DM GROUP developed a new technology for marketing database management, target identification and 1to1 real time communication. As in programmatic buying, DM GROUP reversed the approach for target identification, from “static clusters” on the basis of socio-demographic and behavioral data to “dynamic targets” , according to the reactions to message and /or multichannel communications (e-mail, sms, post, www, …).
The platform was developed on Hadoop framework (H-Base for NoSQL database and Spark for clustering and machine learning). It runs using semantic analysis on text messages sent during marketing activities (in the future also images), in order to identify the best target for each kind of communication. Obviously, the message comes from the insights that are extracted from the database according to classical data mining logic, but then the process is optimized using recommendation algorithms, allowing to define the best audience for every message.
In particular, our platform uses 2 algorithms developed using cutting edge artificial intelligence and machine learning techniques:
– First algorithm analyzes the textual content (tomorrow images) of all messages over time have been sent and creates campaigns cluster. This implies our database is also loaded with full message bodies and complete feedback/action/reaction history.
– Second algorithm creates customer clusters or user, based on their receptivity to a specific content.
The “Synergy” between the two algorithms is designed to identify, in a particular moment, the more receptive target to every message and this is handled automatically and smoothly over time using machine learning techniques. From tests performed on traditional managed campaigns in the past, it was verified we would have been obtained the same results by sending about half of the messages with half the costs.