Bias, Diversity, and Profitability

The first history lesson I ever had in secondary school featured two hours on bias. At the time, I found this a frustrating experience…

Ryan den Rooijen
Ryan den Rooijen

The first history lesson I ever had in secondary school featured two hours on bias. At the time, I found this a frustrating experience. Why sit here talking about the importance of different perspectives when there are so many exciting Roman battles to revisit! I still remember the teacher’s comment how, with all these male historians, it was no wonder we called it history, not herstory — jokes, as the word history comes from the Greek historia.

Now that I find myself working in data and analytics I treasure that memory. After all, everything we do is permeated by bias: from the data we collect, to the analyses we conduct, and the insights we derive. Being aware of this bias is critical to not only avoid painful mistakes but to ensure you can extract the most value out of your data. For example, when analysing consumer behaviours it is easy to fall into the trap of assuming people have the same cultural background as you and draw faulty conclusions or miss the insight.

Even simple data like mobile phone usage can be easily misinterpreted. Photo by tokyoform.

Ensuring diversity in your data is also critical, not only to ensure your analyses will be generalisable, but because failing to do so can perpetuate harmful stereotypes as algorithms encode our existing biases. For example, in a study by the ACLU, people of colour who are members US Congress were roughly twice as likely to be misidentified as criminals compared to their peers. Similarly, translations have run into a number of pronoun related challenges.

This is not academic. How many algorithms are trained daily on non-representative data without people realising?

Whether developing insights or building algorithms, having a team with diverse people, personalities, and perspectives can help mitigate these risks. Even if that proves challenging due to recruitment, at least educate people. While usually people do not appreciate their biases being pointed out, research shows that calling them out leads to a reduction in those biases.

Finding realistic, diverse boardroom stock photographs is not easy. Photo by Roy Bisschops.

It should be noted that this is not only about risks. Research has indicated that companies with diverse leadership teams lead more profitable companies. While promoting inclusion, equality, and fairness are obviously the right things to do, this financial aspect is important to highlight as it can help convince those who are resistant to change. At the end of the day, the first step to building more diverse and profitable organisation starts with awareness of the problem and the personal ownership to address it, especially in analytics.

— Ryan

Ryan den Rooijen

Now Chief Ecommerce Officer; formerly Chief Data Officer.