It’s a pretty safe bet to say that former US President Theodore Roosevelt would have loved machine learning, writes Derek Canfield, the general manager for Business Analytics at TEOCO. He left behind a treasure trove of quotes, but one of his most famous ones is perhaps even truer today than it was for Teddy.
“I believe the more you know about the past, the better prepared you are for the future.”
There’s a lesson right there for today’s communications service providers (CSPs), many of whom are preparing to face the future while struggling to overcome the challenge of digital business transformation. Faced with new potential content partnerships, the introduction of new revenue streams, and the opportunities of the Internet of Things (IoT), the options for CSP reinvention are endless and there are almost too many choices and decisions to make. How can CSPs really know the best approach for their business and for their existing and potential customer base?
Spurred on by the threat of disintermediation from the over-the-top (OTTs) players, the operator community is being encouraged to adopt a fail-fast mentality. Devise something, build something, launch something – almost anything – test the market and see if it flies. Make use of the benefits of virtualisation to launch a raft of new services quickly and virtually leave it to fate to see what sticks.
It is said that one of the great things about this fail-fast mentality is that you can learn quickly from the experience to improve your next offering. Your hindsight informs your next move and gives you a degree of foresight that can be immediately applied.
Teddy Roosevelt would only have been partially satisfied with that approach. His approach represents a thirst for knowledge – ‘the more you know about the past’ is the key element of his philosophy. I get the feeling he wanted to soak up as much information as possible to inform his decisions and planning. And if machine learning and business analytics software had been around to help Teddy, he would have embraced the technology with open arms.
In fact, machine learning has been around a long time – the term was first coined by artificial intelligence pioneer Arthur Samuel back in 1959 – but it is only now that we are beginning to see it really infiltrate our homes on a daily basis. Netflix recently commented that its machine learning algorithm, which recommends personalized TV shows and movies to subscribers, saved the company US$1 billion in customer retention costs because it was able to quickly and accurately provide recommendations that would keep viewers tuned in and coming back for more.
In 2018, Gartner predicts the growth rate for the machine learning and artificial intelligence market will hit 70%. And CSPs are in pole position to take advantage of this technology to reap its potential rewards. They have the volume of data, they have the variety of data – they can get it at speed and it is verified.
By embracing machine learning and business analytics to really understand their customers, CSPs can actually get many of the benefits of hindsight up front. They can gather the intelligence to approach new business decisions with confidence as well as validate and optimise existing business with significantly more insight and confidence. The analysis can help protect existing revenues as well as inform new opportunities.
For example, CSPs can use machine learning capabilities to zero-in on a target customer segment to understand the network conditions that could be causing churn. Recently, we helped a CSP take a deeper dive into its pre-paid customer base to try to determine why the average customer lifespan was only around six months. By analysing data from multiple sources right across all the operations of the business – from the network itself, to customer data plans, and typical account activity – we were able to highlight some patterns that emerged just before a subscriber churned.
Some of these were location specific – such as in-market roaming, dropped calls, and 4G capacity constraints – but they were all impacting service quality. Even more importantly, machine learning was able to determine that another root cause of the problem were handsets not being configured to correctly select the home carrier. Rectifying that problem, and targeting specific LTE cell sites with increased capacity, quickly improved service improved service quality, and now 90% of the pre-paid subscribers have moved from an average life-span of less than six months, to one of more than 13 months.
When it comes to developing and introducing new services in, for example, the IoT space; advanced techniques in analytics and machine learning can be used to help CSPs optimise IoT offerings and smart city applications. Whether it is for an airport, a hospital, a connected car service, or an entire Smart City, machine learning can inform advanced network performance management, automation and orchestration functionalities – including real time key performance indicators (KPI) monitoring, threshold crossing alarms, event and trouble ticket management, and predictive trend analysis and forecasting.
In the same way that CSPs leverage, correlate and analyse the massive amounts of data that traverses their digital networks, smart cities will be able to integrate and correlate data from transportation, utilities, city services, weather tracking systems and even social media. Examining and applying machine learning to the historic data, will help smart cities provide fast, coordinated responses to future emergencies, natural disasters and terrorist attacks effectively and more quickly.
Those service providers that equip themselves with the best analytical tools to understand and learn from the past, will be best placed to manage the challenges of the future. They will better understand their business and be able to achieve the competitive edge they need to survive and thrive during this digital transformation. President Roosevelt might not have understood some of today’s technology, but he would certainly have embraced the thinking, and machine learning in particular.
