Big data at the intersection of building analytics and people analytics

buildings with peopleImagine if you could simulate your building or workplace environment before you built it – not just simulating energy usage or daylighting – but creating a simulation of how the people would behave and work inside your space. And not just a generic sample population, but your actual workforce in a simulation that knows and understands their actual behaviours. Before investing in bricks and mortar (or tables and chairs) – you could test numerous design scenarios and their impact upon not only how the building itself operates, but also how the occupants respond, their use of space, their interactions with one another and more. How would this change the way we design, the way we build and possibly the way we all work?

Many people would think this sounds pretty far fetched, futuristic and certainly a little bit big brother. The reality is that we actually have both the information and the technology available to do this – right now in 2014. Today I’m going to talk about why we would want to look at simulating human behaviour in the built environment and what this could mean for design, as well as discuss the types of data analysis and technologies from different fields that I believe could be brought together to make this kind of simulation of the built environment possible.

My background is in workplace and educational design. A large workplace is probably one of the most complex environments in which to try to predict and understand human behaviour. Unlike a restaurant, a shopping centre or a train station, it is designed to have a large number of diverse activities taking place. Whilst at the same time – and I know this sounds a little strange – there is actually less of a clearly defined purpose in being in a workplace than in many other kinds of enivonrment. Different individuals have different purposes in being there, because they enjoy their work, to socialise or to earn money are just a few. An opposite example of a much more simplified purpose of space would be in a cinema – where almost everyone is in the space for one purpose, which is to see a movie (although they may have different motivations for seeing the movie). In the workplace, because there are so many different activities and behaviours, finding patterns to predict how people work – and even understanding what improves their work is more complex.

The holy grail of workplace design is to be able to prove that certain design elements increase productivity. Most researchers agree that it has historically been almost impossible to measure productivity in knowledge or service oriented workplaces, which today make up the bulk of first world workplaces. We can however measure a lot of approximations of productivity – or things that we expect to have a close correlation with productivity – things such as staff retention, absenteeism or self reported satisfaction and comfort levels. This kind of data is readily available.

Another key issue in workplace design centres around the actual useage of space. Real estate is a significant business cost (though much less significant than the people cost) During the design phase of any project there is great debate over wdifferent kinds of spaces and how and if they will get used. Do we allocate individual offices to sit empty, will staff actually use that large breakout space, will that training room sit empty for half the year? From the workplace designer through to the facilities manager and the CEO, the ability to simulate occupant behaviour in the workplace has a huge potential to impact upon what and how we design our workplaces. To me this could be the next significant game changer in workplace design and productivity.

It’s being made possible by big data. In the past, we have not had access to enough information about either building systems or occupant behaviours to be able to simulate these kinds of complex environments. There is software that can simulate human behaviour – and it has been around for more than 20 years. Commonly used software that simulates vehicular and pedestrian behaviour or fire engineering modelling is all simulation software which is based upon predicting human behaviour. However, the difference between these previous software models, and predicting behaviour of occupants of a workplace or other complex building type is the complexity of the human interactions. Human behaviour in a fire situation or within a train station environment is much simpler than in a workplace. There are less possibilities because of the limited scenario, and also we are essentially only tracking one variable – movement. Workplace design has made very limited to use of this kind of simulation, for example Google campus at Mountainview has been designed to ensure that all staff are within 2 and a half minutes walk of each other. Movement within the workplace, or other building types, is a pretty simple and limiting factor to use to test and simulate our designs. Big data, and in particular, combining information from the fields known as Building Analytics and People Analytics, could give us the opportunity to feed a huge range of different kinds of building and human behaviour data into a simulated building model.

Building Analytics is currently seen as the next big thing in building and asset management as well as an important factor for environmentally sustainable buildings. Probably most people in this room have at least some familiarity with this field. In the past, data gathered from buidling tuning or the BMS was more limited and unlikely to be in real time. However this has been changing. Building managers can now have real time access to a range of data – from factors such as which lights or appliances are in use, to the temperature, CO2 and VOC levels, heat or movement maps of actual occupation coming through motion or heat sensors, lifts that track occupant destinations or individuals movement through security systems via access cards or CCTV. Many of these systems are already commonly available in any new large commercial development. Facilities managers and building owners are using them to understand and predict occupant behaviours in relation to building systems. Historical data from the same systems can then be combined with real time data to predict or simulate things like building energy usage in a given period or what the impact of certain weather conditions might be on occupant comfort.

