Picture of Felipe Gil Written by Felipe Gil
on December 03, 2019

When implementing personalization, a necessary first step is setting up a recommendation system to show the best information to each client at the best time, across all digital channels. The number of clicks is a good indicator of how well you are doing. It will also tell you if there are aspects of what you are displaying that could be further personalized.

An article published by The Netflix Tech Blog suggests that a good approach would be to consider the artwork or imagery used to display different possibilities. To Netflix, this translates into showing different users different thumbnails for the same movie or series. The trick is to shuffle different options to automatically test what works for each user, and based on that, decide what to show in the future to increase click probability.


The Netflix team spent much effort trying to find —through multi-armed bandit algorithms— the single perfect artwork for each of their titles, the one that would earn most plays from the largest portion of their users. And then changed to a much more effective approach. Aided by machine learning algorithms, they started to look for the best artwork for each of their users to highlight the aspects of a title that are specifically relevant to them. A meaningful personalization of artwork should be based on each user's different viewing history, for example, how much they prefer different genres and themes, or even cast members. For example, for the movie Good Will Hunting you might see artwork containing Matt Damon and Minnie Driver, because you've watched many romantic movies, while someone else would see artwork containing Robin Williams, a well-known comedian, because that person is a fan of comedies.

Personalizing artwork is not always clear and obvious, and using hand-derived rules is not an option. According to Netflix, a wide and complex experience that relies on data is the best way to know what aspects to use for each user — for example an artwork that not only highlights different themes in a title but also has different aesthetics— further tailoring the experience.

Personalizing the look and feel of digital experiences can result in meaningful improvements in the way clients relate to products. For Netflix, this meant catching the attention of users in ways that benefit them, as well as the company. By developing solutions in a flexible manner, different user profiles can be accommodated, resulting in a more enjoyable experience. Given their complexity and the sheer amount of ingenuity that Netflix invests in every aspect of their platform, this example can seem overwhelming.

But even if you are not Netflix, you can still play around with the idea of personalizing artwork.

For example, if you are offering a new credit card, how can you use your existing data to personalize a banner on your client's online banking? Wouldn't it be great to know who prefer traveling, shopping or home improvement and show a different artwork accordingly? Maybe you are not there yet, but as they say, a thousand mile journey start right under your feet. Ready for that first step?

PrismaCampaigns_Separador1

Get started in the journey to harness the endless possibilities enabled by data collection and algorithmic customization of the user experience.

The data you need is already there

 

Image credit: shutterstock.