Lies, Damn Lies, and Data Visualization
Predictive Analytics in Healthcare: It's Not Happening
Predictive Analytics Key Component of Customer Experience Management
Predictive Analytics in Higher Education
Trends in Computational Healthcare Research and Digital Therapeutics
Damion Nero, Director of HEOR Research Analytics, Cardinal Health
Transform Decision Making with a Data-Driven Culture
Kamelia Aryafar, Chief Algorithms Officer, Overstock
Thank you for Subscribing to CIO Applications Weekly Brief
Smarter, Faster Outcomes through a Decision-First Approach to Analytics
By Joe Decosmo, Chief Analytics Officer, Enova International
At Enova International, advanced analytics is core to our business. In the four years that I have been Chief Analytics Officer, we have grown the analytics team to over 60 experts in applying predictive and prescriptive analytics to multiple areas including fraud detection, credit risk management, and customer retention. What has driven this growth? Our decision-first approach to analytics. For many businesses, the application of analytics stops at turning big data into insights. For Enova International, analytics is being used to transform big data into action through an approach we call digital decisioning. Since 2016, we have been helping businesses achieve similar outcomes by bringing the same analytics expertise and decisioning technology through our Enova Decisions™ brand.
Digital decisioning is the process of automating and optimizing repetitive but complex decisions that impact how your business interacts with your customers in order to acquire better customers, improve operations, and maximize profitability. In order for this to occur, analytics must work together with business and technology teams.
Elements of a digital decision include the following:
• Business rules: This may include compliance and legal, and they are typically managed by business teams.
• Predictive models: These determine what’s likely to happen given a set of inputs, whether internal or external data and they are created and maintained by analytics teams. These models can be simple regression models or machine learning algorithms which are becoming more popular.
• Optimization: This ensures that you are driving the best set of decisions. They are a set of constraints typically determined by the business teams such as volume or cost and are built into the decisioning layer by analytics teams to optimize the business outcome given these constraints.
• Automation through digital decisioning technology: This is a system that allows you to do everything on-demand and potentially in real-time through software and business integration and is developed and monitored by technology teams.
For many businesses, the application of analytics stops at turning big data into insights
Making the most of digital decisioning to achieve digital transformation requires not only organizational agility from business, analytics, and tech working together but also moving operations into the cloud, continuously building up analytics expertise, and automating more operational decisions over time. However, decision automation does not need to be “all or nothing.” Here are some guidelines to help you determine which decisions are the best candidates for automation.
First of all, you need to understand the type of decision that is being made. In business, there are three types of decisions made on a regular basis: strategic, tactical, and operational. Strategic decisions are complex decisions made by business leaders such as whether to launch a new product or go into a new market. Tactical decisions are complex decisions made more frequently than strategic decisions by lower level managers such as product features or eligibility rules. Operational decisions are simple or complex decisions made routinely, day-to-day by employees or machines such as verifying identity of a customer. Due to the repetitive nature of operational decisions, they are the best candidates for decision automation.
Second of all, you need to understand the risk that is involved in that operational decision. At Enova International, we’ve lent over $20 billion in credit to over five million customers worldwide through our nine online consumer and small business lending brands in the past 14 years. We’re experts in credit underwriting. However, not all of our credit decisions are automated. In cases where there is not a clear “approve” or “deny,” or in small business where we’re dealing with large loan sizes, we will require manual intervention. Even so, the operational decisions prior to the credit decision such as identify verification and fraud detection is automated. This ensures that our Customer Service Representatives are focusing their work on the most impactful cases.
Now that you have a framework for determining what decisions to automate, here’s the million dollar question: to buy or to build? The short answer is that it depends. If you have the analytics and technology expertise in-house, building may be a viable option. If you do not, buying makes the most sense. Regardless, here are some considerations when deciding on a digital decisioning system.
First, realize there’s a tradeoff between efficiency and complexity. Typically, a more complex system leads to more efficiency. However, you need to factor in the time, money, and resources that are needed to integrate this system with your existing systems. In addition, there will reach a point where increased complexity will not gain additional efficiency.
Second, understand the latency of your decisions. The more on-demand and real-time a decision, the more complex the engineering is to make that decision. In addition, real-time, as discussed above, may increase business risk. Therefore, having a clear sense of what decisions should be done in offline; batch processing versus real-time can help you manage complexity and cost.
Third, remember that your outputs are only as good as your inputs. Therefore, a digital decisioning system that enables you to integrate with multiple internal and external data sources will be far superior. However, importing more data will drive up your costs. A smart system creates decision waterfalls that allow you to pull only the data you need, when you need it so you can minimize the cost of the decision.
Finally, evaluate a digital decisioning system by its ability to allow for optimization. Testing is what will enable you to make smarter decisions over time. At Enova International, we believe there are no failures–just results. I encourage my team to think like scientists: create a hypothesis and test that hypothesis. A properly designed system should enable you to consistently test and improve your decisions without disrupting operations.