The Cisco seller process works with digital seller information reports called ‘opportunities’. When a seller has contacted a customer and determined they are potentially interested in purchasing products or services, a seller creates an ‘opportunity’. An opportunity has a 5 stage lifecycle. A question sellers typically have is, ‘which opportunities are the most likely to go through the entire set of 5 stages to booked’. Given this information, a seller would be able to prioritize their focus on the opportunities that need their focus, allowing them to book more opportunities.
Solution:
I developed an end-to-end, automated spark pipeline for gathering multiple pieces of data from salesforce, and other Cisco business information, and developed a highly accurate predictive machine learning model to predict the likelihood of a customer seller leaving. This model was integrated into the seller process, and changed the way Cisco sellers approached opportunities to improve the data.
Methods:
Using large historical salesforce data, and adaptive basis function classification models, along with probability calibration methods.
