Telecoms as a subscription business model has been around for over 100 years, but in the last few decades, this segment of the industry has undergone a period of rapid evolution. The disruption of the Internet has prompted telcos to shift from a consumption model based on minutes into more of a utility provider – exploring a wide variety of ways to add value for consumers. These providers have also been forced to become more agile, a transition which requires quite a bit of acceleration to avoid being overtaken by Internet-native rivals.
Critical back-office functions around subscription management, billing, and a whole host of other functions were originally created three or four decades ago. These critical yet dated systems are often complex and difficult to adapt to the rapid pace of change needed within the digital domain. This spans a whole host of tasks, from launching new products to engaging with existing customers in meaningful ways that help to foster loyalty and boost usage – which, in turn, reduces churn.
Artificial intelligence (AI) offers a potential opportunity that can enable operators to react much more quickly to a dynamic subscriber landscape. AI is changing the way we think about data. As more standalone products become subscription services, the pool of subscriber data offers a potential goldmine for gauging consumer sentiment and targeted advertising. To effectively navigate this era of unprecedented connectivity and data volumes, telcos are turning to AI as a critical catalyst for innovation, operational efficiency, and enhanced customer experiences.
Data-driven
Although AI is the current hot topic in technology, with generative AI and the ChatGPTs of the world the most visible examples, it has been used in the subscription management space for the last decade. AI is powered by data, and in simple terms, it uses software to predict outcomes accurately without being explicitly programmed to do so. Algorithms use historical data as input to predict new output values, with recommendation engines and fraud detection as two of the most common use cases.
The most recent iteration is through the use of generative AI, which allows us to frame natural language questions to interrogate a dataset of tens of billions of data points. It enables us to ask the right question at key moments that help an operator create business value. For example, this can be as simple as predicting which product a person is most likely to buy and respond by offering a smaller selection of relevant options. This delicately designed strategy avoids creating the unnecessary friction that would occur by presenting someone with a long menu of options and hoping they have the time and patience to identify what they need. Operators can also make better informed decisions about which payment methods are more likely to drive sign-up. Another example of how this type of AI technology can be applied is in detecting the signs of likely subscriber churn – along with the engagement methods that have proven to mitigate it from happening. Processes all based on the data collated from millions of user profiles from hundreds of different services worldwide.
Another major consideration when using AI with subscriber management is how to make it extensible to allow new datasets to be combined into the model. From experience, it must be anonymised but linkable, adding more information around demographics, lifestyle, favorite sporting team, and so on. In fact, hundreds of potential data points further increase prediction accuracy.
Prediction also unlocks insights that affect how services are designed and evolve. For example, we know from our own data that SVOD subscribers who own an account across more than one device are less likely to churn after a single year. There are hundreds of these insights that can be gleaned from large datasets that have been processed through machine learning techniques.
Testing times
The last facet to touch upon is experimentation. The process of A/B or split testing is well understood but can be far superior when combined with AI. Not only can it be accelerated, but the results can further feed into models to deliver more accurate future predictions. This iterative step can be highly automated through generative AI that can build personalized engagement assets based on a deeper understanding of each subscriber.
The ability to make predictions, test hypotheses, and adapt service delivery and subscriber engagement is vital for all subscription-based models. This holds true for telecoms, media and entertainment, retail, or many other sectors. As more telecom providers branch out into other services – either directly or through partnerships - the access to new data points allows AI-driven systems to also grow exponentially in their ability to predict outcomes and offer insights.
Adoption of these methods is quietly growing in the background amongst many operators across the globe. Although maybe not as visible as the likes of ChatGPT or Google Bard, these technologies will change the way services engage with subscribers and help to shape business success for many years to come.
