Data makes the world go round. And with the rise of artificial intelligence and machine learning, data is even more critical in business. But some companies struggle to keep up with the latest changes in the industry – understanding data and how to make the best use of it.
In this episode of the Data Point Of View, learn why it's important to hire the right talent and embrace innovation from Anshuman Banerjee, a Domain Chapter Lead of Commercial Data at Spark New Zealand. Anshuman brings a deep understanding of how data plays a key role in understanding customers, and how to fill in the gap between business teams and data teams.
The telecommunications industry will continue to embrace technology and data to drive innovation.
- Being in telecommunications for most of his career, Anshuman has witnessed a lot of the ups and downs that have occurred in the industry. Today, he says he's lucky to see telecom embrace digital innovation and is excited to see what's next on the horizon.
Commercial data is the bridge between business and data.
As a Domain Chapter Lead of Commercial Data, Anshuman and his team use data and advanced analytics to power the business and solve complex issues. In this episode, Anshuman explains how commercial data came about in Spark. "Typically, what we saw at Spark was that we had these groups of people who understood data quite well, and we had a group of people who understood the business really well, but the interconnect between business and data was missing. I thought that the potential of what's possible via data was not clear either to the data group or to the business area, and then this was a gap that was identified not just by me but by a few other people in management as well. And that's how this area called Commercial Data was formed. It is about how you use data to drive commercial benefits for Spark."
Data and information are going to be even more important in 2022.
Anshuman shared that, he sees that data and information are going to become ever more important. He has to think about how to scale a team and the business at the same time. He has a team of highly talented people, and it's a challenge to bring in more people and maintain the same costs and maintain the same degree of drive etc., in the team. Also, when he is bringing in people who are highly capable, they want to be challenged, and they want to work on things that are interesting.
The conversation included a lot of discussion on understanding customers and keeping that at the center of all the work they do at Spark. According to Anshuman, “at Spark, our vision talks about helping all New Zealanders win big in the digital world and keeping the customer at the center of it by understanding when a customer needs something and what they need.” He wants to be able to anticipate customer needs and wants and that using information and data is the way to make them a lot more relevant in terms of the conversations we are having with our customers, as opposed to a world where it's driven by the objectives they want to reach. We don’t want to bombard customers with numerous offers, without doing what really works for them, they want to understand customers better and then providing them with things that they really need and want.
Another area for discussion was scaling teams and use of artificial intelligence. Spark started out with one squad building models and then continued to add squads but realize that they did not really know how to scale. The result was that the squads were independently creating the data features required for models independently. In data science, 80 - 85% of the work is in feature creation, so not finding a way to manage feature creation was resulting in duplication and wasted time. They realized there needed to be a certain degree of streamlining and that they needed to get faster in the way they created models. To address this, Anshuman and his team created a universal layer of features, or a Feature Store. Features are stored, tagged and marked as reliable and can be shared across the company. New models that are being built can immediately tap into this Feature Store and use these features. They have reduced their model creation timelines from over two months to almost a few days which has provided enormous benefit and has helped them scale.