22 Jun Predict Churn And Control Costs With Predictive Analytics: An Interview With Lucas Bernat, Product Manager At NPAW

“USD8.4 million spent on acquiring customers that will leave the platform within the next month.”

Predictive Analytics are a set of tools which enable us to act as psychics. They provide managers with statistical leaps into the future, to help guide future decisions and directives as well as avoid undesired business situations. It seems not even the future is immune from W. Edwards Deming’s wisdom that “Without data, you’re just another man with an opinion.”

YOUBORA Infinity is the online video industry’s Virgil – guiding managers through the unknown and providing insight into the multiple “levels” of the content delivery ecosystem which to some seems like more than the proverbial Hell. Unprecedented insights are delivered with unparalleled precision. With the SmartUsers SmartModule, unique to the YOUBORA platform, the future is now. Customizable dashboards report accurate predictions of future trends and other projections. For the first time, managers can reliably act to reduce user churn and maximize monetization efforts.

To assist us in our deep dive into the realm of predictive analytics, we sat down with NPAW Product Manager Lucas Bernat. With their guidance, we will explore the technology in further detail, and explore its applications in the market today.

 

How is NPAW incorporating Predictive Analytics into its range of services?

At NPAW, the YOUBORA Infinity platform has been outfitted with a tool to predict user churn. It displays the probability of churn per user and isolates users at the highest risk of adding to the distributor’s churn rate. We determine this using three months of QoE data and consumption patterns to predict, for example, the likelihood the user won’t consume video in the next thirty days. Churn prediction has proven to be accurate, and customers noticed a direct impact on their bottom line. The wider the scope of data churn prediction algorithms process, the more powerful a tool it becomes. Essentially, we’re transforming customer’s perspective from playback monitoring to a bird’s eye-view of the video service – correlating QoE and customer behavior data from the moment users log in until they log out. The more robust the data, the more enriching the insights delivered by the churn prediction algorithm. This is just one example of the far-reaching and diverse array of applications YOUBORA Infinity’s predictive capacity is capable of performing.


And what are some benefits of this?

Well, now content providers can proactively manage their user base, assisting Customer Service and Marketing teams, with the ability to anticipate and identify causes of disturbances to and promote incentives, perhaps vouchers, to at-risk users. Marketing positions can forecast customer acquisition winning strategies and related costs and ROI based on data.

By identifying causes of churn, departments across your company can test and successfully launch all sort of changes to boost engagement – from changes on UI/UX changes to content recommendation engines and more.

 

That really is cutting edge. Are there any other kinds of tech applications in the works?

We have a couple of tools in the mix, something disclosable… A  customer lifetime value model to forecast user’s consumption levels for the next week. This will help teams to guide decisions for individually targeted marketing, customer acquisition, and customer retention resource allocation.

 

Wow… Impressive. What raw data do you collect and which processes does it go through to justify your predictions?

YOUBORA collects a vast quantity of diverse data every day, which gives us a massive pool of data to analyze. For instance, to determine our churn model, we use quality metrics like buffer ratio, interruptions, et cetera; and consumption patterns like user’s viewing history, average playtime per user, among others. Results tested very well in simulation and proven well among customers. We are very proud of its accuracy and foresight it delivers.

 

That’s very thorough indeed. So after having SmartUsers on board YOUBORA, what other behaviors can be predicted?

For starters, you could say, “anti-churn” or the size of the audience pool in the next month. Aside from churn related items and customer lifetime value, YOUBORA can cross reference predictions with real-time behavior also in real time, to update our processes which predict hourly, daily, weekly, and larger duration consumption patterns. The predictive algorithms enable more precise market segmentation.  Also, customers can apply precisely segmented Multivariate Testing (MVT), being able to segment by predictive algorithms as well, to ensure a wide range of successful releases to build stronger engagement and, ultimately, reduce churn. For the scientists out there, our algorithm is ideal for further experimentation and help define the measurable organizational value in devising new configurations, UI/UX arrangements, customer care strategies, and any of the other capabilities I’ve already mentioned.

 

Today, most Managers must justify a positive outcome on the initiatives they action.  How can customers improve ROI with Predictive analytics capabilities?

One scenario I would cite is this, one could forecast customer acquisition cost and ROI.  Say having “X” acquisition cost per user, you know how many months they should be subscribed to amortize off such a cost. By predicting user churn and computing lifetime value on our friendly BI platform, you can forecast ROI and anticipate money lost, like the investment of acquiring new customers to keep your user base stable at least, considering churn.  

To go further, assuming an OTT has an average customer acquisition cost of 40USD per user, multiplied by three million subscribers and a 7% average industry churn for this month would yield a USD8.4 million cost of customer acquisition you have lost this month. USD8.4 million invested on acquiring customers that will leave the platform within the next month.

We took an isolated example for one month to give a big picture number to get a sense of how much a VOD could be losing next month. I am conscious it does not include the monthly membership a churned customer has already paid in previous months, or the number of months each user was subscribed, or costs of the service. Depending on their lifetime in the video service, providers may have or may not recover the initial cost of acquisition. Of course, those USD8.4 million may make perfect sense for an OTT on its expansion stages, as long as the investment monitored and optimized…

 

So as long as it is strategic…

Yes.

 

So that will about wrap us up today, do you have any concluding insights based on your experience and the amount of data you are exposed to?

There is always a saying in the industry, “Quality is king but the content is queen”. We help content providers define the best content strategy and identify titles that perform best. And before a user can enjoy the story, it must be conveyed in a meaningful way – “The medium is the message” as the Father of Media Analysis Marshall McLuhan once said. The right content, paired with an effective playback, is the perfect harmony of art and tech.