About the Client
The client is a leading retailer of Dental Savings Plans. The company offers more than 40 of the leading regional and national discount dental plans with more than 100,000 participating dentist listings in combined networks across the country.
As a dental insurance retailer, the client generated prospects through TV Ads, Google Search, and Newspaper Ads etc. With millions of leads accumulated over time, the client needed a solution to contact leads which have a greater chance to convert. The current practice was to use certain prioritization techniques such as LIFO (last In First Out), TV Ads first (since they were considered to be hot leads) etc. With ever increasing number of leads, this prioritizing alone was becoming impractical and did not yield expected results.
Contacting existing members for renewals was another challenge, an effective way that could maximize renewal rate was required.
Need help with an Azure Machine Learning project?Contact Silicus Sales
Silicus developed machine learning algorithms to overcome above challenges. Machine Learning algorithm implemented in Azure uses historical lead attributes vs conversion data to come up with a strategy to determine lead conversion probability.
- Silicus implemented an algorithm which uses 10 different attributes of sales leads to determine probability of lead converting into a member
- After acquiring the data, we performed data cleansing and transformations to ensure that the data can be fed into supervised learning algorithms. This was initially done in RStudio and then in Azure ML.
- Based on the available data, Silicus built supervised learning algorithms to classify lead conversions and estimate the class probability
- Silicus leveraged Azure ML supervised machine learning algorithms, that allowed the team to implement and compare many different models on the same data easily and efficiently as opposed to a single model scenario
- The team compared several different algorithms, including the most popular supervised learning algorithms for scoring – i.e., logistic regression and decision trees
- Azure converts the algorithms to a web service automatically which can in turn be integrated into the sales portal for actual use
Original data Source
Data transformation and cleansing
Machine Learning Algorithms
Customer Profiling and Targeting
Achieved approx. 73% accuracy for converted and non-converted customers, which are the classes of interest.
Improved Sales Efficiency
Predictive nature of the algorithm helps in identifying hot leads. Having concentrate contact effort for these sales leads makes way for better conversion rates, reduced sales personnel time and ultimately more revenue
Self-Sustaining & Self Improving Solution
Machine learning algorithm is self-improving algorithm which will correct itself as more data is being fed into it. This will result into more accuracy on predictions.