A retired professor of artificial intelligence, 11 companies in eight countries, a commitment to open source technologies and an appetite to enable automation may not at first appear to be the ideal ingredients for a group-wide data analytics function but that’s what Axiata has created with its Axiata Analytics business unit. George Malim reports from Digital Transformation World, in Nice, France.
As the discipline of data analytics has matured and become fundamental to the efficient deployment of radio networks – including 5G, in addition to having clearly defined applications from marketing to service assurance, CSP approaches to data analytics are moving away from proprietary tools to embrace open source software and cloud-based platforms – where possible.
Axiata, which has six telecoms operators, three retail companies, one infrastructure company and one business-to-business service provider, within its group, has sought further efficiency by centralising its analytics activities into a separate business unit that operates across all of its entities. “It’s a centre of excellence within Axiata that covers all the data scientists,” explains Pedro Uria Recio, the head of the Axiata Analysis Centre. “It’s a centralised unit that includes the heads of analytics from across the 11 individual businesses. Axiata Analytics is all the data professionals – about 170 – within the group and comprises data scientists, data engineers, developers and some internal analytics consultants.”
In spite of this centralisation there is no intent to bring together data from the different operations. “We don’t put information from the different countries together, the intention is to streamline our data capabilities,” says Uria Recio. “We’re not interested in it being brought together. One reason is regulation; subscriber data has to be kept within countries. In case of war, if your data is outside your country your enemy could end up using your information against you.”
“It’s something people don’t think about,” he adds. “Data ownership is going to become a geo-political game, especially with IoT. Imagine if the control of an electrical system was from another country.”
However, it’s not just regulation that hampers unification of data from multiple nations. There simply aren’t the business cases from cross-border data. “There are not so many use cases for applying data from one country to another,” he says. “I could argue that there may be some sub-use cases like network planning, for example to optimise key performance indicators (KPIs), but I don’t see many marketing use cases.”
Axiata therefore is pooling data analytics resources across its business units while keeping each distinct from one another. This is enabling it to create a larger, centralised data analytics function that the individual businesses would require while also enabling it to keep its data analytics in-house.
“Analytics is like the brain of your company so why would you outsource it?” he asks. “I’m not saying some parts can’t be outsourced, maybe predictive maintenance is better done by the original vendor, but analytics is better done in-house.”
Uria Recio is committed to building and keeping data analytics skills in-house and feels deployments by consultants or vendors which encompass training of staff are not the ideal approach. “Many companies have been with consultants or vendors to develop the use case and then develop the capabilities of the operator but it never happens. The consultant says yes to this requirement to develop staff but then the operator puts them under pressure to deliver a result within three months and the training need is not a priority. Vendors are not going to do it for you, you have to do it yourself.”
To this end, Axiata has developed its own training programme which has engaged a retired professor of artificial intelligence (AI) from Kuala Lumpur University as one of its internal trainers. Uria Recio points out that this isn’t a far-fetched hire. “AI is very old, extending back to the 1950s, and this professor really knows the fundamentals of it, which is linear algebra.”
The skills therefore are being developed. The technology meanwhile is being kept streamlined. “We have a reference architecture which is built on 100% open source technology,” he confirms.
It’s the employee and business culture that present the greatest challenges. “The challenges are all cultural challenges,” Uria Recio adds. “Technology is easy to solve but people are much harder. A key challenge is adopting agile methodologies but this is much broader than analytics alone. Other challenges include how to prioritise and keep focused and how to manage the impact of automation.”
“How to convince people to automate processes when they know the processes are automating their job can be hard,” he explains. “A lot of people can go up the skillset ladder but there are some who are not able to. When it comes to data engineering, for example, it will be fully automated.”
Looking to short term future, Uria Recio identifies external data monetisation and IoT as the next big things for Axiata Analytics. “Both are external and both are adjacencies,” he says. “In the core mobile business what can you do? It will be constrained by the fall of prices so the money is going to be in monetising data externally. 5G is going to be a massive cost so new revenues need to come from elsewhere to support that investment.”
Axiata Analytics is therefore focusing on analytics data for advertising as a new revenue generator and capex planning as a quick means of generating internal cost savings. “Cost reduction is a quick win,” he acknowledges. “Advertising is also a quick win because it uses telecoms data to target advertisers. The data exists and the advertising market is fragmented.”
