Optimizing office space: the role of passive occupancy

Efficient office space use is critical for organisations as real estate is expensive (Knapp et al., 2009; Rusek et al., 2018; Zhou et al., 2022). The office area per worker has declined over the past years due to the increase in remote working and a practical understanding that not everyone needs a dedicated desk (Kim et al., 2016; Laing, 1990). The optimal number of desks and other spaces should be derived from the business plan and support the goals of the organisation (Roper and Payant, 2014). To understand current space usage and optimise the allocation for each team, department or organisation, technology like presence sensors can be used to track desk and meeting room occupancy (Tagliaro et al., 2021). Understanding the current use of space is ever more important as the COVID-19 pandemic accelerated remote working, making office occupancy more dynamic (Hensher et al., 2023; de Lucas Ancillo et al., 2023).

One problem in the measurement of desk utilisation with the current sensor technology is that we can detect if someone is currently sitting at the desk (i.e. the desk is actively occupied) but we cannot differentiate if the desk is unoccupied but unavailable to others due to personal belongings left behind, a state that can be termed as passive occupancy. This concept is missing from academic literature, but there is at least one example from the industry. For instance, the technology company Vergesense has recognised the importance of this state to determine how many desks are needed in the buildings, and they report utilisation data to their customers, including both active and passive occupancy (Vergesense, 2022). To clarify, passive occupancy refers to a desk that is both unoccupied and unavailable to others. When collecting utilisation data to determine how many desks are needed in the buildings, these physically unoccupied but unavailable desks are, for this purpose, considered occupied. Thus, they are called passively occupied. This term should not be confused with methods where occupancy data is collected in a passive way, such as measuring the number of people in an indirect way, such as how many are connected to the Wi-Fi system, as presented in Nyarko and Wright-Brown (2013) and Wang et al. (2020), or by measuring building energy consumption as introduced in Martani et al. (2012).

Passive occupancy can be an issue for organisations aiming to optimise office space. For example, if a person reserves a desk in the morning, they might run from one meeting to another the entire day and still reserve the desk with a water bottle, a jacket or bag on the chair, making it unavailable for everyone else. To prevent this happening and to ensure that desks circulate in a desk-sharing environment, companies have introduced clean desk policies (CDPs) (Babapour Chafi and Rolfö, 2019). CDP is an office policy that defines how long a person can keep the table to oneself without actively occupying it. For instance, if a company has a CDP time delay (TD) of 1 h, the employee should clean the desk if they plan to leave for a meeting longer than 1 h. The TD defines how long the employee can be away from the desk and still keep it for themselves. The key point is that even though the desk is physically unoccupied, it is not available for others to sit at, and they must find an alternative desk. This means that the total amount of desks must be higher than if CDPs did not exist.

Despite the various methods to define and measure utilisation, current approaches, such as those based on presence sensor data, often fail to provide a realistic picture of desk availability because they frequently overlook passive occupancy. Moreover, utilisation studies primarily focus on building technologies such as heating, cooling, ventilation or lighting, rather than on space allocation (Azimi and O’Brien, 2022). There is, however, potential in using this sensor data in space management and to estimate the space need and to determine which spaces are the most popular among the workers (Einola et al., 2023). In addition, the existing literature on CDP is limited to describing its impact on well-functioning desk-sharing, but its impact on office utilisation calculations has not been investigated to the authors’ knowledge.

To address the shortcomings in the previous literature on CDP, this study aims to highlight the impact of different CDP TDs on calculated utilisation rates. It seeks to answer the question:

Understanding passive occupancy and CDP as a tool to encourage efficient space use can help organisations to determine the optimal size of their office space. The rest of the paper is organised as follows:

• Chapter 1 reviews the existing literature on passive occupancy, its connection to space utilisation, methods for measuring office utilisation and the CDP;

• Chapter 2 outlines the research methods;

• Chapter 3 presents the results;

• Chapter 4 discusses the findings and limitations; and finally

• Chapter 5 offers the conclusion and future considerations.

The office space need estimation of an organisation should be an output of the business plan (Roper and Payant, 2014). This means that if the organisation is, for example, planning to grow or make changes in the organisation structure, the space need may either grow, decrease or require some rearrangement of the original space. The new working practices that support desk-sharing, such as flexible work and a desire for different types of workspaces in activity-based offices, have changed the office space needs (Babapour Chafi and Rolfö, 2019; Jayantha and Oladinrin, 2019). This change mainly converts to reduction in number of assigned desks, as modern laptop-based work can be executed from various spaces (Hills and Levy, 2014). Moreover, after the COVID-19 pandemic, organisations have realised that the occupancy levels have not returned to what they were (Hensher et al., 2023; de Lucas Ancillo et al., 2023). Therefore, many organisations have excessive space that costs them resources.

Accurate utilisation rate data helps organisations balance cost savings with productivity when estimating space needs. Incorrectly sizing the office can lead to reduced productivity and well-being if workstations are insufficient (Kim et al., 2016; Mantesi et al., 2022) or high real-estate costs if there are too many desks. Conservative organisations may retain excessive space, incurring unnecessary costs for rent, cleaning and energy. Without accurate data, safety margins added to peak occupancy estimates often result in excess space. Conversely, relying solely on active occupancy data risks underestimating space needs, leading to too few desks and employee dissatisfaction due to inadequate workspace availability.

