Measure for measure

A graded approach to measuring, and acting upon, healthy indoor environmental quality outcomes

Buildings are beautiful, complex objects — even the ones that leak or smell or creak. As a professional working in or on buildings, you have probably received enough “advice” to know that everyone has an opinion about buildings, from your taxi driver to your granduncle. While being bombarded with “suggestions” on how you should do your job is irritating, it is worth understanding why buildings evoke these sorts of passionate, emotional reactions. We live, work, learn, heal, and meet inside them. Some of our dearest memories are made indoors, and increasingly, we are becoming an indoor species [1].

In this article, I will discuss why measuring and improving the quality of the indoor environment is essential to our experience of buildings and their impact on our health. I will discuss easy, small steps we can take from a paradigm of little or no measurement (think thermostats) to one of ubiquitous measurement everywhere, all the time. We will discuss how this can be achieved at a reasonable cost, and how the hardware and data can be managed.

Exhibit A: The Pentagon in Arlington, Virginia, USA is among the world’s largest office buildings at more than 600,000 m2. However, it is rumoured that when it was completed in 1941 it was controlled by only seven thermostats.
Exhibit B: The Edge office building in Amsterdam, completed in 2016, is packed with more than 28,000 sensors.
Exhibit B: The Edge office building in Amsterdam, completed in 2016, is packed with more than 28,000 sensors measuring and controlling all aspects of the indoor environment.

Why measure air quality

The overwhelming scientific evidence indicates that our brains and bodies are still those of a great ape that evolved for open spaces, one that has lived in mud or concrete or steel towers for only a very short period of its existence as a species. This means that, for the sake of our physical and mental wellbeing, we must replicate indoors the ‘natural’ and health-giving aspects of the ‘outdoors’. This can mean several things, including interaction with nature as well as opportunities for exercise, leisure, and socialising. Most importantly, it means developing indoor spaces that promote human health, often by replicating several aspects of the outdoor environment.

A major aspect of a healthy indoor environment is the quality of the indoor air. Simply put, the indoor air should be at the correct temperature and humidity, free from outdoor and indoor pollutants, and with a concentration of CO2 as close to outdoor air as possible. This is not controversial, and I won’t belabour the point here. There are plenty of great resources if would like to learn more about the value of clean air [2–7].

Just in a handful of recent deployments, we have found elevated temperatures and lack of ventilation in break rooms, individual offices, and, more often than we would like, areas whose ‘fresh’ air supply is actually the untreated return air of another zone, like a neighbouring IT room. Thus, we find that installing a distributed network of sensors reveals operational or commissioning issues. In most cases, these issues can be dealt with changes to operational policies, but the point is that these issues are not visible until you start measuring.

Coverage: Or how I learnt to stop trusting the thermostat and embrace density

Thermostats are usually simple and effective devices. The principle is simple: when you are above or below a specified setpoint, the system works to cool or heat a space back to the setpoint. One can propose various refinements like predictive cooling or heating, optimal start, optimal ‘ramp’, etc. but the principle remains the same: cover the ‘enthalpy gap’ between where you are and where you need to be. Traditionally thermostats have focused almost exclusively on temperature, though some now include humidity and CO2. We can play fast and loose with the term ‘thermostat’ to include other sensors commonly included in building automation/management systems (BAS/BMS), such as return air CO2 monitors.

So, where does one start then? Well, here are three aspects worth considering:

1. Measure more places, more often

2. Value trends over accuracy and precision

3. The solution to pollution is (often) dilution


In an ideal world, the thermostat represents the entire zone within which it is placed, and the CO2 reading in the return air represents the concentration at every point in the zone served by that extract. However, field measurements suggest that actual conditions such as local temperature and CO2 levels in various parts of a building can vary widely [8]. Notable trouble spots for ventilation include conference rooms or rooms made up by retrofitted partitions. While measurements cannot perfectly capture every bit of air in a space, there are significant operational gains possible when moving from one measurement in a large-enough zone to several. Recent work by standards-setting bodies [9–11] has suggested the calculation of ‘coverage’ metrics to get a handle. There is a myriad of technical considerations, but the bottom line is that we’re missing a lot of important variation and we would benefit from measuring more places, more often.


While it is very important to have fairly accurate sensors, it is also important to not get stuck in the weeds of accuracy (how correct you are) and precision (how exactly you can repeat measures over time). A system that tells you the temperature is currently 23.99856 C (75.197408 F) is no more useful than one that tells you 24 C (75 F). Or, for that matter, a CO2 reading of 976 ppm is not qualitatively different from one of 995 ppm — both are high enough to warrant some concern, investigation, and action. Even better would be a system that tells you that the current value is between 950 and 1000 ppm, acknowledging that all measurements will carry some error. At the risk of sounding hand-wavy, accuracy is essential but only up to a point.


