How weather data helps manage call volume during catastrophes
Carriers must leverage weather data for First Notice of Loss call volume management to create a better claim experience for insureds.
First Notice of Loss (FNOL) is the backbone of claims handling, and directly affects a policyholder’s experience with an insurer.
One of the top challenges insurers face today is increasing the efficiency of their FNOL processes, especially when insurance contact centers experience spikes in call volume during adverse weather events. The size of these spikes can vary depending on the intensity of the adverse weather event.
When these call-volume spikes aren’t predicted and planned for in advance, they can cause hold times to go up and increase call abandon rates. This can have a considerable impact on the carrier’s brand image. The variability in the complexity of the FNOL calls further complicates contact center management when catastrophic events occur.
Forecasting a better customer experience
For insurers, it is crucial to build an effective forecasting model to predict the inflow of claim-related calls. The projections from these call volume forecast models equip carriers to optimally utilize their contact center resources, which could result in big savings and loyal, satisfied customers.
For instance, consider the case of Hurricane Michael Emergency Order (8-Oct-2018), which led Citizens Insurance to send a formal notice to its vendors for catastrophe support. As a result, after completing six training classes, 120 new employees were added to Citizen’s program between Oct. 9, 2018, and Oct. 10, 2018. On Oct. 11, 2018, 147 Customer Service Representatives secured and trained for Citizens’ Hurricane Michael response efforts.
That month (October 2018) witnessed a 175% increase in call volume (12,697 calls), which comprised mostly of FNOLs and status calls. The average policyholder wait time was clocked in at 8 seconds.
Although this carrier was swift in its response to the catastrophe, not every insurer is lucky enough to have optimum contact center resources to handle the increased call volume during such an event. Therefore, it is essential for all insurers to include additional dimensions such as weather forecasts and warnings as well as data modeling services in order to build a framework for handling weather-related claims effectively.
The current FNOL process
First Notice of Loss is the best opportunity for insurers to provide a differentiated claim service that creates a lasting positive impression. Effective, efficient notice of loss combined with fast-tracked claims handling during FNOL can transform operations while allowing insurers to realign claim resources with more complex claim-handling activities.
Insurers can supplement their FNOL call volume forecast modeling solutions using time interval-based historical FNOL call data, along with weather information and other dimensions.
Several insurers are already engaging with analytics firms on this to help them optimize their FNOL call management process. One U.S.-based insurer, for example, realized that relying solely on univariate time series call volume data may be insufficient to arrive at accurate future call volume forecasts.
Forecast model success
Insurers can have short-term and long-term goals in mind while building the forecast models. For instance, if adverse weather conditions play a significant role in a particular scenario, then the most accurate dimensions of weather information should be explored and eventually used in models.
The following two factors determine the success of forecast models in weather-specific catastrophic events:
- Precise, location-specific data: Insurers that send timely alerts and accurate weather information help policyholders take preventative measures to mitigate damages and reduce claim volume. Carriers can better equip their contact center resources based on precise weather information. Consider, for example, a scenario in which a hailstorm occurs in an unexpected area. The carrier’s contact center may remain understaffed or ill-equipped to capture the claims information due to an unexpectedly high volume of FNOL calls.
- Data coming in a timely manner: Timely information about shifts in wind direction, changes in temperature and precipitation, and changes in the intensity of a catastrophe can positively impact the claim handling process if the contact center agents follow the right scripts to handle the updated forecasts and predicted FNOL call volume. This enables insurers to gain a better view of the claims arising out of a weather catastrophe.
Multidimensional call-volume forecasting
The first step toward improving FNOL responses using weather data involves analyzing call volume data along with other dimensions. Multidimensional forecasting is an approach that develops forecasts based on historical call volumes along with additional dimensions. Additionally, time-series forecasting takes into consideration missing call data, outlier cases such as the impact of holidays, and other factors.
When choosing the right weather data to add as additional dimensions in the call volume forecast models, several variables must be kept in mind.
- The data should contain common metrics including temperature, sunshine, pressure, humidity, cloudiness, visibility, and other information
- The data contains adverse metrics such as heavy rain fall, heavy snowfall, storms, and other events which impact FNOL call volumes
- The weather data is accurate in terms of forecasting weather metrics and has been validated based on historical events captured
- Future forecast data should be available for the required future forecast time period
Insurers can also develop two separate models for long-term and short-term forecasts. These can include additional dimensions such as weather-related information, forecasts intervals of 15 minutes, 30 minutes, hourly, daily, or other time periods based on the nature of the data, and use different time series modeling techniques like triple exponential smoothing, ARIMAX (Auto Regressive Integrated Moving Average with exogenous variable), or other approaches. Both traditional and machine learning time series model algorithms can be explored for this type of forecast problems.
However, using traditional forecasting techniques over machine learning techniques have disadvantages including:
- Triple Exponential Smoothing: Forecast models using this technique are easy to build, but there is a great risk of over-fitting for historical anomalies such as outages
- ARIMA(Auto Regressive Integrated Moving Average): ARIMA cannot detect multiple seasonality in the data. This limits the use of this method of forecasting for long-term trends on high-frequency data.
The role of machine learning
Machine learning techniques avoid the pitfalls of traditional time-series forecasting. Two such highly effective techniques for call volume forecasting are deep learning time series (LSTM) and seasonal ARIMAX:
- Deep Learning Time Series-LSTM (Long Short Term Memory): This approach is used for high frequency and long-horizons. It can be difficult to implement as it uses complex network structures like RNN (Recurrent Neural Network) and CNN (Convolutional Neural network), and choosing the optimal parameters is not simple.
- Seasonal ARIMAX (SARIMAX): SARIMA supports the direct modeling of the seasonal component of the ARIMA series. When this method also takes into account additional exogenous dimensions it is called SARIMAX. The optimal hyper parameters for this model can be decided based on Auto ARIMA or grid search framework.
Additional measures to optimize call volume management
Along with optimizing contact center resources, insurers can increase FNOL efficiency by:
- Enabling Digitization: Multi-channel intake and claims routing can influence cycle time and claim outcomes. An end-to-end omnichannel claims experience helps carriers to cut costs and settle claims quicker. Digital FNOL increases the volume of claims intake through digital channels and increases efficiency in managing the claim process. As a result, the contact center staff has relatively fewer FNOL call volumes to handle and can focus on capturing more complex claims.
- Data-Driven Segmentation: Analytics-driven segmentation rules can quickly assess whether a claim is complex. Segmenting claims based on complexity can help insurers manage the FNOL process by reducing effort and increasing customer satisfaction. The faster a complex claim is detected, the quicker it can be assigned to an adjuster for appropriate care and handling, leading to better mitigation efforts and increased customer satisfaction.
The opportunity for insurers
Insurers can work with their internal analytics department and analytics vendors to augment their FNOL call volume management capabilities and manage their FNOL operations for a faster, more accurate claim resolutions by preparing for high call volume scenarios. Smooth FNOL call volume management not only improves customer satisfaction among policyholders but promotes loyalty and creates a scope for cross-selling other relevant policies relevant. From an enterprise angle, it introduces a sense of agility in an insurer’s claim handling processes that eventually leads to a reduction in the number of claims.
Soumyajit Dasgupta is assistant vice president and engagement manager at the insurance analytics provider EXL. These opinions are the author’s own.
To find out more about using data to manage First Notice of Loss call volume, call Vivek Gupta, EXL’s vice president of Insurance Analytics, at (848) 236-8185 or send email to Vivek.Gupta@exlservice.com.
See also: