Otto is the world’s second-largest online retailer in a highly competitive market space. Superior customer experience in terms of higher empathy, relevance, and speed is key to positive customer experience, and this is where AI comes into play. Rupert Steffner explores the cornerstones retailers have to focus when building their customers’ experience on artificial intelligence. It starts with having clear goals and a value system that finds the right balance between customer retention and revenue optimization. Even if AI is hereby built from the seller’s perspective, retailers will need a “good AI” approach that treats consumers fairly and as partners to optimize long-term customer equity.
Technology talent is needed to organize the AI system for implementing agents that receive percepts from customer’s touchpoints and perform actions. Rupert outlines Otto’s system, which is built on a Dockerized microservice architecture and consists of reactive real-time bots working as a little army, and offers a pattern of how to orchestrate the topology. Although far from a generalized AI approach of managing customers in real-time, all dedicated bots are acting on the same belief states by now. The higher the scope of business functions, the more essential the need to act on conformed states to assure decision conformity.
Rupert explains the necessity of implementing coherence between AI’s capabilities and the business functions, perception to build empathy based on behavioral sensor data, and reasoning and learning to manage customer’s risks and opportunities in real time. Rupert then discusses the need for data locality for speed, achieved by implementing in-memory data grids and sharding by customer ID. Rupert concludes by shedding a light on the algorithms executed in real time—ranging from neural networks to random forest. Apart from narrow-task models, Rupert covers how Otto attacks a more generalized model to calculate the real-time conversion propensity along the overall site journey with every click. Analytic models perform within 3–4 ms, which allows for parallel models operating the same task. The design pattern for automated decisioning splits the model execution from the cutoff and the callback/action trigger. This enables a history-versus-real-time negotiation and the proposition of a more generalized next-best action compared to a single-purpose optimization.
Rupert Steffner is chief platform architect of Otto Group’s new business intelligence platform, BRAIN. In this role, Rupert is responsible for the entire setup as well as for initiating and managing the major change projects. Previously, he was head of the Marketing Department at the University of Applied Sciences, Salzburg and worked as a business intelligence leader for several European and US companies in a range of industries from ecommerce and retail to finance and telco. Rupert has over 25 years of experience in designing and implementing highly sophisticated technical and business solutions with a focus on customer-centric marketing. He holds an MBA from WU Vienna.
Comments on this page are now closed.
©2016, O'Reilly Media, Inc. • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • email@example.com
Apache Hadoop, Hadoop, Apache Spark, Spark, and Apache are either registered trademarks or trademarks of the Apache Software Foundation in the United States and/or other countries, and are used with permission. The Apache Software Foundation has no affiliation with and does not endorse, or review the materials provided at this event, which is managed by O'Reilly Media and/or Cloudera.