This type of building analytics does take into occupant behaviours, but only at a very simple level, things like is the space occupied or not occupied – because this is key information for the running of building services such as lighting and air conditioning. Whilst this data is firstly being used to control the systems and secondly to predict building performance it also provides us with real time reliable data on occupied versus unoccupied space. The ability to use web based booking systems for rooms or desks was the first step that created some kind of data around anticipated space usage, but it wasn’t real usage data, only a prediction of usage. Today BMS data can be combined with this kind of booking system, and it is possible to not only track advance bookings but real time actual usage – if someone doesn’t turn up to use the booked space it can be reallocated to somebody else. Whilst this kind of information can help manage a building it doesn’t predict behaviours or improve occupant performance.

This is where People Analytics can start to work with building analytics to create a fuller picture of how space is actually being used, and what this means for the occupants.

So what is people analytics? People analytics looks at data generated by people and companies rather than data generated by building systems. It is a growing field of social science with implications in particular for human resources and recruitment – and in my view for designers. People anayltics starts to look at and analyse any available kind of data in order to find patterns and understand human behaviours – its anthropology using computer generated information. Today, data generated by people can include anything from emails, to social media usage, to bluetooth and movement tracking, voice recordings, computer data logs, organisational plans, charts and documents or google searches. If you think about your electronic footprint, even without anyone planning on tracking what you do, there is a lot out there. The more we use the web or cloud based services, the more data exists about our habits, our performance or our personalities. In the past the quantities of data have been so much smaller, that there was not sufficient amounts of data to generate patterns or the computing power to crunch it. Today there is.

By analysing huge amounts of historical data it is possible to identify patterns or characteristics of certain groups of people, or how to predict or promote certain behaviours. Once the historic data set has been created, it is then possible to analyse data of new people to identify who fits the patterns. We still don’t always know what the data can mean on its own though. One of my personal favourite odd ball data correlations is that super guru computer programmers apparently have a tendency to like a certain Japanese manga website! You can see the applications to recruitment and HR immediately.

Another fun example of the use of large samples of aggregate data is the Twitter happiness index. This website analyses the use of certain words on Twitter every day since 2008. Words are assigned values from 1 to 9 to signify sad to happy. The overall happiness score for each day is then calculated and graphed. There are also Twitter election indexes, oscar indexes and many more aimed at trying to predict outcomes based upon twitter traffic. Elections polling has been a high profile example demonstrating people analytics to the public. In the 2012 US presidential election, big data was used by a number of forecasters to accurately predict the results in all 51 states. These are all examples of different uses of people analytics.

But how does all of this relate to buildings, and workplaces in particular?

Lets start with a really simply example of using other kinds of data in combination with building systems data which was undertaken by Immersive, a big data company based in Melbourne, Australia. By taking the historic heat sensor data from a workplace BMS and analysing it against the organisation’s project planning data for the same time period, it was possible to determine what the actual space usage and occupancy loads had been over the period compared to the predicted project staffing levels. Using the same forward project planning data, it was then possible to predict the organisations actual future space needs. Whilst this takes into account some level of occupant behaviour – space occupancy – again its a single variable, where we are still looking at physical space more than actual occupant behaviours.

But what if we could take multiple kinds of data – data that is more specifically tracking behaviours in the specific context of the workplace? And not just data about electronic interactions – what if we can gather the same types and quantities of data about our face to face interactions as our electronic interactions? In analyising workplace productivity, this tracking of real time physical interactions is important – because in most companies, much of the informal collaboration still happens face to face. The theory is that in organisations where knowledge work is undertaken, social networks define how information is transferred informally across the organisation, and that this informal sharing is creating a transfer of knowledge. This new knowledge then has a significant influence on how the work gets done and therefore on productivity – kind of like how you learn just as much by going for drinks at RTC as you do in the presentations – people are sharing what they already know.

If organisations can find ways to firstly understand these social collaborative networks and then secondly promote them, social scientists believe that the organisations productivity can be enhanced. The office space itself then becomes one means of modifying social behaviours in order to promote certain kinds of interactions. But how to collect information on face to face interactions – we are not all going to suddenly start skyping the person sitting next to us.

