Next best offer (NBO) with personalized marketing based on predictive analytics

Challenge

Being able to present consumers with the right offer at the right time is critical in personalized marketing. When implementing NBO use cases, we at Scigility proceed as follows when applying our frameworks:

• Gather and understand historical data about customer behavior and marketing campaigns. Accurate predictions require a large amount of data and a good understanding of product offerings, the number of products, and how often customers buy new products.
• Consider the scope and how the model will be used. Will the predictions be used to define marketing campaigns? Or for recommendations on a website?

Personalized marketing is constantly evolving, and carefully implemented NBO recommendations produce better conversion rates.

Solution

Scigility supports and enables data science, engineering, and marketing teams by:

• Sourcing and consolidating data from multiple internal/external sources on-premise or on cloud: e.g., CRM systems, Google Analytics, streaming data from Kafka topics, data warehouses, etc.
• Exploring and aggregating behavioral data for feature engineering on single or clusters of machines (using tools such as Apache Spark, Presto, Dask, or pandas)
• Training ML models (e.g., using XGBoost, LightGBM, CatBoost) that align with marketing goals and can be explained to stakeholders.
• Deploying models, maintenance, and operations (monitoring, versioning, validation, etc.) both on public cloud (e.g., Azure ML, AWS Sagemaker) and private cloud or on-premise (e.g., Kubeflow, MLflow).