Musings from a Data Scientist After The Personalization Summit

Couldn’t make it to The Personalization Summit in Boston last week? Read on for my takeaways and insights from Evergage’s event. This year’s theme? To demystify machine learning.

The blender is a staple of any modern kitchen. Fundamentally, it’s a simple concept. It’s a set of blades that rotates quickly to crush, slice and mix. However, if I asked you to build one from scratch, I bet confusion and chaos would ensue. Alternatively, if I asked you to make a smoothie, you could figure out how to put the food inside, turn a few knobs, and hit the blend button.

Evergage’s event, The Personalization Summit, was an immersive experience in personalization tools, techniques, process and strategy, but the most exciting thing was the realization that Evergage has created the personalization “blender” for companies to use on their websites. While others in the industry may push a one-pixel solution to personalization, Evergage has invested in the knowledge and expertise within your business by empowering you to create and manage your own personalized experiences.

The one-day event was a great experience, and below are a few takeaways that stuck with me.

  • Just like the blender, personalization is about mixing the right ingredients and recipes.

An ingredient is a type of data relationship such as co-buy, co-browse, or trending. Those relationships are constructed using machine learning algorithms, such as association analyses, collaborative filters, clustering methods and decision trees. The beauty of Evergage’s approach is its ability to surface these relationships in a clean user interface for business owners, who can combine any number of these relationships, apply filters and boosters to create their “recipes.” Combined with a robust user profile, business owners can mine insights about their online visitors to target recipes for specific user profiles.

Having led a team of data scientists building recommendations for a Fortune 50 company, I am particularly interested to see how Evergage continues to evolve it’s “white-box” approach. The platform could serve as a fantastic interface for data scientists within organizations to build more advanced ingredients and serve those relationships to the business owners to create and manage more recipes.

  • Organizations are really starting to take ownership of their personalization strategy.

This is an exciting time for personalization strategy. Companies have talked about the pursuit of personalization for the past few years, but they’re just now able to begin actualizing their goals.

I attended a few e-commerce sessions because I work with several similar clients, and the consensus was how personalization extends beyond just recommendations. Companies can tailor their use-cases for their businesses, whether it’s banners, messaging, product content, individualized home pages and beyond.

  • The complex algorithms that fuel personalization quickly gather enough intel to fuel action.

To get super-technical on you, Evergage’s foundational algorithm is predicated on Collaborative Filtering with Alternating Least Squares. Most websites collect enough data to train the Collaborative Filtering in a short period of time – which is great news. The model fit— a simplified representation of reality— includes a typical train/test split utilizing K-Folds Cross Validation, a highly scientific method of validating random sampling to produce a single estimation.

  • It’s worth it to take the time to build the right framework for your personalization program.

This wasn’t new to me, but it’s incredibly important – especially for businesses that are just starting up their personalization program.

To stick with the food analogies, imagine that you’re hosting Thanksgiving dinner, and you want to help create a memorable experience for your family. You have to plan for how many people are dining with you, what dishes you want to make and where everyone is going to sit. You might not add stuffing to the menu if your family is gluten-free or you might opt to make a family member’s favorite dish. You don’t start each dish at the same time – you prioritize your cooking based on how long each item will take. Building the framework of your personalization program is similar.

First, companies should start with a planning and prioritization process. That process should be aligned to business objectives for targeted channels, customer segments and points of interaction. Finally, companies should establish key performance indicators (KPIs) after determining campaign objectives because those KPIs measure if you’re making a dent in those goals. With Thanksgiving, the KPIs are always empty plates and full bellies.

  • Testing is an imperative part of the experience iteration process.

The Evergage platform comes with a built-in testing suite that utilizes Bayesian statistics, which is a theory in statistics that uses data to assign probability to predict outcomes – and continues to update those probabilities as new evidence comes into play. This statistical theory allows for data to be used to predict which online experience is going to be best for you, and the experience changes with more available data.

However, it’s a theory for a reason, which is why testing is so important. As an analyst, it’s important to consider the introduction of bias and how that comes into play for your personalization campaign both before and after a campaign launches.

Of course, with ease-of-execution comes a world of possibility. At Brooks Bell, we’re excited to partner with Evergage. Together, we offer a cutting-edge personalization platform and the expertise in optimization and experimentation to help enterprise clients’ answer questions about how to integrate and evaluate effective personalization strategies into their experimentation mix.

 

Aaron Baker, Data Scientist
As a results-driven data scientist, Aaron brings a wealth of knowledge in statistics, advanced algorithms and their role in experimentation to Brooks Bell’s analytics team and enterprise clients.  Before joining Brooks Bell, Aaron worked for such top brands as Lowe’s, Eastman and Hanes.  A true NC State fan, he holds a bachelor’s degree in in engineering and master’s degree in analytics.