Vertafore, an insurance software and services provider, has launched a new risk retention prediction tool to help independent insurance agencies improve their client retention rates. The analytics tool, called RiskMatch Retention Prediction, is an integrated part of Vertafore’s RiskMatch platform. It uses data modeling and adaptive machine learning to identify policies at risk of cancelation or non-renewal, thus equipping agencies for more proactive client engagement and portfolio management.
According to aggregated data from 3,700 agencies in Vertafore’s RiskMatch platform, the average client retention rate for independent agencies across the United States is 83%. This means agencies must replace 17% of their clients each year if they want to maintain their business at the same level – a task that requires a lot of time and money. With the new RiskMatch Retention Prediction tool, Vertafore hopes to reduce that client turnover – sometimes called ‘churn’ – by giving agencies insights into historical retention trends and mapping those insights alongside upcoming renewals.
Scott Ziemke, director of data science at Vertafore, explained: “At Vertafore, we process a significant quantity of industry policy data through our agency management systems (AMS) and through RiskMatch. With RiskMatch Retention Prediction, we leverage that extremely large data set of policies to learn what factors or what features about a policy contribute more or less to its risk of retention. The machine learning model then learns from the past, and we can use that data to predict retention in the future.”
The first version of the model was released in August 2020. It gives insights down to a policy level, ranking renewals as low, medium or high risk of churning. According to Vertafore, this enables agencies to focus their retention efforts on customers with the highest savable value to maximize resources and business growth. The model also shows client retention trends by carrier, product, geography, industry sector, office location and agency rep, helping agencies identify opportunities for client outreach, training or other solutions.
“We’re currently working on a second version of the model that we aim to release before the end of the year, which will focus more on the prescriptive corrective actions that agencies can make to improve their client retention,” Ziemke told Insurance Business. “At the moment, the model analyses over 250 internal and external factors across market, client, and agency data to produce the policy risk retention score of low, medium or high. Next, we want to provide more insights as to what caused that risk score. If a policy is identified as high risk, we will indicate what features caused that risk so that agencies can make corrective action.”
With Vertafore’s RiskMatch Retention Prediction tool, the value comes from both the risk prediction and from the historical causes of non-renewal. Ziemke explained: “We hope that agencies will look at the information we provide with a broad strategic lens, to say: ‘Historically, we’ve struggled in this area. Let’s think strategically about how we can make changes, either to our markets or to the way we service our clients.’ With the new feature we’re releasing at the end of the year that will allow agencies to look at a broader picture of the causes [of client churn]. For example, if they have 100 policies that are high risk of non-renewal in the next 120 days, they will be able to look in aggregate at the causes [putting those 100 policies at risk] and potentially act strategically to change something at a broader scale rather than on a policy by policy basis.”
The risk retention tool uses machine learning capabilities to adapt to current market and industry factors and deliver more accurate future-looking insights — such as how major events like a global pandemic might impact future retention. “Our model is continually retaining itself based on the latest data available,” Ziemke commented. “As we learn more about the impacts of COVID-19, we will capture that data and use it to predict forward. Obviously, we couldn’t predict back in January that COVID-19 was going to happen, but over the past few months, we have been able to capture data and learn from the immediate impacts of COVID.”
