The Knowledge Exchange / Risk Management / Blending the network soup

Blending the network soup

How big data analytics helps HMRC prevent fraud, recoup revenue, boost productivity and improve taxpayer services

Take unfathomable amounts of system data from a government tax agency. Mix it with third-party data and commercial data sets about taxpayers. Blend well. Now, serve this “network soup” to more than 3,000 frontline investigators and 150 data specialists who use a mix of data mining and visual analytics to taste test the combination.

What happens?

At HM Revenue & Customs (HMRC), this blend had produced an extra 2 billion pounds in tax revenue by the end of 2012. And that’s just for starters.

Our story began about five years ago, when we merged the Customs and Excise division with the Inland Revenue division to form HMRC. Both groups dealt with tax information already. We thought we could be more efficient at managing our data and compliance efforts – and serve our taxpayers better – if we merged.

As part of the merger, we did a proof-of-concept project that linked subsets of data from different departments and tax regimes. There were several vendors – Capgemini, Detica and SAS – who helped us set this up, mine our data, and use analytics and visualization software to get the most we could out of our combined data.

As it turned out, the proof was in the pudding – or, should I say: “the soup.”

Right away we saw the value of blending our data. The pilot project turned up some real cases that we couldn’t ignore; some risks that we needed to tackle right away. This real-life evidence made a very compelling business case.

How it works

Once the business case was approved, we worked with our vendors to build a full-fledged solution. We literally took the pilot and shifted it onto an enterprise platform that we call Connect. Today, we use Connect for several things. For example:

  • We use it to segment our population and identify behavior patterns. Those patterns indicate things ranging from simple errors, to people who need help, to outright fraud – and all flavors in between.
  • We also use it to assess whether we need to make a repayment to a tax customer.

There are two main components in Connect.

  • There’s an analytical platform for our data specialists. They look for patterns and footprints in the data to find certain behaviors. Once they’ve spotted this behavior, they link it in the network view and send it as part of the case pack directly to an investigator, where the research continues.
  • The front-line investigators navigate around the data using a visualization tool. This gives them an intuitive view of the taxpayer, or the entities associated with the taxpayer, so it’s faster and easier for them to find associations in the data.

The advantages

The benefits we’ve received from Connect are truly impressive, with cost savings at the top. Going back four years to include the benefits we gained during the pilot, we had gained an extra 2 billion pounds of tax revenue by the end of 2012. That was from our initial investment of around 45 million pounds.

The second benefit is productivity, which is dramatically improved for us. For instance, the number of people at HMRC carrying out bulk risking has been reduced by 40 percent. And our compliance investigator teams are much better targeted on the highest risk cases. Plus, they can use electronic case referral and workflow to work on cases anywhere in the country.

Finally, because we can make better use of our big data, we can potentially improve customer service more. To give an example, some tax agencies around the world prepopulate the registrations for tax return forms, and they give taxpayers an account that lets them see what they owe or what they should get back across multiple types of taxes – from multiple devices.

Lessons learned

A proof of concept is definitely the way to go for any organization that wants to start a big data project. It gives you tremendous insight about the value of your data. But linking the data to create a good customer view depends on the quality of your data. And it may depend on legacy systems your organization has built up over a period of time. So it can become very complex.

Here are some things to keep in mind as you get started:

  • The key to getting the most out of the technology is the innovation and creativity of the individuals who use it. Those skills are hard to retain, because data scientists are in high demand. You may have to compete to get and keep the right skill sets.
  • In terms of actually using the tool, don’t be too prescriptive early on. We had some preconceived ideas of how we might use the tool. But we found that once we brought the data together, it often surprised us to discover the kinds of questions the data could answer.
  • For those of you starting a new business, it would be ideal to establish your digital data standards right up front. That way you can effectively stitch your data together from the start – that will make it easier to create the customer views you want.

A look at what’s coming

Being able to monitor customer behaviors, profile customers in real time and target very personalized offers has huge implications in tax and welfare. For instance, it could improve the relevance of our outbound communications and our compliance outcomes. We might be able to nudge people into making more accurate declarations of their income – and that would save expenses for post-event compliance investigations later.

In the future, we want to put these capabilities – these views of our data – directly into the hands of our chief head of department and other key decision makers. We want them to have dashboard views so they can do sophisticated drill downs and analytics themselves. That way, they’ll be able to manipulate the data as they’d like – without asking for any help.

Dig a little deeper into how HMRC used analytics to revamp data collection and analysis to find billions in additional tax revenue. Read Recovering £7 billion in additional tax revenues.

Tags: , , , ,
  • Facebook
  • Twitter
  • Digg
  • LinkedIn
  • email

Post a Comment

Your email is never published nor shared. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>