What to do when the lights are on but no one is there

May 9, 2017   |  

Ever wake up in an office with bright lights shining, high-tension music building to a crescendo, and you’re eerily all alone?

If you answered “yes,” don’t worry, no one is going to get hurt today. This story has a hero, a data scientist who sweeps into the room, crunches some numbers, sprinkles a pinch of machine learning on the room, and suddenly everything is back to normal.

Humorous, maybe, but not all that far from the truth when it comes to explaining what the data scientists on my team at Microsoft do day in and day out.

Today I want to share with you how my Microsoft IT Data and Decision Sciences team – I think of them as data artists – are helping the people who manage buildings at Microsoft optimize the way they manage the company’s 765 buildings.

Like other companies, Microsoft spends millions of dollars heating, cooling, lighting, and supporting its buildings. In many situations, there are few or no people in certain low-traffic parts of buildings. Hence the need for a data scientist or two to pinpoint where all the people are (and are not).

Realizing that they could save millions of dollars by fine-tuning the way they manage the company’s buildings, the Microsoft Real Estate and Facilities (RE&F) group asked my team for help. They had data on when people enter buildings because Microsoft employees must badge-in every time they enter a building. They don’t badge-out, however, so there’s no way to tell how many people are in the building at any one time, right?


When RE&F came to us for help, we went over all the sources of people-in-building data that they had, which admittedly wasn’t much (Microsoft is very serious about protecting employee privacy, and rightfully so).

But they had more than they realized.

As mentioned, RE&F had solid info on when people entered buildings. Thanks to the pervasive use of Wi-Fi, they also could tell when devices were carried into a building (in pockets and bags). They didn’t know who the phones and laptops belonged to, but there were patterns that could be sniffed out.

Using modeling, machine learning, and regular old sleuthing, my team figured out how to match a phone and PC to a badge. We had to build our own data lake to store the data, and after about 90 days of tracking, we could reliably pair devices with badges. Once we did this, we could triangulate how long a given person stayed in a building. And by using Wi-Fi data linked to devices, we could track where people went in their building.

Note that when an employee used a particular space, all we sought to learn was the organization that he or she rolled up to – we didn’t seek out any identifying information about the individual. All private data was protected in this effort – we didn’t track any identifying information for privacy reasons.

Needless to say, this kind of melded-together information is powerful!

With it, RE&F can track how people use various spaces in a building (and where they don’t), and very importantly, when (as in what time of day). They can use this information to perform long-term space planning, regulate temperature, and better stock break rooms as a means of improving the overall employee experience.

When fully deployed across the company, estimates are that this technology will allow the company to cut its building management costs by as much as 20 percent which could translate into savings of $270 million per year. As a bonus, the information also gives RE&F much better insight into how employees are using offices, meeting rooms, and common spaces, allowing us to alleviate pressure where there is too much use, and to rethink how to use spaces that are going unused.

I invite you to check back over the coming months as I use this space to share more stories on how we are using data analytics to change the way our customers work, customers inside Microsoft and externally. Next time I’ll go deep into another example of data analytics work we did for another one of our customers. Meanwhile, read my first post on how I started in this role and why Microsoft IT is home to my team and my second post on what it takes to put together a data analytics team.

Tags: ,