Wilcompute's data science practitioners enable an industrial distributor to gain insight into their opportunities.
Machine learning can be leveraged to identify insights in data that would be difficult for a human to identify. Wilcompute's data science practitioners have experience analyzing large data sets to pull out actionable business intelligence. This experience was put to the test by a large international distribution client that wanted to understand their opportunity pipeline.
Traditionally, managers determined priority and work effort in the opportunity funnel for their 219 team members through a combination of opportunity revenue, key account status, or a gut feeling on win percentage. The problem being, how do you know you're spending time working on opportunities that will actually pay off? If these opportunities don't successfully convert, how should you plan the business and forecast revenues?
Wilcompute hypothesized that this problem could be solved if only there were a more accurate method of predicting opportunity outcomes. In the hopes of surfacing properties that contribute to a win or loss, we put this theory to the test by applying a machine learning algorithm to the vast quantity of data accumulated in the client's CRM: 210K historical opportunities, 480K customer visits, and 3.2M sales records. At this scale, even basic activities require heavy computational and processing power.
The data analysis identified a number of key insights. For example, most user contributed data, such as opportunity confidence level percentage, were not statistically significant. So much for gut feeling! Opportunity criteria such as the number of days open, time since last contact, and number of active opportunities were more reliable indicators of success or failure.
Once we identified the most useful statistically significant input factors (18 of them) we deployed a neural network algorithm on thousands of historical examples to learn what contributes to a win or loss. The experiment was a success, resulting in predictive accuracy rates of 75-85%. Wilcompute deployed this trained model to the client's business intelligence dashboard to clearly mark current open opportunities with a future win or loss flag. We store predictions as they are made (570K and counting), allowing us to continuously measure how previous predictions stack up against reality.
The client is now able to use this predictive data for business planning. Managers can decide if they want to spend energy on consistently unsuccessful opportunities, and pipeline calculations are simpler with the addition of an opportunity outcome confidence level.
Leverage Wilcompute's data science and machine learning practioners for your own business problem. Get in touch to discuss how we can help.