The Objective
For an insurance company shipping thousands of devices, a shortfall in inventory can have damaging financial implications including increased cost to obtain more inventory to service unexpected demand at a lower margin or decreased claim fulfillment times. Synechron recognized macro-market events such as a new phone launch result in an influx of phone trade-ins and purchases create cascading impacts on inventory needs to meet future claims. If the supply chain management team were able to look at historical launch data, claims, and re-enrollment records, it could better forecast potential future upgrades on a per customer basis, and therefore, enhance inventory planning and management.

The Solution
Synechron created a data lake and used a combination of Data Science and Machine Learning techniques to analyze historical policy data for 100K customers with almost two years of data available combined with existing data on how many subscribers have upgraded already. The models analyzed this data using a variety of Binary Classification models with supervised learning, including: Decision Tree, Logistic Regression, Naive Bayes, Random Forest, and Support Vector Machine (SVM).

The models examined additional data including the age of the device, and customer demographic data such as phone provider affinity (eg. an Apple loyalist), personal phone vs. business phone, and small vs. large city (across 100 cities) to predict the customers most likely to upgrade to a newer model.

Using 75k customer records for training and 25k for testing, Synechron was able to predict most accurately using the Random Forest algorithm that the age of the current device and loyalty toward Apple were the most important factors to predict the likelihood of an upgrade when a new Apple phone model is launched – with .94 ACC, .92 Precision, .93 Recall and .92 F1.

Benefits

• Data-driven ability to manage risk, inventory, and profitability
• Enhanced business intelligence related to customer behaviors and buying trends
• Target audience likely to upgrade with new policies
• Prediction of revenue impact related to new phone model releases with extendable models
• Enhanced inventory management
• Savings in inventory cost
• Preventing out of stock situations