Today, we are pleased to announce two exciting new capabilities: Evergage Decisions and Evergage Data Science Workbench. Both capabilities represent major advancements in the way Evergage customers can use machine learning to better understand their audiences and deliver 1-to-1 personalized experiences across channels.
Building on our existing machine learning capabilities, Evergage Decisions enables B2C and B2B companies to deliver the most relevant promotions, offers, hero images – or complete experiences – to individual website visitors, application users and email recipients. The first algorithm being released under the Evergage Decisions umbrella is Contextual Bandit.
Contextual Bandit uses machine learning to evaluate both the probability of engagement at the per-person level and the revenue opportunity or synthetic value (for non e-commerce use cases) at the business level, and then presents the optimal promotion, image, offer or experience. For example, if a company has 15 different homepage hero images it could potentially show a visitor, the algorithm considers all the data available about the person, factors in the value of someone converting on each image (clickthrough, offer acceptance, purchase, etc.), makes the best choice for that person at that point in time, and then presents the image – all in milliseconds.
Contextual Bandit provides tremendous value not only by optimizing what to show different people but also by freeing up time for business users to focus on creating powerful messaging and offers, rather than trying to manually match each offer to each audience segment (the best most businesses can do today, since manually matching at the 1-to-1 level is impossible). Thanks to the advanced machine-learning capabilities of Contextual Bandit, companies can now let the system determine and render the optimal offer each time, for each person.
Evergage Data Science Workbench
The Evergage Data Science Workbench, which is part of our recently announced Evergage Gears framework, gives data scientists a way to access and work with the rich data stored on Evergage servers.
With Evergage’s Data Science Workbench, companies can access a treasure trove of rich customer data. For every visitor, customer and account who engages with your website or mobile app, Evergage maintains a unified customer profile where engagement details as well as a contextual understanding of a visitor’s behavior is stored. Profiles may also include explicit data collected from surveys and attribute data passed from external sources into Evergage’s customer data platform (CDP).
Users of the Data Science Workbench are provided with a dedicated cluster, where they can access Evergage’s data through a safe and secure read-only proxy. The cluster is pre-installed with a suite of familiar tools which run on top of Apache Spark. Apache Zeppelin provides a notebook in which Python, R and Scala can be used together, sharing data across languages. Once set up, your data scientists can run data transformations, numerical simulations, statistical modeling and data visualizations on source data in the same way Evergage’s own data scientists do.
Not only are your data scientists able to create custom models using the Evergage-housed data, they are also provided a pathway to integrate those learnings back into Evergage. This occurs by uploading the results of your modeling as custom attributes. For instance, you could flag particular customers as “high value,” “seasonable buyers” or “likely to churn” based on your analysis.
To learn more about Evergage Decisions and the Contextual Bandit algorithm, watch our on-demand webinar, Contextual Bandit: Advanced Machine Learning for Selecting the “Optimal Offer.” In the webinar, Cliff Lyon (VP of Engineering at Evergage) and Jordan Bentley (Senior Data Scientist at Evergage) discussed and demonstrated the power, benefits and uses of the new algorithm.