In Brief

Coming up with the right sequence of actions and offers at the right time and the right place is now a make-or-break matter for the enterprise. Furthermore, it is not just about you and your customer, the pace of innovation and competition mandates immediate responses to changes and internal agility.
At T1A we identified the lack of solutions that can fully address this challenge and provide marketers with a scalable self-learning solution. And thus we build an innovative product based on Reinforcement Learning which has already allowed one of our customers to achieve a 27 percent increase in campaign response rates.

Assessing What is Used Today

Imagine a major enterprise with a long-standing developed customer loyalty program. A program boasting a wide range of diverse campaigns generating customer interactions, 1–1 personalized offers and rewards. Sounds familiar?
While the major loyalty programs may be sophisticated enough to support an elaborate customer segmentation, most of them are still lacking the mechanics to really optimize the ROI in a true intelligent automated fashion. It is still manual cumbersome expert or statistically driven process which never results in a truly optimal solution across the board. Both approaches has obvious drawbacks.

Hence, a better product should meet all of the following:
● Easy to use, implement and maintain
● Best-in-class ROI given the business constraints
● Automatically adapt and self-learn in a rapid-changing environment
● Blending real-time digital data streams, 1st party offline data and existing expert knowledge

AI Journey of One Retailer

We faced this challenge while working with one of our major customers, a Global Top 50 retail chain from Eastern Europe called X5 Retail Group (think Kroger Co or Albertsons Companies, Inc.).
With a snowballing number of segments and offer types for millions of customers, choosing the most beneficial strategy to increase engagement eventually become a shady game of luck where you bet on analysts gut-feeling.
Our data science team, led by Valentin Marenich and principal data scientist Emily Volynets taken on the challenge to change that. They’ve went on a journey to develop brand-new next-best-action solution using truly innovative approach, that would be easy to use and at the same time be highly adaptable to specific business goals. On one hand, the system was meant to set a new straightforward process for surfacing and scaling insights. On the other, align with a company’s monetary KPIs. In addition, the decision-making process should be fully explainable to business users and C-suite.

We combined our expertise in retail and knowledge of machine learning techniques and developed a unique product that satisfies all mentioned criteria and shows an amazing profit. Reinforcement learning techniques under the hood enable self-learning and ease of deployment and maintenance. Furthermore, unlike most black-box AI products on the market, our system is transparent and interpretable, a perfect example of what is now being called Explainable AI.

What’s Next

Starting as an R&D project, we have managed to beat even the most optimistic ROI estimations, consequently becoming the core CRM analytics tool, used by X5 marketers to automate testing and deployment of successful hypotheses, enabling them to finally focus on business and creativity.
Now, after months of improvements, usability and value testing, we are glad to announce that we would be offering this product to the market. An easy-to-use and deploy solution available for any business trying to bridge the gap between hoarding data and tangible business results through the power of next-gen AI.