"What does good look like?" the client asked us on the eve of the launch of their new service.
Networks and MVPDs have a playbook for what they should do to improve their core linear TV services, based on years of experience of operating one. They know, within a few months of a customer joining, to which segment of users they belong. They know how to target them and at what time to upsell them to other products. And they know when customers are most likely to churn, and what to do to stop that from happening.
But when it comes to OTT services, the playbook can be empty. The kind of content that watched on OTT differs from their core services. Quality of service expectations vary. Acquisition, retention, and upsell are very different for a month-by-month service that costs $10 a month than for a $100 monthly service with a 12-month minimum contract. And when service providers use OTT to take their content outside of their traditional markets or geographies it opens up a whole new set of parameters.
Making sense of OTT
We worked with a client that had launched a stand-alone OTT service in their local market, and in several markets outside their home territory. They wanted to understand how usage of the service was developing, emerging clusters of usage segments and actionable strategies to aid customer acquisition and retention. In short, they wanted to know "what good looked like."
Bring on the data science
Our data science consulting team took six months’ worth of OTT TV data and mapped it to CRM data, and some first-party data that the client had access to. We excluded data which we knew was subject to data quality issues.
We used a bespoke clustering methodology, developed by our data science consulting team, which produced distinctive usage segments. We analyzed each part, classified the usage and analyzed the data to provide specific and actionable recommendations on how to better target each group.
Combining what we saw from the TV data, our knowledge of our client's strategy, and benchmarketing from other operators, we came up with recommendations.
How to make friends and influence people, OTT-style
The segmentation surfaced eight distinct groups of OTT users. Some of these were segments that the client immediately recognized and expected to exist. There was a large group of users who were relatively new to the service and watched it only lightly. These users were the most critical to retain because they were also the ones who were most likely to stop using the service. The client used the model to target offers, using content touchpoints that particularly resonated with this segment, rather than those that were popular across the rest of the customer base.
Another segment was driven heavily by live events. Viewers were using the service to catch up on content that they would otherwise have missed by being away from the main set. The client created a campaign designed to improve the engagement of this segment of users, highlighting "unmissable content."
Two segments were heavily sport-focused. The client used the segmentation to communicate with its sports-based differently. One portion was a much lighter user of the service, and more at risk of churning. The client communicated with this segment by reinforcing the core, showpiece events that were available on the service. It targeted heavier users with non-sports content, to try and broaden the appeal of the service for this segment beyond just sports.
We also identified a “Pay-TV-like” segment – customers who, based on the data, were using the service in a set-top box-like way, with a mix of content genres, and a combination of live and on-demand viewing. This segment was also affluent, and loyal. The customer agreed with our assessment and attempted to upsell this group to the core, and more profitable, Pay TV service.
The final segment watched the service heavily on connected TVs and large-screen devices. This group was less affluent and lighter viewers, so were deemed a less relevant target for upgrading to the core service. They were instead targetied with offers around transactional movies, as the segment over-indexed in the viewing of both the core movie channels and transactional VoD.
What good looks like
Before we rolled out our recommendations to the entire based, we tested them on a group of subscribers. Subscribers who received marketing targeted to them based on the segmentation were 60% more likely to re-engage with the service and watched 20% more content than those that received the standard marketing offers.
That's what good looks like. Data science is now at the heart of the client's efforts to grow the OTT business, and the client is using it as a first step to offer different forms of marketing communications to different customers.