After years of outsourcing projects, Financial Times built its own customer segmentation model. The FT Customer Segmentation tool uses FT’s own data to understand which behaviours predict value so we can better acquire and engage those specific types of customers.
This segmentation replaces and improves upon an existing audience segmentation created by a third-party several years ago. That version used market research to group users and was, at the time, a helpful tool for us to think of audiences by their motivations for using FT.
However, we could not tie our FT subscribers to these third-party-created segments, so we had no idea how valuable these audiences were to the FT in the long run. Creating our own segments, using our own internal data, has allowed us to focus on the customers most valuable to the FT rather than optimising for pure volume.
Identifying user groups
Our data-led customer segmentation has identified nine distinct audience groups. Our prospective (non-subscriber) audience falls into three of these. The model applied:
- Unsupervised clustering (k-prototypes), using a combination of demographics (e.g. what industry they work in).
- FT product traits (e.g. if they are a standard or premium subscriber to the FT).
- And behavioural data (e.g. when they visit the FT, if they click on notifications and if they hit barriers to subscribe).
By overlaying additional quantitative and qualitative data, we have been able to determine which of these clusters have the highest lifetime value with the FT — as well as their attitudes and motivations — to provide a clearer view of why they interact with the FT.
Like many publishers, we at the FT are trying to adapt our product landscape and strategy to reach and engage new audiences. One drawback of outsourcing these types of projects in the past was a disconnect between the methods used by third parties and our own data.
These third-party segmentations provided a snapshot view that was increasingly outdated and out of our control to refresh as our product landscape and audiences changed. By applying this new data-led technique, we now have control of our segmentation. We can ensure that our segments always reflect our evolving audiences and their needs and wants.
This is particularly important as we launch and release new products.
The introduction of the new Customer Segmentation tool has also allowed the B2C department to start taking the necessary steps to mitigate the risks of losing third-party cookies. The acquisition team has a pipeline of follow-up A/B tests planned for 2023, focused on showing the effectiveness of the tool on acquisition:
- Lookalike targeting: Working with our partners, we plan to use the Customer Segmentation tool to target prospective users who look similar to high-value FT subscribers, tailoring the product, price, and proposition messaging for each cluster to increase conversion rates.
- Registered users engagement: The Customer Segmentation tool is the first time we have created segments for our registered (i.e. non-subscriber) user base. We will experiment with registered user engagement tactics and the warm-up journey to becoming a paying subscriber.
This new approach has involved a lot of coordination, collaboration, and up-front work to create and validate the Customer Segmentation model. But we believe the additional benefits will, in time, be well worth the effort.