The pandemic has exposed a dirty secret about buildings, particularly those modern ones with the Hi-Fi, fully-featured, humming, thrilling HVAC systems. That there is no substitute for clean, fresh air, and some past ‘optimisations’ have created a system where recommended lower limits on ventilation (i.e., the minimum amount of ventilation a person is entitled to) have become targets (i.e., we will try but probably never reach them). International and national norms [4,5,12] already recommend differing levels of ventilation to ensure healthy, productive environments. These have been supercharged by the need to reduce indoor aerosol transmission [3,13,14].

As with everything else in building design, design intent is only half the picture. You may intend to increase ventilation to recommended levels at the central plant, but how sure are you of the outcomes? That each room or zone of interest is actually receiving enough air to dilute any viral particles ejected by a host (conference rooms are particularly bad). Installing a sensor is the first step in analysing the effectiveness of mitigation measures.

Some readers may balk at the notion of bringing in more outside air, particularly those currently dealing with forest fires or chronic pollution. I think that Lane Burt put it best in his recent article: “use the cleanest air you can” [Accessed 21 September 2020]. That is, operational decisions have to be dynamic, which often means acting on real-time data rather than policies based on averages or “typical” inputs.

The first step is always hardest

My own background as a civil engineer did not prepare me for distributed, continuous, IoT-based sensing. In our initial pilots, I literally did not know how to handle the mass of data being generated. The most important step in the learning process, however, was actually installing a sensor and observing the data that came from it. There was nothing quite like being able to correlate a physical phenomenon, like spraying perfume in a space, to a change in a physical parameter, in this case the VOC (Volatile Organic Compounds) concentration.

An example of a display of real-time data with interactivity.
Exhibit C: Displaying real-time data can sometimes be a good idea to interact with occupants and let them participate in healthy building operation. However, it is also fine to not set up displays during the first few weeks of monitoring while you take time to understand the data.

When you see data from a building you actually know, they begin to make sense bit by bit (har har…). You don’t need to start with a high-density, high-fidelity network of perfectly-calibrated devices measuring at MHz frequencies (unless maybe you’re building a rocket ship for Mars, in that case: knock yourself out). Going for a platinum-level sensor network on the first shot is not just a question of budget but also one of information overload. In any given building the variation of environmental conditions across space and time will be unknown but it is safe to say that many assumptions on homogeneity or uniformity can be made. Any sensing specialist or vendor worth their salt should be able to recommend some level of coverage or density between a minimum and optimum.


Sitting at a window during a dreich[†] Scottish summer, the thought of recreating outdoor conditions feels mildly heretical. It is pointless to romanticise the discomfort and suffering of living outdoors in inhospitable environments. As a species, we are better off for the infrastructure we have built — homes, hospitals, schools, transport, and so on. However, that infrastructure now exerts a powerful influence on our health. This makes it all the more important to measure, visualise, and control the conditions in our buildings. Measure for measure, indoor environmental quality can often be the most impactful investment you can make in a building.

[†] Adjective — Scottish: (especially of weather) dreary; bleak. Middle English (in the sense ‘patient, long-suffering’): of Germanic origin, corresponding to Old Norse drjúgr ‘enduring, lasting’.


Parag would like to thank Dr Chris Pyke, for his constructive comments, helpful inputs, and free proofing. Also the team at arbnco for motivating the discussions, helping test the hypotheses, and generally providing the support for the work on which this article is based.


1. Velux. The indoor generation: the effects of modern indoor living on health, wellbeing and productivity. 8 (2018).

2. Allen, J. G. & Macomber, J. D. Healthy buildings: how indoor spaces drive performance and productivity. (Harvard University Press, 2020).

3. ASHRAE. ASHRAE Position Document on Infectious Aerosols. (2020).

4. CIBSE. TM40: Health and wellbeing in building services. (2020).

5. CIBSE. Environmental design: CIBSE Guide A: 2015 (corr. 2018). (Chartered Institution of Building Services Engineers, 2015).

6. ASHRAE Handbook: Fundamentals. (American Society of Heating Refrigerating and Air-Conditioning Engineers, 2019).

7. NICE. Indoor air quality at home. (2020).

8. Pantelic, J. et al. Personal CO 2 cloud: laboratory measurements of metabolic CO2 inhalation zone concentration and dispersion in a typical office desk setting. Journal of Exposure Science & Environmental Epidemiology 30, 328–337 (2020).

9. Pyke, C. R. Guide to Arc Re-Entry. 24 (2020).

10. RESET & GIGA. RESET Air Standard 2.0 for Commercial Interiors.

11. IWBI. WELL Building Standard v2 (Q2 2020). International Well Building Institute (2020).

12. Olesen, B. W. Revision of EN 15251: indoor environmental criteria. REHVA European HVAC Journal 49, 6–12 (2012).

13. REHVA. COVID-19 Guidance. REHVA: Federation of European Heating, Ventilation and Air Conditioning Associations (2020).

14. CIBSE. CIBSE COVID-19 Ventilation Guidance — v3. 16 (2020).

I work on health and wellbeing in buildings, IoT-based controls, and the use of machine learning and data science in building performance evaluation.