A Look into Microsoft’s Data-Driven Approach to Improving Sales

Companies are beginning to utilize their employees’ behavioral data — generally known as people analytics — to better understand and improve their sales operations, with strong results. Microsoft, where we work, is no exception, and B2B sales is one of the areas where we are seeing the most value. Our findings, and the ways we came to them, can be useful to other sales organizations looking to make internal changes of this type or optimize how their salespeople relate to customers.

In mid-2017, we executed a major redesign of our sales organization in response to what our customers needed from us, and to better align our selling approach with cloud services sales model (in this model, customers pay based on usage versus a traditional fixed licensing deal). We knew we needed a fast and effective transition to the new model without dropping the ball with our customers, but the undertaking was daunting and the stakes were high: With a complex sales organization of 20,000-plus salespeople covering large enterprises to small business customer segments, and spanning 100 countries, it was important to see how these changes impacted our customer collaboration and partnerships. We needed to get answers to some of our biggest questions, including:

Are we spending enough time with our most important customers?
Are new hires ramping up and collaborating with customers as quickly as expected?
Are they growing their internal and customer networks?
Are salespeople collaborating with one another effectively?
How is all this impacting our customers’ own business success?

Our hunt for answers started by using our own Workplace Analytics product to aggregate de-identified calendar and email metadata for thousands of enterprise salespeople. We then combined that with organizational and customer relationship management data to determine how the people selling via the cloud sales model were collaborating with their internal teams, customers, and partners. The next step was to correlate sales outcomes with these behaviors to identify the patterns that correlated with better results. These analyses were done in part to help us through a massive transformation and in part to better align us in responding to our customers’ needs and expectations. To date, the analyses revealed several actionable insights, which we came to with the help of our colleagues Ben Boatman, Chris Moss, Gabriel Zhou, Jared Baker, and Fabio Correa.

1. Networks are vital — and a reorg could destabilize them. One of the first things we learned is that salespeople with larger, more inclusive networks tended to have better outcomes. This is consistent with a number of other similar studies. Based on this finding, we initiated a program to coach our sales teams to focus on efficiently building and growing their internal and external networks. By looking at network size relative to tenure within the company, we were further able to establish that it typically takes roughly 12 months for most people to build these networks.

This underpins the importance of stability in roles over that time period, and beyond. It also left us concerned that the reorganization was forcing the salesforce to rebuild their networks from scratch, which could be costly and sub-optimal for our customers. To mitigate this cost, we rolled out programs to emphasize manager coaching and invested in facilitating rapid network growth for new hires.

2. We engage very differently with high-growth accounts. Another key aspect of the re-org was to ensure continued growth of our business and the right level of engagement with customers. Looking at the amount of time teams spent interacting with each of their accounts, as well as the number of individual contacts they were connecting with, allowed us to identify statistically significant differences in how teams engaged with the different account segments. On average, teams engaged with twice the number of customer contacts in our higher growth accounts, and collaborated double the amount of time with these customers as compared to lower growth accounts.

To make sure this wasn’t just a one-time anomaly, we also confirmed that this pattern was consistent month over month. Correlation vs causation is always an open question with an initial finding like this: are the accounts higher growth because we spend more time with them? Or do we spend more time with them because they are higher growth? Deeper analysis showed that investing more time and energy into partnering with some of these lower growth accounts could improve them. As a result, we adjusted our sales coverage models to enable more face time with these previously underserved customers.

3. Relationship investments correlate with customer satisfaction. It was important that the new sales model also drives happier customers and partners. Therefore, our next step was to look for patterns associated with customer satisfaction. We found that customer satisfaction is directly correlated with customer collaboration time (email and meetings) across all Microsoft roles and teams engaging with customers, including product engineering and marketing teams.

In the enterprise segment specifically, satisfied customers are the ones we spend the most time with and the least satisfied are the ones we barely keep in touch with. This and other findings encouraged our sales leaders to revamp internal business processes such as business reviews and forecasting meetings to be more efficient. We also reduced the number of enterprise accounts per seller to allow for more customer interaction. This enabled our sales teams to spend more time building and maintaining relationships across their entire account portfolios. We also observed behavioral differences in different countries — some use email more frequently than others, for example U.S. and Canada sellers directly schedule meetings with customers through Outlook, while in Japan customer meetings are more formal and scheduled via executive assistants. This confirmed our understanding of various cultural norms and collaboration patterns which was an important input to our analysis.

4. Customer satisfaction (and churn) can be predicted. As part of our ongoing organizational efforts to better understand our customers, one of our teams built a machine learning model that uses more than 100 features to predict customer satisfaction. We worked closely with this team to add the behavioral data about collaboration we gathered into the model. After our analysis, we discovered that collaboration became the top feature in predicting customer satisfaction, and helped increase the accuracy of the model from 78% to 93%.

Being able to predict satisfaction of each of our customers at any given time with this level of accuracy was a groundbreaking discovery for us. Further, having a deeper understanding of how our team’s interactions influence customer satisfaction by segment has huge upsides: it enables us to intervene in time to change high-risk customers into low-risk ones, and to offer new opportunities to highly satisfied customers. Our ability to predict customer satisfaction with this level of accuracy will help us keep an ongoing pulse on our transformation and intervene in a timely manner to ensure customer satisfaction at all times.

What’s next. Our goal is to arm each seller with these four insights on an ongoing basis, setting them up to be as successful as possible in creating value for our customers. We are currently testing a prototype in which a customized and automated email is sent monthly to each seller to help guide them toward behaviors that drive higher outcomes. Importantly, the data sent to each seller is set up for their eyes only; to protect everyone’s privacy and retain trust in the system, no one else, not even upper management, can see anyone else’s data.

Salespeople are provided the following every month:

Predicted satisfaction scores for their customers
Reminders to connect with customers they’ve lost touch with
Internal and external network size in comparison with benchmarks in their local areas
Recommendations on how to grow their customer networks through LinkedIn Sales Navigator
Time spent with each of their customers as compared to addressable market
Top internal collaborators and reminders to connect with other sales roles that are also working with their customers

We believe this information will empower our sellers with nudges and recommendations that are simple, actionable, and effective. Early reactions are extremely positive. If we continue do our jobs well, our sellers will be empowered to be as successful as possible, and will get better and increasingly connected to customers over time.

We also learned a few things along the way that were critical in helping us shape the story and vision to drive business impact.

Executive sponsorship is critical, and we couldn’t have gotten our analysis off the ground without it. Their support helped us get the right level of visibility for continuous analysis and digging deeper, which ultimately got us to something more meaningful and actionable.
Investment in business analyst, data science, and data engineering talent was essential. It takes a real commitment to unlock and operationalize the most powerful insights, and it takes a lot of people to do it. We believe bringing the right people onboard is worth it.
Freeing data from silos and cross-team collaboration was key to our success. As for any analytics project, we needed to source data from multiple sources across the company to correlate behaviors with sales outcomes. Without this, our efforts would be fruitless.

We’ve invested a lot of time and resources in building out our behavioral data capabilities, and they’re already generating tremendous value. However, we believe we are still in the very early phases of uncovering what’s possible. We have a long way to go, but so far, our transformation is working. Pushing the envelope on behavioral analytics has been a key ingredient to our success, and hopefully our insights can help your salespeople, too.