Personalize or stay behind… the key to great user experience of digital products

Personalization of digital products is becoming the norm, yet we have all had those eye-rolling moments when a brand that should know us completely misses the mark — for example your favorite online shopping platform showing products that would never ever be in your consideration span or even worse, offering us products that we just bought. Oh yeah Amazon, I’d love to buy a second coffee machine the week after I bought a new one. The outcome of such situations is a frustrated customer. A situation that brands today can not afford, as consumers have a wide array of options and will quickly switch to a place where they feel recognized, appreciated and valued.

Personalization is a process of tailoring content to individual users’ characteristics and preferences. It enhances users/customers experience by meeting their needs more accurately, making interactions faster and easier. If done correctly, the personalization process also can shapes users’ future preferences and helps them further identify themselves with the brand.

THE KEY TO SUCCESSFUL PERSONALIZATION: DATA, DATA, DATA!

Personalized digital products deploy a technology based on algorithms that learn from the collected user’s data, and based on that provide a unique experience to them. However, the personalization strategy is only effective if the data informing it is collected and interpreted well. Even the most advanced algorithms can’t work magic if they have incorrect, inadequate or outdated data.

In order to do so brands have to track an individual’s behavior across different touch points and merge that information with customer data from other systems. Once this is done, brands should automate the interpretation of data to determine affinities and intent, create profiles for each person and finally act on data in real time.

Artificial intelligence and in particular machine learning technology have changed the personalization game. They enable processing of vests amounts of data instantly and actively make decisions about what to show to each person. With machine-learning-based algorithms and predictive analytics brands can present the most relevant experience to each and every visitor — and they can do so at scale.

MY FAVORITE DIGITAL ENTERTAINMENT PRODUCTS AND THEIR PERSONALIZATION STRATEGY

1. Netflix

Netflix is a brand that is featured in numerous personalization examples as one of the best practices. I couldn’t agree more. The brand is famous for using an algorithm that’s consistently being developed and improved. Moreover, Netflix uses personalization for different actions — viewing recommendations, search results and even creating content. The whole home page is basically personalized: the banner, carousels, order, artwork (yes, also every artwork), text and search (red marks on the screenshot below). The brand considers: what, how, when and where a title is recommended.

At Netflix they figured that a typical Netflix member loses interest after 60 to 90 seconds of choosing something to watch, after reviewing 10 to 20 titles. This means that Netflix has to recommend the right content fast in order not to loose their interest. The goal of personalization at Netflix is therefore to help their users finding content to watch and enjoy to maximize satisfaction and retention.

Netflix estimates that only 20% of its subscriber video choices come from search, with the other 80% being from recommendations. So it’s essential that Netflix gets this right.

Did you take that in? 80% of users, that is 109.6 million users, make their decision based on recommendations. Mind-blowing..

Netflix gets data by constantly testing and screening millions of their users and instantly improving the algorithms. Whenever they rate, click play, pause, stop watching, search, browse (image below)… Every single action feeds back into what the users will see when they next open the application. In addition to that, Netflix chooses a significant number of people randomly and runs A/B tests constantly in order to get feedback on their product and measures the response. The brand uses these tests for everything from their adaptive streaming and content delivery network algorithms, to redesigns of the UI layout and new personalized homepages. Every product change goes through a rigorous A/B testing process before becoming the default user experience (yes, every change, also images associated to many titles).

In my eyes Netflix is the best example of successful personalization.

2. Spotify

Spotify offers its users personalized recommendations in its Discover Weekly playlist. This is a curated personalized playlist based on tracks the user listened to in the past — the system develops a “taste profile” and identifies songs in other similar taste profiles that could be a good fit. Spotify thus finds songs that could be a good fit, but haven’t been listened to by that user and collects them in a new “Discover weekly” playlist.

What is very interesting in this case is how Spotify is identifying the songs that appear on those lists. The platform uses three types of recommendation models that are based on screening the behavior of its’ users, online texts and raw audio models. They all work separately, but are connected to Spotify’s larger ecosystem.

Similar to Netflix, also Spotify is analyzing the behavior of its’ users, but in this case it measures implicit feedback (the stream counts of the track, whether the users have saved the track to their playlist, visited the artist page after listening etc.). Images below explain how it works perfectly.

In addition to that Spotify also uses Natural Language Processing, which means that it screens online texts about music to figure out what people are saying about specific artists and songs (which adjectives and what particular language is frequently used in reference to specific artists and songs and which other artists and songs are also being discussed alongside).

In order to create new recommendation for its users, Spotify is also using raw audio models, which analyze the raw audio. This model improves the accuracy of the recommendations, but more importantly takes new songs into consideration, that are not considered by the first two models. This is very important for songs that are new on the platform and have, for example, just 30 listens, so there are only a few other listeners to collaboratively filter it against it. The song also isn’t mentioned anywhere on the internet yet, so other models can’t pick it up. With this method also new songs can land on the Discover Weekly playlist alongside popular songs.

By combining these methods Spotify can offer its 83 million users a fresh new playlist called Discover Weekly every Monday morning. A custom mixtape of 30 songs they’ve never listened to before but probably will love, and it’s pretty awesome.

PERSONALIZATION WORKS, WHEN BASED ON GOOD DATA… BUT IT’S NOT EASY

Personalization is a core competency that enables companies to serve each user a tailor-made experience. Users desire personalized services that maximize their enjoyment and minimize annoying search time. Moreover, it can help drive sales, build customer loyalty and help users learn and discover digital tasks easily. The final goal of personalization is an endless and seamless journey for every user. Applying personalization across all relevant channels ensures that consumers are recognized and can continue their journey where they last left off.

Discussed examples prove that personalization works and can effect users’ adoption rate, increase conversion rates, customer life value and finally also marketing ROI. However, basic personalization tactics are no longer enough to engage users. Advanced personalization based on real and well interpreted data is the best way to predict and shape users’ behavior.

Sources:

• Digital Growth Unleashed (The Future of Personalization with AI and Machine Learning)
• Forbes (The Amazing Ways Spotify Uses Big Data, AI And Machine Learning To Drive Business Success)
• McKinsey (Personalization at scale: First steps in a profitable journey to growth)
• Netflix Tech Blog (It’s all about testing, Learning personalized homepage, Artwork Personalization, It’s all A/Bout Testing, Product Integration Testing at the Speed of Netflix)