Problem statement

The client is one of the leading healthcare providers in North America. The company was looking to expand branches in 5 provinces in Canada and wanted to identify potential zip codes with their target customers.

Their target customers included people above certain annual income, age group, population level and other such parameters. Other than the target customers, they were also looking for availability of skilled resources with certain ethnicity in the zip codes.

The selected zip codes should also ensure presence of their competitors and other infrastructures required to function. The scope also included forecast of ROI for each with respect to selected parameters.

Our Solution

The approach was to identify homogenous zip codes which satisfy the criteria of their target customers. The Demographic dataset of five provinces in Canada was prepared from various sources like in house data, Census data and third party data providers.

Segregated the geography at Zip code level and checked for variable importance using Random Forest Algorithm. Then unsupervised learning technique (Algorithm-K Means clustering) was used segmented homogeneous population within the dataset of heterogeneity. From the selected cluster, we concluded 5 geographic ZIP Code areas which satisfy all client requirements.

Result & Value Adds

• Decision making time reduced by 70%
• Increase in marketing ROI by 25% with respect to previous initiatives
• Narrowed down to 5 zip codes from a data of 900,000 zip codes
• Predicated ROI in each new branch using customer behaviors in each zip codes