Jan 25, 2018

ASS Pyramid (Approach to Solving Stuff)

You probably understood from the caption that this article would be naughty. In fact it’s purposefully provocative and you may even want to smack it. It originated in a very similar nature. Our Head of Research, Roman Rožník wanted to smack a lot of people.

Imagine this: you are walking around the office and a huge problem jumps at you from behind the corridor. You want to tackle it with all your IT force, but the problem is new and you don’t know much about it. It might appear bigger than it really is or vice versa. So far, the only thing you know is that you are tackling an illusion of a problem.

That’s where our Approach to Solving Stuff (ASS — terminus technicus) comes into play.

It’s good to have a variety of options on how to solve the beast. Terms such as predictive modelling or machine learning come to mind. And these are okay, as long as there are simpler options preceding these two. The picture below illustrates the cost and complexity of problem solving.

At the heart of the “ASS Pyramid” (no, it’s not the girl’s ass), there’s a very basic function of the cost of each approach. We like to start at the bottom when considering which technique to use. Often, you only need the animal part of your brain to find a solution that will bring you the majority of the benefits. Every step up requires more resources and further complexity in this order:

  • Pen and paper.
  • Clean database and a skilled analyst.
  • Clearer, structured big data ready for production, data scientists, capacity to put your ideas into production, skills to understand what the ML is doing and an ability to explain it to others.
  • A very strong AB testing framework, valid data, people who understand how AB tests can be biased and the countermeasures to this, a strong culture and a very, very strong belief that you know what you’re doing. Here’s a nice video explaining how they do it at Pinterest)

Choosing the simplest option has many pros including: understanding the problem, understanding the reality you operate in, massive savings and avoiding a lot of pain with the complex solutions that can be found.

I advise you to follow the simplest approach if you want to live a happier life and guide your company to its full potential.

I feel that this principle can reap great rewards — even if it seems trivial and very basic. And when you use it, you can still come up with many tasks which you’ll be able to use machine learning on, no matter what business you’re in. How come? I’ll tell you in the next episode 😉


Andrej Makovický, Head of Search

Featured articles
Generating SwiftUI snapshot tests with Swift macros
Don’t Fix Bad Data, Do This Instead