Enter the sociometric badge. Developed by a team at MIT, this device contains a number of sensors. An IR transreciver allows the devices to sense one another, bluetooth records their physical location in space, an accelerometer can figure out if you are sitting or standing and a microphone detects audio. Right now this device is approximately the size of your building access card although slightly thicker and can be worn around your neck. In the future your smart phone will probably be able to track all of this anyway – its actually already got almost all of the same sensors. The sociometric badges have been used to track and record the behaviour of building occupants in a number of studies investigating the way we work. The outcomes have been published in a great book called “People Analytics: How Social Sensing Technology Will Transform Business and What It Tells Us about the Future of Work” by Ben Waber.

One of the interesting things is that the microphone doesn’t even record what you actually say. It records things like tone, change in volume and speaking speed – which are considered social signals, and which are in fact more important in our face to face interactions than the words we actually speak. Early tests in laboratory environments included speed dating and salary negotiation simulations. Computers were able to predict outcomes with over 85% accuracy based upon 5 minutes or less of these recorded social signals.

These devices have since been utilised in a variety of actual real workplace studies. So far sociometric badging has found that call centre productivity is enhanced when team members take breaks together and that the amount of time spent interacting and the amount of physical movement are god predictors of creative days.

These studies, and most in the book, are based around understanding and modifying behaviours rather than modifying environments, but as any architect or designer knows, if you modify the environment, you have the opportunity to modify the behaviours. One of the studies of most interest to us, looking at how changing physical space impacts on occupant behaviours, was a study which investigated the size of lunch tables in a workplaces cafe spaces. Using data from the sociometric badges within an online travel company, it was found that staff that sat at larger lunch tables were more productive. Within the existing office environment there were 2 different spaces staff could choose to eat lunch – one had small tables for 4 people, and the other larger tables seating up to 12 (or they could chose to eat by themselves at their desk). The data quickly showed that the people who ate lunch together would then tend to communicate further that day. The staff that sat at the larger tables were more likely to speak with others outside of the group they had arranged to lunch with, and formed larger conversational groups at the lunch tables. These wider lunch time conversations led to links and collaborations in the organisation that were not otherwise being formed. These links were part of the knowledge sharing that led to greater productivity.

In another MIT project, the cubicles themselves were equipped with sensors so both the office environment and the people within it were being analysed. The cubicles were fitted with blinds instead of typical workstation cubicle screens, in order to provide privacy or allow collaboration. Based upon the collaborations that the data had identified as being most beneficial, the automated blinds would open or close overnight. In this way the building itself can even automatically respond to data analysis.

Often, the data coming out of these studies is not surprising the social scientists or the building designers. What is is doing though, is proving things we know instinctively, the things we have seen work before.

When you think about this information about your day, what you do, where you go and who you talk to is then combined with your electronic footprint, the information about your colleagues and then possibly also the building data – its a pretty full story of what happens inside a given workplace or building in a day. The possibilities for analysis and experimentation will be endless. Why is this so important to design and construction though? So far this is all about modifying existing environments. Being able to test and prove what works is the next step.

In an example that initially does not seem to be related to physical space, but to health, the sociometric badge data is combined with data about how disease spreads. The impact of sickness on the work environment, the interactions and the productivity can then be simulated across a range of scenarios with different people being the disease originator and different simulated responses such as stay home versus solider on being tested. One suggested solution to minimise spread of disease was to change the regular seating layout, which has the effect of reducing the level of interactions between people who already knew each other.

Moving into the not so distant future – there is no reason why the possibilities of physical environments could not be tested inside a BIM, with the algorithms behind the behaviours of the sims being developed from these kinds of behavioral data sets. We have the technology available to us already.

While this isn’t about BIM as we know it today, the link between the the building model and the simulations is obvious. But will architectural practices embrace these technologies or will this lead to another new kind of consultant in our team?

Imagine the value of the design and simulation team who can prove to the client organisation that workplace productivity could be enhanced simply by working with them? Translate that to all kinds of building typologies – and the whole definition if what architecture is or could be may change. Perhaps big data is going to have an even more significant impact on change in our industry than BIM, in ways we haven’t even imagined yet.

Ceilidh Higgins

This blog is the text from my presentation at RTC North America last month, as part of the session BIMx: Big ideas around big data.  I had a great time over there and attended some excellent classes.  If you are in Europe, RTC will be in Dublin later in the year.

Image Credits: Via Flickr Creative Commons
Big Buildings https://www.flickr.com/photos/neilarmstrong2/5480543083/
The New York Times on the New Art of Flikr https://www.flickr.com/photos/thomashawk/2442371176/