Embrace the errors

Product managers are advised to "fall in love” with the problem, not the solution. Similarly, those working in applied AI, such as ML engineers, should “fall in love” with data. Data is a reflection and footprint of the problem at hand. A thorough understanding of it – features, correlations, missing values – helps solve the problem. Looking at individual data samples inspires more insightful questions that can be then addressed using a large sample size.

Most crucially, invest time in analyzing errors. Errors are the remaining problem to solve. Once you've developed your initial algorithm and backtested it on your data – achieving a 60% accuracy, for example – focus on the remaining 40%. Let’s rephrase it then: fall in love with the errors, so you can bust them.