Why data scientists solve irrelevant problems

13 November 2018

In today’s challenging financial market businesses need to utilize every available resource and technology to reduce risk when processing payments and credit applications. One area in particular that is often talked about, but still either under or incorrectly utilized, is data science.

While many financial institutions are working towards implementing data science within their risk decisioning processes many are still working on creating an environment and culture that allows data scientists to be fully effective.

In a recent webinar Ken Schultz, VP of data science at Elevate Credit - better known under their brand Sunny in the UK -discussed the benefits data science can bring to a financial services organization, including the opportunity to increase accuracy, expand your market, and reduce fraud, all of which can be used to drive business growth. However, Schultz highlighted one key area in particular where businesses struggle to best utilize data science to optimize risk decisioning:

A potential issue with using data science in an organization is solving problems that don’t need to be solved, or solving irrelevant problems, and that’s never the fault of the data scientist,” he said.

If it’s not the fault of the data scientist, whose fault is it?

One of the most valuable skills data scientists possess is the ability to find stories in data. These data-backed stories can be used to drive changes in risk decisioning processes that lead to more reliable risk analysis and lower incidences of fraud and credit defaults. However, to fully maximize the effectiveness of the data scientist’s input they need to understand what stories they are looking for in the data and why. Without this knowledge they have no way of knowing whether they are finding solutions to relevant problems or wasting their and the business’ time.

For example, two data scientists are given a set of data and asked to find actionable insights from that data. Data scientist A looks at the data and identifies that applications peak at certain times of the day. Data scientist B looks at the data and finds that applicants that meet a certain set of criteria are more likely to default.

Neither scientist is incorrect, however only one of them produced results that were valuable to the business at that time. To understand why, the results need to be taken in context:

The business where the data scientists work has a world-class risk decisioning solution that is capable of handling thousands of applications at a time and is fully prepared for any spike in volume. However, the business is unhappy with its decisioning accuracy and is looking for additional data trends that can be used to improve accuracy.

So, while both findings could be considered to have value, only data scientist B’s results are valuable in this particular situation. B’s findings provided actionable business insights that solved an existing business problem - they could be used to improve risk scoring and lower default rate.

However, data scientist A is not at fault. Why? Because the business had failed to provide the business context that would enable the scientist to identify a solution in the data.

Three steps to getting the most value from your data science team

Providing the business context your data science team needs to work efficiently and provide the deepest possible insights into solving business problems isn’t as simple as giving them information about the business. It needs a three-pronged approach that ensures that your data science team is fully integrated into business activities.

  1. Get business buy-in

While organizations are drowning in data, without context it can’t provide all of the information a data scientist needs to be an effective part of a business. The data science team can’t be used to its full power if it exists in a bubble separate from other departments within the business.

The first step in taking the data science team out of the bubble starts with the business fully embracing data science. Data science cannot be seen as a service function, it needs to be fully integrated into the business which starts with the company and its executives buying-in to data science as an integral part of the company.

  1. Provide the bigger picture

Your data science team needs to understand the role they play in the business not just on a day-to-day basis but in its long-term goals. This goes deeper than bringing data science into business processes, it’s about making sure that your data science leader is involved in business decisions that affect both short-term and long-term activities. With this information they’re fully aware of where the business is going and the risks it’s exposed to, which means that the data science team are prepared to answer the right questions to help the business mitigate risk and reach its goals.

  1. Create communication within the organization to provide in-depth context

Even if your data science team know the right problems to solve, they’ll need a thorough understanding of business processes to identify suitable solutions. To find the right answers they’ll need to understand:

  1. The business’ goals
  2. The marketplace it operates in
  3. The processes that power it
  4. Its products
  5. Its customers
  6. The future roadmap
  7. The risks these things expose it to

Again, to facilitate this process you need to ensure that your data science team collaborates with other key teams, including the risk team, finance team, etc. In the webinar Schultz shared the surprisingly simple solution Elevate uses to ensure that its data science team is exposed to and understands business processes. “At team meetings every week, we have 30 minutes assigned for a guest speaker from strategy, finance, or marketing to come in and talk about what they’re working on," he said. "And what I’ve found is that if you can get your team and your data scientists to really understand the business, and to know where they’re going, to know what other people are working on, they’re way more effective."

The old idea of a data scientist is someone who sits in a corner, just hacking away at the keyboard is not going to allow the team to be as successful as they can be. Our data scientists sit with our risk analysts, sit with our fraud team. Fostering collaboration is huge and in forcing that collaboration in some sense, people will open up to it and really start to enjoy it. And everyone gets better for it.”

To successfully design something, you need to understand how it fits and what its role is in the bigger picture. It’s up to a business to ensure that its data science team understands everything they need to know to help the organization succeed in today’s turbulent financial markets.

Become a bobsguide member to access the following

1. Unrestricted access to bobsguide
2. Send a proposal request
3. Insights delivered daily to your inbox
4. Career development