ReCoTech: How WeWork Uses Data to Make Buildings Better

HESA-SAFA hosted the afternoon presentations at ReCoTech in Helsinki on November 30, 2016. One of the excellent presenters was Daniel Davis, Lead Researcher at WeWork. He talked about how his company uses data to make offices better.

wework-products

“We make products. We’re a product company, our product is a platform for community!”

According to Daniel, WeWork is “designing spaces that bring together entrepreneurs, creatives, and artists; people who want to create businesses and new things in the world.” He and his research group are  especially interested in understanding what makes the spaces and people inside WeWork successful. The way they do that is quite innovative.

Architecture elements will become intelligent

Daniel started by showing images from the Venice Biennale in 2014. At the Biennale, Rem Koolhaas, a world-renowned architect, claimed that in the future, every element of architecture, whether it’s a door, window, or fireplace, will produce and exchange data. Every component in a building will gather, analyze, and use data to make itself better. Nest, the learning thermostat, was presented as an example of this future. Over time, it collects data on inhabitants’ behaviors and controls heating and cooling accordingly.

Daniel went on to describe how sensors and cloud services are used to make all sorts of things smart. “If we’re putting sensors into something as trivial as a toothbrush, then something as important and critical to our lives as a piece of architecture is going to embed itself with these sensors and gather data about what’s going on in the world,” he said. He noted that we don’t necessarily have the mechanisms in place to deal with this new world of data in architecture.

Data will be available, and we should use it to make design better

Buildings = data

Buildings = data

The state-of-the art in architecture is building information modeling (BIM). A large BIM model might be 400 megabytes in size. The largest BIM databases, containing data from several buildings, can be around 10 gigabytes in size. Compare this to Facebook, which processes 10 gigabytes every second. The amount of data that a company like Facebook can manage is orders of magnitude more than architects are currently using.

The other thing that architects struggle with is evaluating their buildings. Daniel illustrated this dilemma with a few examples. He clearly showed how architects’ visions and project results were very far apart. However, there’s very little emphasis in professional journals and architecture discourse on post-occupancy evaluation. These methods have been known for decades, but they are seldom used to gain insight for future projects.

If architects would get feedback and repeat the process several times, their results would get better with each iteration. Unfortunately, architects value novelty, not repetition.

Case study: Meeting rooms become better through data collection and analysis

WeWork builds and manages offices. They operate 100 buildings around the world. Daniel said they like to describe themselves as a kind of product company. Like Apple, they’re improving their product in every version. To WeWork, a building is an object that produces data that influences how they create buildings.

wework-chart

Prediction by designers vs. prediction by computer

Daniel shared an example of how WeWork uses data to design meeting rooms. They started off about a year ago by walking past meeting rooms in their offices and manually counting how many people were in the room and noting what they were doing. They found that some interesting patterns emerged. One company could have several meetings during a day in different rooms and on different floors. The number of people attending changed during meeting. They found that 99 percent of the time, there were one to four people in a meeting room regardless of its size.

WeWork currently offers a mobile app that collects data on their customers’ meeting room experiences. Users rate rooms on a scale of five stars. If the room gets three stars or below, the app asks the customer to provide a comment.

Every day, they receive feedback on how people feel about their meeting rooms. They also receive emails from customers. Using machine learning, WeWork can identify and fix problems quickly and improve their product standards constantly. They also track how people travel between buildings and use their spaces.

When collected over time, data is also valuable in deciding how many and what type of meeting rooms should be built. Daniel showed how WeWork uses neural networks to predict meeting room usage in each building. They have compared the information provided by the neural network with designers’ predictions. The computer is much more accurate in matching projections with the actuals measurements.

“When we’re going to design a room, we can be pretty sure that the thing we are designing is actually needed and going to be used by the people who are going to inhabit the building,” Daniel concluded.

Learn more about WeWork at wework.com.

Title image: ReCoTech panelists, from left: Hilla Rudanko, Daniel Davis, Idil Gaziulusoy, Annie Locke Scherer

Photos by Aarni Heiskanen