Analytics, machine learning and automation are the keys to running smarter, more efficient networks
Machine learning and artificial intelligence (AI) are two technologies that are making a huge impact on virtually every industry, writes Brendan Gill, the chief executive of OpenSignal. Smart tools are being used by organisations across the globe – within current hardware and software – to improve both the decision making and automation processes.
One industry that is feeling the benefit of machine learning and AI is telecoms – and specifically within the communications service provider (CSP) community. Data-driven, automated decisions can help CSPs manage large estates of both network infrastructure and optimise services, enabling them to run better, smarter networks and more efficient services. The efficiencies gained have a huge impact on a CSP’s bottom line, and are also intrinsic to keeping customers happy, reducing costs and opening up new market opportunities. Something the industry is shouting out for as it embraces digital transformation.
Customer expectations are high – anything less than an always-on, 24/7 service is deemed unacceptable. In today’s connected world, even a short service interruption can create customer backlash that can negatively impact the reputation of that brand and/or company.
In the case of over-the-top (OTT) services, heavily consumed by users to communicate and collaborate, any infrastructure failure would be – and has been – deemed catastrophic, leading to certain damage to perceived credibility. In the same vein, if the network the consumer is using to access this service fails, then the CSP is likely to feel the brunt of customer negativity. And this is often communicated and amplified across social media channels for the world to see. With so many of these services accessed on a mobile device, smart networking technology can give the mobile operator the power to improve service reliability. And ultimately, deliver a better and more enhanced customer experience.
Automated, data-driven decisions
There are a number of ways that CSPs can use machine learning to ensure service reliability, allowing the network to make automated, data-driven decisions. Potential demand services during peak traffic times – i.e. major sporting events – can be predicted by using data modelling. This allows the CSP to deploy additional network capacity when and where it is needed. In addition, self-optimised networks (SON) will analyse call, text and data quality, then make remote adjustments using equipment such as tilting antennas.
Machine learning algorithms can also be programmed by CSPs to predict any infrastructure failures – i.e. transmission issues or power outages – before they happen, by using historical and real-time data. This could even lead to the model recognising symptoms such as unexpected packet loss and issuing an alert for maintenance teams to address before it’s noticed by the customer.
In addition, AI-powered software defined networking (SDN) can allocate more bandwidth to a customer as and when there are peaks in demand at specific times of year. A good example of this would be a broadcaster during a live sporting event such as Wimbledon, Ascot or the Rugby Six Nations tournament. It would be business-destroying if the broadcaster did not have enough capacity to upload and distribute content to its viewers.
Frontline services can also get a much-needed boost from AI and data analytics. OTT streaming services such as the likes of Netflix and Amazon Prime are already using machine learning algorithms to encourage customer to make the most of their subscriptions by simply recommending new content that relates to the genres they’ve been interested in in the past.
When used in a similar way, AI can be invaluable to customer service. AI-powered chatbots are not only a simple, cost effective way of handling common queries, but they can also redirect more complex issues to, for example, a call centre A number of CSPs have repatriated their previously outsourced call centres with the intention of bringing agents closer to the customer. The potential increase in costs incurred by this decision can be offset by services that analyse data – including customer history, credit scores and social media behaviour. This ensures people are matched to the most suited agent, helping to boost customer satisfaction and reducing call times.
AI is also being investigated to understand how it could possibly augment human-led customer service through offering automated suggestions. This can speed up resolution times, while also retaining the ‘human touch’ that some customers prefer.
Beyond the benefits to call centre customer service, AI can also be used to analyse customer relationships by helping to spot any early signs of discontent. This serves to head off any potential issues and allows the company to take pre-emptive action to stop attrition, and even launch highly personalized customer retention campaigns.
Quality of data
Machine learning models are only as good as the data they are fed. While CSPs have a plethora of information to draw on, the traditional network-centric metrics they’ve used to capture the data have their limitations. The machine learning model needs to be programmed with a customer-centric perspective, if the goal is to improve customer experience. And this requires the data to actually be customer experience data.
Performance data that is network-related may suggest that a customer is receiving an optimal level of service. But it does not inform the CSP of how buildings, Wi-Fi quality and congestion from other users in the same location are all impacting the experience.
Put simply, the data is missing a true customer centric picture. By collecting and analysing billions of measurements collected from actual users, CSPs can close the data gap. This gives them the edge in building AI that has a positive impact on customer experience.
The real competitive advantage for CSPs, when it comes to AI, lies with using the most relevant and extensive data to program the AI engine, and not from creating a better machine learning algorithm. CSPs have no intention of ever being Google DeepMind, but they do have access to some of the most unique and comprehensive datasets available. Those who focus on unlocking the value in the data they have access to, and then use it to close customer-centric data gaps, will be the true winners in the age of AI.