Incorporating Intent Propensities in Personalized Next Best Action Recommendation
Incorporating Intent Propensities in Personalized Next Best Action Recommendation


Next best action (NBA) is a technique that takes unique user history and characteristics into consideration and recommends the next actions that help the customer progressing towards business goals as quickly and smoothly as possible. It is not easy to design such a AI powered NBA engine. Ideally a hand-free NBA engine needs to handle the following problems. a) It should deal with incomplete historical feedback that are skewed towards a small set of actions; b) It should adapt to dynamic actions, which can be added or removed frequently due to seasonal changes or shifts in business strategies; c) It needs to optimize for multiple complex business objectives, which usually consist of reaching a set of target events or moving users to next more preferred stage; d) Most importantly, it has to learn and make decisions in realtime and at massive scale. Most of the solutions in the market only addresses a few of those challenges, due to model or technical challenges. In this presentation, we will show how we address all those issues at Salesforce Marketing Cloud Einstein. We will present a hybrid model based on reinforcement learning that balances both online and offline learning. We will show how we utilize distributed big data processing technologies and services to train and make predictions at massive scale. We will also discuss an offline evaluation mechanism to provide bounded expected performance, which has been a hard problem for reinforcement machine learning in general.


Kexin Xie is a Senior Director of Data Science and Software Engineering at Salesforce, responsible for data science research, practices, and computing architecture for Marketing Cloud Einstein. He leads the team enabling large scale AI initiatives, processing over 7 billion monthly unique users, and making tens of trillions of weekly predictions. Before Salesforce, Kexin has worked in different sectors building large-scale data science platforms that perform machine learning, data mining and online/offline analytics in the space of data management, real-time bidding, intelligent marketing, anti-fraud and anti-money laundering. He holds a Ph.D. in computer science and has publications in top tier journals and conferences like ACM TODS and VLDB

Privacy Settings
We use cookies to enhance your experience while using our website. If you are using our Services via a browser you can restrict, block or remove cookies through your web browser settings. We also use content and scripts from third parties that may use tracking technologies. You can selectively provide your consent below to allow such third party embeds. For complete information about the cookies we use, data we collect and how we process them, please check our Privacy Policy
Consent to display content from - Youtube
Consent to display content from - Vimeo
Google Maps
Consent to display content from - Google