The objective of the initial project phase was to identify the prospects, who were most likely to purchase payroll services. During this project, I was responsible for gathering the business requirements, understanding the business and the sales process as well as building the predictive models and setting up an architecture that allowed for the regular, automated scoring of prospects and the integration of the scores directly into ADP's cloud-based sales automation system.​

The project included data from several different and disparate data sources. These were both internal to ADP - such as the cloud-based sales automation system - and external sources such as data feeds from Dun & Bradstreet and InfoUSA to include information about new and existing small businesses.

The prospects were first assigned to different segments based on the sales channel (telesales or field sales) and the information which was missing for each of the customers. We were working with small and often newly opened businesses and it would sometimes take some time before all of the relevant information would make its way into the various external data sources. The propensity models used to evaluate each prospect's likelihood to purchase were ensemble models using as many as 11 different algorithms and would be trained separately for each of the customer segments.​

After this customer acquisition solution was deployed and the sales associates started planning their sales calls and visits based on the estimated propensity to purchase, it was demonstrated that the top 5% of the prospects were about 5x more likely to purchase than the average prospect. This was a massive boost to the efficiency of the sales associates and it gave ADP a real competitive edge over the other providers in the market.​

Following the initial project, I was able to work closely with the project team and the stakeholder's at ADP and convince them to extend the project to extend the solution to also predict the service bundle that each prospect would be most likely to buy.​

Because of the proven success of the project and the tangible impact that the results had on the business, ADP created their own Center of Excellence for Analytics and I helped them with setting it up and with hiring people with the correct skills for the team. They went on to develop solutions similar to the original project for their Mid-size business segment as well as their National Accounts using a combination of in-house and IBM resources.​

The project was prominently featured at the IBM Information On Demand conference in both 2010 and 2011.

”Kenneth worked for me when I was at ADP. We brought him in to assist with our predictive analytics project for customer acquisition. Within a very short period of time, he got up to speed on our processes to understand our business model and our sales model. He developed a comprehensive model to help our sales associates identify those prospects that were most likely to purchase ADP products and integrate those results in our sales automation tool. I highly recommend Kenneth not only for his expertise in analytics, but also his excellent work ethic.”

Timothy Barnes
Director, Enterprise Applications, Business Intelligence
ADP, Inc.