With the world changing so rapidly, every company or organization that adapts to the changes becomes an example for all others. One very recent example for all companies to learn from and implement in their own decision making is the transformation of RELX Group to a leading global information and analytics company. I mentioned in my previous article, and the first part of this series, that the transformation at RELX Group has been interesting, but somewhat surprising at the same time. Considering the questions this brings to mind, I decided to delve into this concept even more.
For those new to the organization, RELX Group is a global company that provides information and analytics for business customers and professionals across industries. They help scientists make new discoveries, lawyers win cases, doctors save lives, and insurance companies offer customers lower prices. In short, they enable their customers to make better decisions, get better results, and be more productive.
Last time around, I had the pleasure of talking with Kumsal Bayazit, who is the head of the CTO forum at RELX Group. She has been at the organization for over 13 years, which is why her insight was deeply appreciated by me and the readers. Recognizing the interest among readers about this astonishing transformation, and the hunger for more insight, I have organized an interview with Vijay Raghavan this time around. Raghavan is the CTO for the LexisNexis Risk Solutions and Reed Business Information divisions at RELX Group and has been associated with the group for more than 15 years. Having already discussed the first two pillars that RELX focuses on – deep customer understanding and large data sets - in detail with Kumsal Bayazit, I wanted to question Raghavan about RELX’s two additional capabilities:
Vijay Raghavan started by emphasizing how RELX Group first embraced Big Data analytics by funding and cultivating a solution called HPCC Systems (High Performance Computer Cluster). The technology behind HPCC Systems was initially acquired by RELX Group in 2004. For more than a decade, the platform has been further modified and enhanced to make it what it is today.
“We created HPCC Systems because back in the late 1990s, not very many businesses had a big data problem. But we had a big data problem by virtue of our vast collection of public records data that we needed to ingest and make sense out of, so we built a revolutionary platform to deal with all that data.”
According to Raghavan, RELX Group has since altered its policies by using HPCC along with other third party applications and making it open source. Technologists are currently attracted towards RELX and the setup that it has, as they can flex their talents and knowledge with better opportunity for innovation.
“Our strategy is to use the appropriate technologies and apply them to our content sets to drive the right outcomes for our professional customers. Ultimately, the biggest value we provide and our secret sauce are the algorithms we write that enable us to analyse the underlying content sets to lead to actionable insights.”
RELX Group ingests thousands of quality data sets from numerous sources in varying formats. The data then comes into the company, and follows a funnel-like process. The data that is coming in is reduced in volume and increased in quality. This specifically means that the data that are considered irrelevant or unreliable are suppressed in favour of consistent data. Even though RELX Group deals in Big Data, its customers have a very specific set of needs and want to see “small data” that is accurate and actionable enough for them to make concrete decisions.
“Our technology enables us to do so much more with content today than even five years ago. We can take all our high-quality content, structured and unstructured, and we crank through it with Big Data technology. We employ our linking algorithms to create clusters of entities, which could be people or businesses or properties etc.”
Once the process of linking the data is over, the group members direct their attention towards ensuring that customers get the insight they need in the form of scoring models or attributes. The linking process and the construction of models and attributes involve the usage of advanced algorithms that have been created or enhanced by RELX Group. The insights that are garnered after this step are calculated to get the best output for customers, whether it is an insurance company calculating the right auto insurance premium for a consumer, or a financial institution approving a mortgage loan for a consumer, a pharmacist making sure that a prescription being filled is legitimate, or a law enforcement agency attempting to track down a fugitive. One of the secrets behind generating these insights is the use of machine learning techniques. These techniques are extremely useful when it comes to the construction of scoring models, but it all starts with the depth and breadth of the data that RELX Group collects.
“We collect data from over 10,000 different data sources to assemble a full picture of a consumer or a business. We make it a point to get the broadest possible set of data we possibly can, so the data does not artificially skew the insights that we derive from it. For example, if we collect property records only from the Southeastern United States and not from the rest of the country, the value of that data is questionable. Collecting the breadth of data that we do is an integral part of what makes our solutions so compelling.”
The total amount of data that is currently being used by RELX Group is in excess of 65 billion records. Indeed, it is their superior data quality and data maintenance techniques that make them handle this Big Data with few hiccups.
“Machine Learning (ML) and Deep Learning do not replace our Big Data infrastructure HPCC Systems. They augment it.”
As we transition from the stage of Advanced Analytics to the stage of Predictive Analytics, Machine Learning is a powerful way to deliver decision tools quickly and effectively, rather than exclusively relying on humans. According to Raghavan, the modelers at RELX Group use Machine Learning to their advantage in several ways.
In a recent success story, Machine Learning is used by LexisNexis Risk Solutions to improve modeler productivity in the Small Business Financial Exchange (SBFE), by way of improving the speed of data discovery in attribute development and model development. LexisNexis Risk Solutions has effectively set up a credit bureau over the last two years, to provide scores and attributes to Financial Services customers who want to understand the risk of doing business with small businesses via the SBFE platform.
In addition, LexisNexis Risk Solutions uses Machine Learning techniques to improve the effectiveness of its Telematics products that help its auto insurance customers provide Usage Based Insurance. Specifically, these techniques are used to create a “Driver Signature” model to enable consumers to have more flexibility in using Telematics products, e.g. by automatically detecting whether the person in the car is in fact the driver or merely a passenger.
A final example: RELX Group always strives to improve the efficacy of its scoring models. In the context of helping its customers address the problem of fraud, waste and abuse, Machine Learning techniques have come in handy in helping LexisNexis Risk Solutions boost the performance of the scoring model. This has enabled customers to examine a smaller subset of transactions and detect the maximum amount of potential fraud.
The interesting information springing from the transformation at RELX Group is not only exciting but also educational. This Interview with Vijay Raghavan, CTO for Risk and Business Analytics at RELX Group, was an in-depth account of what inspires the recent success of RELX Group, their control over Big Data and their exemplary use of Machine Learning. To learn more about this, you can visit the website for the RELX Group.