- Customer Relationship Management (CRM) systems have only recently gained wide-spread adoption. CRM systems store key data on each sales opportunity and are the foundation for sales data analytics.
- Many salespeople still rely on their familiar “tried and true” process of spreadsheet roll-ups and “gut feels” to make forecasts. Senior salespeople develop several rules of thumb to predict their quarterly sales number.
- Change is difficult, especially in such an important area. Relying on a statistical program to make estimates without fully understanding the details is a significant change for most sales organizations.
What is especially interesting about the situation today is the CRM system was supposed to become the foundation for sales opportunity management and forecasting. CRMs have done a great job on the opportunity management side by tracking sales activities and providing feedback and follow-up support to the sales team. However, the forecasting quality has many challenges primarily due to poor discipline in data input and a tendency for salespeople to “game” the system to ensure they can meet quarterly commitments. The difficulty level increases in forecasting as the length of forecast increases and several studies have shown when predicting on a deal-by-deal basis at the start of a quarter the average accuracy rate is less than 50%.
Most companies today try to cope with these issues by investing significant time and resource in the Sales Operations team who adjust for the various issues, typically in Excel to enhance the forecast. However, there is a significantly better way to achieve accuracy and objectivity in your Sales Forecasting. The answer is machine learning. Machine learning is a tool used by data scientists to make predictions about the future. This technique uses historic data to train a model and then makes predictions about the future when new data is observed. This technique is great for making predictions about whether a sales opportunity will win or lose. There is a wealth of historic data in the CRM that can train a model, so when a new opportunity is observed the model can predict if it will result in a win or a loss. The other clear benefit of this approach is that it is totally objective as it relies on the actual historical outcome of sales opportunities and it is not influenced by any salesperson bias.
Most companies today can take advantage of a machine learning approach to sales forecasting. This process can run in parallel to the sales team’s existing process and over time the reasons for lack of adoption will confirm the benefit of using this new process. Given the rich data contained in the CRM the machine learning system can leverage this data and will track all opportunities through time. It will also track each time the Stage, Close Date, or Amount change and create a new observation for that opportunity. Other data attributes like how long has an opportunity has been in its current stage, how many days until the end of the quarter, the dollar value of the opportunity, and many others. In all, we use a minimum of 19 different data attributes to make our prediction of won or loss in our machine learning models.
By using machine learning to make predictions about sales opportunities, we are able to predict if an opportunity will win or lose the first time we ever see it with 85% accuracy. This is a significant improvement over the less than 50% industry average noted above. In addition to the benefit of getting a more accurate forecast for the quarter, you will also get an objective list of deals that your sales team should focus on for the quarter based on the higher probability to win. If you would like to know how you can use machine learning to predict the value of your pipeline or would like information about ORM Technologies, please let us know at info@orm-tech.com.