When Industrial IoT Comes to the Office
On the complexity of IoT in the seemingly straightforward scenario of determining whether a conference room is occupied.
One cool aspect of being part of Favorite Medium is that we get to work with emerging technology, such as industrial IoT. We’ve had the opportunity to work on IoT in many spheres from security to energy grids to physical space analytics. Our IoT projects have spanned the gamut. Sometimes we’re just trying to make an ordinary task a touch easier, such as ordering lunch with a button. Other times, we’re changing the way that people manage a chronic health condition like diabetes via a medical device.
With the Internet of Things, it feels like we’re entering a new era. Chips are getting smaller, more powerful, and less costly to produce. Likewise, the software networks, data storage, and computational power needed to enable devices to communicate are also becoming cheaper to build and maintain. It’s clear that the potential to transform the operations of many industries — and the way that we live — is vast.
Today, we’re sharing about a project that takes place in a location that’s familiar to many: the office space.
The Conference Room Conundrum
The question posed to us was: How can we make the best use of space in an office?
Our client was interested in finding out if conference rooms were being used well. With real estate at a premium, using space efficiently is essential.
Yet figuring out the most productive office configuration has proved to be a vexing problem for facilities managers. They didn’t have meaningful data to help them make informed decisions. Were the number of rooms and the amount of people they could hold the right number for the space? What if there was a way to look and see which conference rooms were available at any given time? On its face, this seemed like a straightforward issue. As we dug into these questions, we found that they became increasingly complex.
To determine whether or not a room was occupied, we outfitted them with PIR (passive infrared) sensors that could detect motion. We built an app that integrated the client’s existing calendaring system. One could search for a room in real time, generating a list of what was available at that moment. Not only that, but if you had booked a meeting and weren’t in the room at the appointed time, the system would ask if you were still planning on using the space. If you didn’t respond within 10 minutes, it would cancel the reservation, freeing the room up for someone else to use.
The sensors in each room connected to a gateway via Bluetooth. However, we found that wireless Bluetooth sensors can be unreliable in buildings with a large amount of structural steel or with very thick walls. These can disrupt a sensor’s Bluetooth connection to a gateway. The pilot space we were testing happened to be located in an older building that had been retrofitted with large steel beams. Thus, the sensors constantly lost connection to the gateways. We conducted a lot of complicated field testing just on the reliability of the sensors.
We also found that motion sensors by themselves were not sufficient. Some people like to fidget, but others sit very still. And in the case of the latter, sensors could miss them. There were times when a room became unlocked even though people were in it because there were insufficient sensors placed throughout the room to detect very still people or low occupancy (for example, two people taking up a 10-person conference room). To truly determine if humans were in the room, it would be ideal to add sensors that tracked temperature too.
Another challenge was human. People had concerns about their privacy and were nervous about their calendars interacting with the reservation system. Some people didn’t want to participate in the system at all. At other times, people removed sensors. (A good inventory management system to track devices was important for this project!) Ultimately, any system needs buy-in on the individual and organizational levels in order for it to succeed.
Data Engineering in Office Space Analytics
The conference room efficiency idea was then expanded to the larger space. What if we could generate analytics for the entire office space and see how well (or not) it was being utilized? To do this, we had to carefully design where we placed sensors. Then we collected and stored data for several months. Sensors help capture data in a way that’s far more accurate (and cost effective) than humans with clipboards observing the space.
One key thing we did was separate the raw sensor data from the aggregated space data. So much data was being generated, but not all of it was necessarily meaningful. Cleaning up the data was crucial, as it allowed us to converge on an aggregated data model that supported a wide variety of queries. (Garbage data = garbage insights.) Thus, we could develop a wide variety of visualizations very quickly.
Because the data was “flexible” we could manipulate it in many ways to deliver insights. Our designers and facilities consultants used the data to make recommendations on reconfiguring spaces to optimize how employees actually use the space. For example: the data might show that phone booths were often occupied, particularly at certain times of the day, but that hot desks were not. So, from a facilities standpoint, removing some desks to create more room for phone banks would increase space efficiency.
IoT doesn’t mean that you can just slap sensors on something to make it smart and expect it to start generating data. Even with something as seemingly straightforward as room reservations, we found a level of complexity that isn’t in many traditional web- or app-based software projects. Not only do you have data-related software engineering, you have networking and physics. (When steel and Bluetooth collide!) You also need a good understanding of human behavior related to the objects in their sphere that are collecting data about them. And then you need to understand how to sort and clean that data for it to generate meaning.
You must deal with all of this just to let people know that a conference room is open or to tell them how they are using the spaces in their office.
No matter what the situation or industry, when we approach this kind of platform, we’ve learned that it’s important to understand the broad challenges, not just the software challenges. We may work in the digital realm — but whether it’s analyzing something as ordinary as office space or something as important as medical data that could potentially save lives — at the end of the day, our work is centered on making people’s lives a little better and a little easier.