Improve Customer Relationships with Predictive Analytics

Digital platforms have given consumers a vast array of products and services to choose from. It can be a daunting task to sift through the seemingly endless number of choices. Organizations that can predict which products best match the needs of their customers not only delight them, they also succeed in driving more profitable customer relationships.

Our proprietary Recommender System enables you to connect your customers with the right items in your large-scale inventories (tens of thousands to millions) of products or content. It is customizable to your unique business needs.

  • Proprietary methodology
  • Customizable to unique business needs
  • Optimal selection from multiple recommenders
  • Tuned for real-time operation
  • Built for scale
  • Continuous model improvement through integration with Test & Learn

Antuit’s Recommender System can integrate real-time feeds from multiple structured (ERP, point-of-sale data, in-store transactions) and unstructured (search engines, clickstream, web, mobile, emails, call center logs) data sources, and analyze them to arrive at the most relevant product recommendation for each targeted customer interaction. At its core is a machine learning engine that employs a proprietary methodology to create an optimal ensemble of multiple algorithms (supervised and unsupervised).

Customizable Platform for Multiple Objectives

It is built on a highly customizable platform that can be tailored to meet multiple, sometimes contradictory, objectives in a recommendation system such as: 1. rewarding existing customers, 2. winning-back old customers, 3. acquiring new customers, and 4. cross-selling new products to existing customers. Additional user-defined objectives can also be specified.

The Data–Product Features, Customer Genome, Product Relevancy Scores

The content based recommender is built on a foundation of multidimensional metrics that characterize a product or customer. For product features, our deep domain expertise in targeted industry segments has enabled us to build a core set of features for multiple use cases. This feature matrix can be augmented by additional user-defined features through a customizable interface. Further, for certain use cases, automatic feature extraction using Antuit’s proprietary Deep Learning algorithms may also be used.

The recommendation engine also generates a real-time customer genome, or profile, using a combination of standard hard data (such as demographics), behavioral data (web interaction metrics, sentiment), and supplemental external data. The customer profile is further augmented by a set of derived metrics using our inference engine. The inference engine uses machine-learning-based models to compute metrics such as technical sophistication, channel preferences, and lifestyle propensity.

Finally, the recommendation engine uses multiple customer-product relevancy scores (such as frequency of past buys, propensity to buy, learning propensity, etc.)  to make the best targeted recommendation. These relevancy scores can also be customized to specific use cases.

Integration with Test & Learn

The Recommender System can be integrated with our Test & Learn platform to improve the Recommender System by validating new hypotheses using A/B testing.