tunities for email campaigns. In one case, Pinterest sent one that said:
You’re getting married! And because we love wedding planning—especially all the lovely stationery—we invite you to browse our best boards curated by graphic designers, photographers and fellow brides-to-be, all Pinners with a keen eye and marriage on the mind.2
The problem: Plenty of the recipients of the email were already married or not expecting to marry anytime soon.
Pinterest did act quickly and put out this apology:
Every week, we email collections of category-specific pins and boards to pinners we hope will be interested in them. Unfortunately, one of these recent emails suggested that pinners were actually getting married, rather than just potentially interested in wedding-related content. We’re sorry we came off like an overbearing mother who is always asking when you’ll find a nice boy or girl.
It’s an important lesson. Even some of the most tech-savvy companies blow it.
For example, there are some cases where the data may be spot-on but the outcome could still be an epic failure. Consider the case with Target. The company leveraged its massive data to send personalized offers to expectant mothers. This was based on those customers who made certain types of purchases, such as for unscented lotions. Target’s system would create a pregnancy score that even provided estimates of due dates.
Well, the father of one of the customers saw the email and was furious, saying his daughter was not pregnant.3
But she was—and yes, she had been hiding this fact from her father.
There’s no doubt that data is extremely powerful and critical for AI. But you need to be thoughtful and understand the risks. In this chapter, we’ll take a look at some of the things you need to know.
Leave a Reply