As my affair with data anlytics continues to unfurl, I realize that data and algorithms are naughty and if not handled properly, they are trouble makers. Even those self-claimed data masters have their awkward moments and let me list just a couple that I’ve witnessed recently.
- Amazon and its recommendation system
Admittedly, I fancy some exotic exercise, such as….pole dancing. Seriously, you get some comfortable stretches through practice which also makes you feel more confident about yourself. So I bought a portable dancing pole through Amazon to skip the crowded studio pole sharing scenes. A week later, Amazon started to send me new recommendations based on my “historical purchases and items I’ve owned”, the highlighted feature recommendation was a pair of silver NIPPLE RINGS! As astonished as I was, getting to know there is a type of product called nipple rings on earth, I wondered what algorithms Amazon’s recommendation system used to jump to the conclusion that me a dancing pole buyer would need nipple rings. Seems to me there is a pretty strong correlation coefficient between “dancing pole” and “strippers”.
- Pandora and its customized ads
Admittedly, again. I fancy some 90s-00s Hip Hop and R&B. While I’m playing my classic 90s Hip Hop mix on Pandora, I am forced to listen to this ads that basically urges black women to purchase an in home HIV test as they are a particularly vulnerable group to AIDs. So again, another strong correlation between “female”+“90s Hip Hop” and risk of AIDs infection.
While I write these little incidents out rather light-minded, I feel offended by the inherent stereotypes the collaborative filtering system has been coded in predicting customer needs. By far, algorithm still falls short in recommending real custom fit contents, due to the fact, which I believe, of the gap between understanding the complexity of consumer behavior and normalization of certain segments.
Outliers that we often choose to neglect have strong story-telling power, also a secret key to unveiling a hidden new segment.