Using activity data to improve opportunity scoring predictions in Dynamics 365 Sales

Listening to the challenges many of our customers are facing with increasingly working online, we’ve developed new capabilities in Dynamics 365 to discover email messages, meetings, and phone call activities related to an opportunity, and then to factor these signals into the score.

Capturing activities to determine opportunity health

A healthy opportunity will usually have activities happening in it like email messages, meetings, and phone calls. Many of our predictive opportunity scoring customers were gathering these signals manually to determine opportunity health. We’ve replaced the need for customers to manually track activities by using AI to gather activities, approving prediction accuracy and compensating for cases where sellers do not diligently update the opportunity fields.

There is additional value in activities – they provide actionable explanations for opportunity scoring predictions. It can help the seller make more educated choices about what the best next action might be – for example, a high level of activity may indicate that the opportunity is heating up and is worth attention. Likewise, an opportunity with where the activity level is low may need immediate attention.

Here’s an example showing the opportunity score improving:

Screenshot showing predictive opportunity score improving

Here’s an example showing the opportunity score declining:

Screenshot showing predictive opportunity score declining

Connecting opportunities with activities

In addition to activities that are explicitly connected to opportunities, we developed algorithms to infer indirect connections between activities and opportunities, using Contact and Account activity timelines. Since these are not trivial connections (for example, an account may have multiple opportunities), AI is assigning them appropriate predictive weights.

How the predictive opportunity scoring model is trained

First, we make sure, automatically, that the data we have is meaningful and contributing to the prediction. Then we look at each recent activity level of every opportunity and try and find a correlation between this signal and the likelihood of winning the opportunity.

After the model is trained, we can view the open opportunities in the system and give them a score based on past examples. What the AI does is learn how recent activity level affects the likelihood to win an opportunity in your organization.

What Clover Imaging Group has to say about predictive opportunity scoring

“The predictive opportunity scoring model in Dynamics 365 Sales offered sales leaders at Clover Imaging Group valuable machine learning into what has traditionally been a manual process. With the predictive opportunity scoring model implemented, sellers were able to more effectively prioritize their opportunities based on metrics that are unique to our industry and organization. Beyond that, the predictive opportunity scoring model also gave detailed explanations as to why an opportunity was scored as it was, and what needs to be done to achieve a higher success probability. Plus, the Microsoft support and development team took the time to understand the nuances of our business, which allowed the model to pick up and evaluate the metrics most important to us.”

Next steps and continued learning

We encourage you to explore predictive opportunity scoring in Dynamics 365 Sales to see how it can help your sales team to prioritize work more effectively.

To understand the full capabilities of Dynamics 365 Sales and the value they bring to Dynamics 365 customers, visit Dynamics 365 Sales. To get started with activity suggestion, visit Overview of Dynamics 365 Sales. you’re using contact capture and have feedback, questions, or new needs, contact us at

Sign up for a free trial of Dynamics 365 Marketing Sales.