THE PUNCH LIST Step-by-step guides for knockout results Implementing Analytics Guide 3 steps to better underwriting profits
KEYS TO 3ANALYTICS Adding Science to the Art of Underwriting Implementing predictive analytics in an underwriting organization is a significant undertaking. Many companies focus primarily on selecting the right predictive model. In reality, the model itself is one part of a larger process that touches many parts of the organization. Whether you are developing a model in-house or using a external firm, there are components common to every predictive analytics project that are necessary to ensure success. This guide will help you: 1. Make a solid case to your executive team 2. Prioritize, plan and get buy-in from the organization 3. Prepare your organization to execute
Answer Why First 1. Senior Level Commitment There should be no more doubt that predictive analytics are table stakes for accurate underwriting and pricing models. The two most important questions are how best to use them and why. The Punch List: 3 steps to implementing analytics Let s be real information is a business enabler, and any data analytics project must produce meaningful insights that will solve particular problems and achieve specific objectives. Make sure all the relevant stakeholders understand the business goals from the beginning and that you have secured executive commitment and sponsorship. This holds true whether you are the CEO or a department leader. Questions to answer: How will we prove that a predictive model will produce results? What is our proof of concept? What are the agreed upon metrics to measure success? Loss ratio, price competitiveness, premium growth? What management reporting will you put in place? If it can be measured, it can be managed. How do we know if a predictive model is giving us new insights vs. telling us what we already know? What is our risk appetite for this initiative? What are the assumptions and sensitivities in our model and how will those impact projected results? What is the plan to integrate the model within our existing workflow?
Will They Follow? 2. Organizational Buy-In Data analytics can only be successful if developed and deployed in the right environment. You have to retool your people so that underwriters don t feel that data analytics are a threat to their expertise, or actuaries to their tried-and-true pricing models. Following a thoughtful, straightforward process that involves all the stakeholders early and often goes a long way. 82% of P/C execs say that underwriter adoption is a big concern when deciding to implement predictive analytics. Source: Valen Analytics 2015 Summit Survey Build a solid plan Make the Case Outline the business problem you are solving with a simple, compelling story. Your people will support change with a solid road map and clearly defined success metrics. Manage Culture & Process Seek the appropriate influencers who can help communicate and provide feedback. Start small, build on early wins, and bring the team along in phases. Have written processes in place to empower your employees to take action. Create Transparency A predictive model shouldn t be a black box, explain what data went into the model. Underwriters who can interpret the predictive scores are better able to guide discussions with agents and policyholders, and make more informed decisions. Play for the Win The devil is in the details. A structured training program is critical. When things fail, it s rarely because of faulty algorithms and almost always because the implementation went awry. Provide Support Have a clear process for answering technical questions like who will respond when the scoring engine/database isn t working correctly. Adjust Over Time Rolling out anything new takes time. Enable a test and learn culture, allowing a couple months to iron out the kinks. Then assess whether process changes are needed. TThe Punch List: 3 steps to implementing analytics
Have the Right Ingredients? 3. Assess Resources & Capabilities A predictive analytics initiative typically takes one of these paths: 1. Developed in-house, hosted in-house 2. Developed by consultant, hosted in-house 3. Developed and hosted by modeling firm Regardless of whether the data analytics project will be internally or externally developed, your assessment should be equally rigorous. Data Considerations Sample Size - Having ample data to build an accurate model is particularly relevant for commercial lines. Selection Bias - Risk appetites and growth/retention strategies skew the data. How will you round out the data to address biases? Blind Spots - If growing into new markets is a priority, thirdparty data can fill in the gaps. Modeling Best Practices Data Custody - Establish a process to standardize, normalize data. Data Partitions - A/B testing is not sufficient and can result in erroneous conclusions. Four partitions are recommended. Model Validation - Is it predictive? Accurate? Is it better than what we have now? Model Type - Identify how the model will be used - is it an automated use case or will human beings need to understand the results? Time to Market - Market conditions change. If it takes 18-24 months to deploy, it will have lost significant business value. IT Resources Scope & Priority - Evaluate the scope and where this initiative fits in the queue. Deployment Tools - Do you have a deployment platform? Can it be incorporated into the existing u/w workflow? The Punch List: 3 steps to implementing analytics
wins to gain momentum. Involve all the relevant stakeholders along the way and find internal champions to share your progress. Recognize that whether you are building a data analytics solution internally, hiring a solution provider, or doing some of both, there are substantial costs involved. Having objective criteria to evaluate your options will help you make the right decisions, and arm you with the necessary data to justify the investment down the road. What is the best way to get started? The most important first step is to make sure you have aligned your data analytics initiative to a strategic business priority. Once you do that, you will be able to garner the time and attention required across the organization. Make your first steps doable and measurable. Define a small pilot project, test and learn, and create early THE PUNCH LIST RECAP 1. Secure senior level commitment 2. Create a plan to gain buy-in 3. Assess your resources and capabilities Don t boil the ocean - just get in the game! The Punch List: 3 steps to implementing analytics
Valen Analytics is an advanced data and analytics provider for property and casualty insurance companies. We work with insurers who are actively looking to improve underwriting profits by driving growth, lowering loss ratio or creating efficiencies. If you are focused on increasing competitive pressures and adverse selection, an experience gap with your tenured underwriters nearing retirement, or scarce IT resources to move your initiatives forward, we can help. Our customers span many lines of business including Workers Compensation, Commercial Auto, Commercial Package, Commercial Property, BOP, and Homeowners. Learn more online at valen.com. 800.280.3304 I valen.com 1730 Blake Street, Suite 300 I Denver, CO 80202 Copyright 2015 Valen Analytics. All Rights Reserved.