Easy Personalization Tip of the Week: Anti-Personalization Wild Card

Here is part three in our series:“Easy Personalization Tip of the Week”. If you find these tips useful, we’d love to hear from you!

People are complex. So is personalization

Do you hate being pigeon holed into narrow descriptions? Youthful, conservative, trendy. Mother, hipster, bargain-hunter. Sure, most of the time, you make choices that are consistent with a handful of specific categories or classifications. However, your true self is more complex. You are a multi-faceted person whose tastes and desires are constantly evolving. Just because you buy household items most of the time, you might still be interested in splurging on a pair of leather boots or a dress every now and then.

As individuals we connect with the visceral joy of finding something unique and entirely “unlike” what we normally choose, ideally uncovered in the strangest of places. We cherish those discoveries and are willing to spend time and money on them.

Yet personalization engines used for product recommendations are often the antithesis of such discovery. The measure of most personalization engines is based on how effectively they can “stereotype” a consumer based on their transaction history – and typically, that approach works! In A/B tests, category-based targeting and past purchase-based personalization succeed. They result in higher engagement and conversion rates because the products offered are directly relevant to the consumer.

The merits of occasional anti-personalization

Companies like Costco and Zulily have taken a slightly different approach. Now and then they sprinkle in wild card product options. This means that the consumer mostly sees items that fit her predicted preferences (as determined by past behavior and other consumers who share similar attributes). But occasionally she is shown “wild cards”. These are anti-personalization products, curated to delight and surprise – providing her the experience of discovering something novel and new.

At Zulily, we conducted a simple experiment to help understand the short and long term effects of anti-personalization. We added a section on the personalized homepage experience called “Wild Card Products”. This section contained 3-5 products that were completely different from what the product recommendation services suggested. For instance, if we knew a particular consumer only shopped for children and home, we made the Wild Card Products might show quirky and fun items the Women’s and Men’s product categories.


Here’s what we noticed:

  1. Conversion on these items was low → at a glance you’d think that this would mean the experiment failed. But hold on…
  2. Long term customer engagement (as measured in return visits, time spent on the site browsing across product categories) increased.

We already had predictive models that suggested long term customer engagement was a key driver of higher lifetime value. So the experiment was successful and the practice of including Wild Card Products was incorporated into our daily site experience.


Let us know what you think about Easy Personalization Tip of the Week: Anti-Personalization Wild Card on Twitter.

Signup to receive new blog posts as they're published.