Brian Seery, Singlepoint’s technical consulting director, spoke to Ian Campbell from The Sunday Business Post about the adoption of real-time analytics. Read the transcript of this interview below.
“The problem with the “the next big thing” in business technology is that you always have to wade through an ocean of hype to get a clear picture of what organisations are actually doing with it. In the case of analytics, data may well turn out to be “the new oil” or facilitate robots taking our jobs, but for now it’s an area of IT that some companies are using for business advantage while many more are struggling to unlock its potential. There’s a widespread understanding that accessing the increasing amounts of data companies hold in their systems will yield valuable business insights and inform better decisions, but organisations find it hard to execute on strategies to optimise the information that’s available to them.
Advances in technology and the evolution of the cloud in particular have made data analytics accessible to more organisations, but it’s not a magic button. Singlepoint is in the business of building infrastructure to support data-driven businesses and often starts customer conversations with a reality check.
“There’s a big buzz around machine learning and AI, and organisations see it as panacea for fixing all their problem. But if they don’t get the basics right then everything that’s built on top is going to be built on a foundation of sand,” warned Brian Seery, technical consulting director at Singlepoint. The basics, in Seery’s view, include addressing the age-old mantra of “garbage in, garbage out”, ensuring that the data quality is good. Essentially, it’s a reminder that a fundamental part of the data collection and cleansing process, ETL (extract, transform, load), is still important.
“The models that companies build are only going to be as good as the data feeding them. You have to get the data in the right place and consistent across lots of different sources. We work for a lot of large companies and that’s by far the biggest challenge that they’re facing,” he said. Another basic is having the skillsets to execute on projects — one area of data analytics that is probably getting harder not easier, as Seery explained. “In most cases, if you are going to put machine learning and predictive analytics into production you are going to do it in the cloud. So you will need dev op skillsets to get a model up and running that’s usable,” he said.
This is in addition to having expertise in core IT, cloud engineering and data science. “You are going to need all these skills together to do machine learning effectively. That’s a big challenge for companies,” Seery said. There are further challenges after a project goes into production, when a good understanding of predictive analytics is essential to ensure insights stay useful.
Seery explained that a fundamental part of the process is continually tweaking the production process, refining models over time as the data changes. Another complication is the increasing need for explainable artificial intelligence, or XAI. Machine learning uses algorithms to automatically make decisions without any human involvement, such as setting an insurance customer’s premium. “If you’re going to allow AI to make decisions around offerings for customers, then you have to have an understanding of why those decisions were made. That’s going to be a big area of need,” he said.”
If you would like to speak with Brian or one of his team, please contact +353 (0) 1 562 0027 or email@example.com