Propensity Modelling Services.

We get a clear understanding of the business domain, and how this translates to the data captured. We understand where the transactional data is held, how we would access it and how to transform for usage. We ensure that we can link a visitor to the end purchase, and what the key business use cases for the propensity model will be. We agree the output KPIs, and what will the program improve.

Leveraging the business domain knowledge garnered, we select features from the data that we will use to test the first iterations of the model. We identify any gaps in the data and what will we need to do to close these gaps to get an MVP up and running. A key differentiator over out of the box propensity models is our ability to engineer new features to create more sophisticated models.

We build and train the model using real world data so that we can test the accuracy of the model and test our initial features. We calculate the propensity scores and test the feature importance. Modelling on which features make the best contribution to the model's prediction and propensity scores of customers.

In this phase we integrate the outputs of the program into the existing or new personalisation ecosystem, whether this is CRM, marketing automation, CDP, or a marketing cloud. We specify how the test and learn program will run to prove the incremental improvements from the program.

We deploy the new propensity model into production in a controlled tested manner to statistically prove that the outputs from the model generate incremental lift, before rolling out the program to all customers and moving into a continuous optimisation program.