Newspaper Group: Increasing Subscription Rates Through Deeper Customer Understanding

Outra developed a Custom Customer Segmentation and highly targeted content strategy to recognise micro-triggers indicative of subscriber behaviour.


Managing and analysing vast amounts and types of structured and unstructured data, often across legacy systems, is a situation many companies find themselves in. Brands need to find a way to work with disparate data sources and customer touch points to be able to deliver a better customer experience and drive loyalty and business profitability.


This British newspaper had 28 data tables of data schema covering 25 million individuals, 250 million interactions and 267 million email actions. They wanted to find out how they could drive value across multiple data sets and legacy systems with the overall objective of increasing subscription rates. They needed to gain a deeper understanding of their highest potential prospect groups and understand what kind of content they would be interested in. They needed to understand subscribers from many different angles but lacked a framework to do so.


Outra cleaned and analysed data which included 267 million email actions, to review their approach to data before applying deep learning to build a multi-layered Custom Customer Segmentation to help drive up subscriptions rates through a deeper customer understanding.

The solution was a multi-step process which began with developing a new Engagement Index. We drew in key engagement data into the analysis which then enabled us to create a benchmark. Pen Portraits were provided against each of the bands that indicated vital behaviours per engagement group.

From this baseline we will build a content segmentation from the new data available and developed content categories and themes that are important to different groups which will inform content development in the future as well as recommending other related content.

The next phase in this project builds in deep learning to help develop acquisition and churn models based on the learning of the custom segmentations.


This newspaper now understands who is likely to subscribe and for which subscription offer. Combined with the content data we are now able to recognise micro-triggers in behaviour which may be indicative of larger actions both positive and negative. i.e. increase in shopper behaviour or an unsubscribe.