, we add risk and limit our ability to understand what we have. Simplicity is a mindset. If something isn’t working well, the first thing we should ask is whether we can simplify it, not complicate it. Simple solutions may not necessarily give the most accurate answer, but they often produce the best answer. The best answer is one that is useful, easy to understand and easy to maintain.

One of the contributors to the La La Land error was a complex process involving two people each with an envelope for each category. Their solution: three people and three envelopes!

Adrian Harris
Head of Operations