Home Blog DEM optimisation wit ...

DEM optimisation with a Random Forest based approach

Matteo Bregonzio

Authors: Giulia De Poli, Maria Angélica Lobo Paulino, Stefania Tola, Manuela Bazzarelli, Leone De Marco, Matteo

Optimisation in digital advertising is a complex task deemed to increase customers engagement and satisfaction. In more
detail, optimisation involves not only identifying the right images, template and timing to engage a given customer, but
also understanding the context and judge whether the message is actually relevant to the recipient.

To be successful advertising optimisation requires taking into account lots of data coming from multiple digital sources
such as Customer Relationship Management (CRM), web analytics, and advertising interactions. Although this process could
be performed manually by marketing specialists, more recently data-driven methodologies have shown promising results.

In this direction, our study proposes an automated system addressing advertising optimisation via a supervised learning
approach where decision-making is performed accounting for the latest customers interactions in a near-real-time
fashion. Specifically, this work presents a solution for direct email marketing (DEM) composed of three modules:
monitoring, decision-making and automation. Monitoring is provided through a web dashboard showing historical
performance of relevant Key Performance Indicators (KPI). The decision-making module computes a relevance score
predicting how a given email message or sequence of messages are suitable for a specific customer or cluster of
customers. Subsequently, this score is used to support the decision process within the automation module in order to
deliver fully personalised messages. Experimental results confirm that the proposed DEM management system promotes
customer satisfaction minimising perceived spamming. Moreover, DEM activities contribute to boost the revenue without
sacrificing the customer’s experience.

Learn how 3rdPlace can optimise your DEM. READ NOW THE FULL PAPER