A Fortune 500 insurance, banking, and financial services company
Success Story
A century-old insurer automates demand forecasting, and improves policy development and claims management

The Client
Profile: An insurance, banking, and financial services company with 12 million members, operating in all 50 states
Structure: Owned by policyholders
Year founded: 1909
Yearly revenue: $8.7 billion

Predicting how consumers will use an insurance policy is critical in serving policyholders and containing claim rates.
Like most insurers of its caliber, this company has an expert analytics team with a successful track record in predictive modeling. But after claims for a particular policy significantly surpassed anticipated volumes, the team wanted to learn more about holders of the policy, understand their reasons for initiating claims, and equip decision makers with the insights needed for change.
In any industry, long established statistical methods, long accepted domain knowledge, and long standing drivers are important — but they’re not the whole story anymore.
This company’s analytics team had been delivering reliable policy intelligence using yearly historical claims data and linear modeling. These don’t capture the monthly movements in macroeconomic or claims activity that reflect shifts in the population, the environment, and the healthcare space.
The team was asking, who is driving the increase in claims? What do they have in common? And how can we better plan for their behavior in the future?
The Nousot solution: Demand Forecasting
Nousot partnered with this client to customize a Demand Forecasting model that more accurately predicts claims against the policy over a 12-month horizon, and identifies the drivers of those claims.
Client data used
The previous year of claims data consisting of basic policyholder demographics and policy-related activity
How we enrich it
Nousot combined internal claims data with hundreds of GlobalPulse data sets related to medical insurance, hospital visits, and other healthcare activity, deepening the company’s policyholder knowledge. Then we used our autonomous clustering engine to identify similar policyholders and create 15 different member groups based on claims patterns.
Deliverables
An in-production, autonomous demand forecast that predicts overall policy usage and learns from internal and external data, reducing the forecast error rate to an average of 2-3%
A new policyholder segmentation scheme that more precisely groups members for targeted treatment by all business functions such as underwriting and marketing
A custom typing tool for entering and properly segmenting new members
An autonomous, per-cluster demand forecast for each of the 15 new member groups
Timeline
1 week
to source client internal data
1-2 days
to produce initial forecast and clustering models
6 weeks
to iterate with client on curating GlobalPulse data and using it to enhance, stabilize, and validate demand forecasts
2 weeks
to develop and test the typing tool

Business outcomes
Predictive policy intelligence, updated monthly, regarding expected usage of the product over the next 12 months
The ability to make proactive adjustments to policy elements and policyholder treatments in order maintain high standards of member service and claims management
Increased departmental collaboration via detailed, actionable, and shareable member segments
An enhanced, more productive analytics team without a single new hire