SO1’s Research Wins EHI Retail Science Award 2020
Science and research have always been at the core of all our endeavors. We collaborate with prestigious universities such as MIT, Chicago Booth and ETH Zürich to devise truly state-of-the-art technologies and funnel them into retail practice. As a result of this fundamental research work, our Chief AI Officer, Dr. Sebastian Gabel, won this year’s EHI Science Prize in the category “Best Dissertation”. And we think that’s a pretty big deal!
This prize is only awarded to the best out of hundreds of retail researchers and scientists across Europe. The jury consists of high-profile industry and science experts from companies and institutions such as Ikea, Edeka, Douglas, EHI Stiftung, GS1, WHU, and more. The award was handed out on February 18th at EuroShop 2020 – Europe’s largest retail fair.
So what is the work about? Sebastian’s dissertation describes deep neural networks (DNNs) and innovative ways to enable those for “1-to-1 Marketing in Grocery Retail”. These DNNs are also used in the SO1 Engine, and explores many innovative ways to apply them. Keep reading to learn more:
“When we talk about grocery retailing, we’re talking about lots of data. Hundreds of categories, thousands of products, and millions of consumers purchasing 50 or more different items every week. That data contains plenty of exciting insights”, explains Sebastian.
“At the same time, it is very hard to find the relevant signal in the noise that data always contains, so fully utilizing and extracting consumer-level insights from it is a big challenge.”
“Most of today’s recommender systems fail, because they still have to rely on time-consuming manual analyses based on simple rules, assumptions, and human interpretation, and are thus very limited in terms of personalization, efficiency and impact. They might be able to consider 2-3 different store types, look at high-level product categories, and manage a handful of customer segments. But that’s a far cry from “one-to-one marketing“, Sebastian comments.
The deep neural networks (DNNs) used in the research reveal complex and valuable insights from raw retail data, such as shopping baskets. These DNNs are based on fundamental marketing theory and learn autonomously, without requiring any human intervention.
In the first part of his dissertation, Sebastian shows that AI is capable of mapping market structures for large retail assortments from the perspective of a customer, simply by observing the co-occurrence of products in shopping baskets.
But why is this so important in retail and in one-to-one marketing?
“Automatically understanding market structures from a customer’s perspective is the key to efficient marketing personalization. Just think about a couple of examples, like understanding market structures to manage substitution, i.e. to identify the one yoghurt among hundreds of competing options that the consumer actually likes. Or to use insights on market structures in bundle promotions, by directly or indirectly promoting complementary products such as pasta and tomato sauce to increase revenue”, Sebastian explains.
“But there’s more, think about cross-selling. If you understand how consumers view products and markets, you can recommend products with a similar use case from across multiple different categories, such as Italian food, vegan products or products for school children that include small juice boxes, bologna in the shape of a teddy bear or Disney yogurt.”
“More examples of using the insights on market structures include helping retailers to redesign shelves and aisles, identify redundant products, and support brands to customize their market positioning to better fit consumers’ perceptions.”
Sebastian’s paper has been also acknowledged by the well-respected Journal of Marketing Research, which published it in August 2019. You can examine a detailed description of the approach in the full scientific paper or the summary on our blog Let the Data Speak: Revealing Market Structures in FMCG Retail.
Sebastian is as passionate about bringing new methods to industry as he is about academia. While the first part of the work is more theoretical, aimed at business intelligence, the second part is about a revolutionary recommender system that personalizes offerings for each customer individually. You might have guessed it – we’re talking about the SO1 Engine and its business applications at some of the major grocery and drug retailers in Europe and the US.
While the system was originally developed for automated promotion personalization, the research work by Sebastian proves that it is not just a black box and that there are many ways to extract valuable business intelligence.
Winning this prestigious award is a great achievement and a well-deserved acknowledgement for many years of hard work. But of course what’s most important is how the technology actually performs in the real retail environment. If you want to learn more about its business applications, visit our website.