Predicted lifetime value
Retaining customers is vital for any company to maximise investment, because customer retention is usually much cheaper than acquisition. Effective budgeting for customer retention, in order to maximise revenue, is predicated on the ability to prioritise customers based on their future value. The recent, rapid increase in customer interaction data has provided a new range of opportunities to predict lifetime value for both individual customers and broad segments.
Dynamically predicting the future revenue from an existing customer, and identifying highly valuable new customers before they have made multiple purchases, is an inherently challenging problem. The difficulty is compounded by the scale, complexity, and diversity of data about customer behaviour that must be integrated to make accurate predictions.
We overlay an individual customer’s purchase and interaction history with the cumulative history of all customers to find the most similar long standing customers. Observing these select long standing customers gives us a window into the future purchases and revenue of a new customer.
THE TECHIE BIT
Our machine is built with business in mind, connecting to all relational databases, cloud platforms such as AWS and Azure, Google Analytics and social media profiles – as well as scraping infinite amounts of data from the web. We bring all of this information together into a single cloud data store within our machine and, from there, customers can work directly with their new streamlined data sets or choose to use our other solutions to drive further value.
The Peak AI System ingests all of the data you have about your customers and their transaction histories, and unifies it in one place. It then derives a combination of individual customer features, and cohort features, to use in the prediction of the near future behaviour of your customers. These predictions are completed using an ensemble of approaches, including tree-based machine learning algorithms, regression techniques, and survival analysis.