It’s no secret that many mobile operators consider the need to engage with customers as individuals to be one of their biggest commercial challenges. But due to the sheer size of the task, many focus on price and network asset management to entice customers and limit churn – on a “herd” basis, not a personal one.
However as Digital Service Providers enter the telecoms market, operators face fierce competition for customers’ attention and spend. To keep up, operators must focus on creating a personalised experience for each subscriber; one that communicates value, increases loyalty, and recognises them as an individuals. Something like that of Netflix, Google and Facebook do very well, say Jane Zavalishina, CEO of Yandex Data Factory
This need is understood too well by most operators. They know equally well that achieving this relies on collecting and interpreting data from multiple sources – subscriber profiles, network performances, and individual customer traffic data to name just a few.
But isn’t this a story that has been told many times before? Isn’t using data to improve customer engagement and drive profits old news? Yes and no.
The current analytics methods being used by operators are underutilising the immense amounts of data being generated by and within the network. Operators tend to aggregate the data, providing instant assess and visualisations, and examine the datasets to uncover useful business information – to inform and support human decision-making. While providing certain business value, this approach has it flaws.
Human decision-making – whether based on big data analysis or not – is highly susceptible to bias and largely reliant on experience over facts. For humans to use big data, the information is generalised to the point it loses its power. Quite simply, human brains lack the processing power to interpret the sheer volume of data managed, and to create thousands of hypotheses that are compared against each other to determine the most suitable course of action for each subscriber. With this, big data’s usefulness is often being relegated to supporting infrequent, localised, biased and late decisions.
The volume and complexity of the data on which we are trying to make decisions has reached the point where a lot of operational decision-making needs to be handed over to the machines.
Machine Learning is the technology that brings real disruption to business, by taking a scientific approach that allows not only analysing the data, but also predicting, recommending and making automated decisions. This is a complementary technology that goes beyond traditional descriptive analytics (“here’s what used to happen before”), allowing us to predict and prescribe actions (“here’s what is likely to happen and what you should do with it”). Machine Learning allows incorporating all the existing data to choose the next best action and actively learns from the outcomes in order to feed future automated decision-making.
Machine Learning’s greatest advantage over previous incarnations of data analytics is that it does not require the operator to have a deep understanding of the domain itself, nor to have data interpretation capabilities. The key is simply the provision of relevant historic data – which operators are blessed with a wealth of – for the “machine” to interpret, test, act upon and then refine its outcomes.
For example, when applied to personalisation or churn prevention, this has a rapid and positive impact on a customer’s lifetime value. Consider being able to identify and describe what the warning signs are of a churning customer. This is where traditional analytics tends to stop, leaving marketers to decide how to run the churn prevention campaign or target an offer to one of a few chosen segments. In contrast, Machine Learning means that the operator can foresee when a customer is about to stop using the network and automatically deduce an informed suggestion of how to respond to prevent this customer from churning, and then act upon it. This is based on how other similar customers have acted and the predicted probability of conversion for each possible course of action.
Mobile operators need to embrace Machine Learning, fast. It’s not enough to rely on technology just for extracting data. Machines need to be empowered to not only store, cross-reference and present data, but to also act upon it – calculating the most suitable action from hundreds of thousands of hypotheses – and then also learn from every action and reaction. The benefits are the same as have always been claimed for big data – more activity, faster responses and lower costs – but it is only through Machine Learning that they can realistically become reality.
