AI algorithm for Send Time Optimization (STO)

When designing a marketing campaign, marketers have to choose the best content, channel, and time to send a communication in order to increase conversion rate (e.g. open rate, click rate). What is actually “best” changes between one person to another. Among the factors that mostly affect conversion rate there is sending time (e.g. somebody is most likely to engage in the morning, others in the evening, etc.).

Wonkit has designed, implemented, tested, and deployed an AI algorithm for Send Time Optimization (STO) in partnership with Diennea, an Italian marketing company. Diennea expects the STO algorithm to be integrated with MagNews, their marketing automation platform.

A Proof of Concept of the AI algorithm has been delivered to Diennea. In the A/B tests performed the click rate has experienced an increase ranging from +58.94% to +135.43%. Wonkit has currently been integrating the AI algorithm with MagNews. The architecture of the solution is represented below.

Technology stack and architecture

The STO AI algorithm and supporting logic are based on Python and widely used open source libraries for data processing, machine learning and deep learning (NumPy, TensorFlow, …). The whole software is integrated into a Docker container which can run directly on Diennea systems.

The software is backed by a database providing data about events collected by MagNews (sent mails, clicked mails, bounces, …) and storing times recommended by the algorithm.

Wonkit constantly monitors the conversion rates of campaigns driven by the AI algorithm in order to measure its performances.

Diagram of AI integration in MagNews