A conservative approach also fails to address energy-related carbon emissions from heating, cooling, ventilation and lighting of surplus space. There seems to be a strong relationship between the amount of space reserved by an organisation for employees to work and energy consumption (Francart et al., 2018, 2020). For instance, providing 44 desks for 66 employees reduced energy consumption by 30% (Dooley, 2017), while closing underused university lecture halls based on occupancy data saved 19% in energy (Azizi et al., 2020). Finding the optimal balance between reducing the space as much as possible while considering productivity and well-being is key to reducing carbon emissions related to the energy consumption of office buildings. This is why trustworthy and accurate data is needed.

Traditionally, office space utilisation has been estimated through space audits or surveys (Rusek et al., 2018). However, advancements in sensor technology now enable real-time occupancy measurement, and facility managers increasingly show interest in sensor-gathered utilisation data (Einola et al., 2023; Zhou et al., 2022). Despite this progress, estimating office building utilisation rates has remained a long-standing challenge (Knapp et al., 2009; Stjelja et al., 2022). Several attempts have been made to estimate utilisation via imperfect methods and data sources. For example, CO₂ measurement, access control data, Wi-Fi devices, water consumption, restaurant purchases (Alsafery et al., 2023) or other technologies such as passive infrared (PIR) motion sensors, people counting cameras or wearable sensors (Sun et al., 2019; Zhou et al., 2022). Most of these technologies do not measure accurately the number of people in the space, but rather if the space is occupied. In addition, previously the utilisation measurements have not been made on a desk level but rather floor or room level. Therefore, applying CDP TD on top of utilisation data has not been feasible, which may explain why passive occupancy has been overlooked in previous research.

From presence detection technologies, PIR is the most widely applied (Zhou et al., 2022). PIR sensors are typically installed on ceilings or walls to cover entire rooms, reducing the number of sensors needed but making it impossible to determine the number of occupants. Hence, this approach is not suitable for open offices if number of people is of interest. For such environments, desk-mounted sensors, usually PIR-based, are a better option as they detect occupancy at individual workstations. Sheikh Khan et al. (2021) tested desk-mounted sensors, reporting 87.5% accuracy compared to manual observations. These sensors are commonly used in smart workplace solutions (SWS), which are increasingly adopted in modern offices (Remes et al., 2022). SWS can also provide employees with real-time desk availability, supporting practices like hot-desking (Einola et al., 2023).

Another challenge in accurately defining utilisation levels stems from inconsistencies in the levels of detail and units of measurement across different methods (Tagliaro et al., 2021). Office utilisation calculation methods vary depending on the type and placement of sensors. The simplest approach divides daily usage hours by total available hours (Forsström et al., 2011). Alternatively, the utlisation, frequency, and occupancy method calculates utilisation by multiplying frequency (time the room is used divided by total availability) by occupancy (number of users divided by activities) (Rusek et al., 2018). However, this method may be unsuitable for sensors that do not detect the number of users or activity types. Imperfect data and inconsistencies in reporting pose challenges for monitoring occupancy and space usage (Zhou et al., 2022). These methods also usually fail to take passive occupancy into account, which may result in utilisation rate that does not reflect the real availability of desks.

The importance of CDP has been acknowledged for well-functioning desk-sharing (Oygür et al., 2022; Wohlers and Hertel, 2017). Most often, CDP refers to the requirement of cleaning desk after the day (Göçer et al., 2018; Kingma, 2019; Oygür et al., 2022). However, this study discusses CDP as a tool to circulate desks during the day. Therefore, this study uses shorter CDPs with relation to utilisation rate calculations.

CDP was identified as one of the five key rules of activity-based offices by Babapour Chafi and Rolfö (2019). In their study, the CDP TD, defined as the time a desk can remain unoccupied without being reassigned, was inconsistent across the four offices examined. In two cases, employees were required to remove all their belongings at the end of the day, while in one case, the CDP was set to two hours. The last office had a CDP requiring employees to clean their desks when unattended, but it did not specify a timeframe for doing so. They found out that the workers complied with the CDP rules and argued that it is especially important if the company has a high employee-to-desk ratio.

CDP can impact employee satisfaction both positively and negatively. Positive effects include the perception of a clean environment due to uncluttered desks (Søiland, 2021) and improved organisation and documentation resulting from reduced printing (Van Marrewijk and Van den Ende, 2018). Negative effects are associated with lack of space, reduced personalisation and increased time use. Göçer et al. (2018) found that female employees were particularly affected by the lack of storage space, as they were required to clear their desks daily and use a plastic box for personal belongings. CDP was also linked to reduced attachment to the workplace due to the inability to personalise the space, a finding echoed by Van Marrewijk and Van den Ende (2018). In addition, Vos and van der Voordt (2002) noted that CDP can have adverse effects by requiring employees to spend time cleaning their desks; however, in a modern paperless office, this process should not be as time-consuming. Similarly, Babapour Chafi and Rolfö (2019) found that overly strict CDP policies could be time-consuming if employees are required to clear their desks before attending any meeting. While the experiences of implementing CDP have been explored, no previous studies have examined its implications for calculating space usage. To address this shortcoming, the aim of this study is to introduce the concept of passive occupancy, its relation to CDP and how they impact space utilisation.

Desk-level occupancy data was continuously collected with sensors over 6-month periods from October 2022 to March 2023 from ten office building sites located in Finland, Norway and Sweden. The data were collected from working days between 9 a.m. and 4 p.m. and the period was chosen based on the peak office hours in the Nordics. The observation time was 130 days. Each site was using the same commercial SWS platform for collecting the occupancy data and calculating utilisation. In total the data was collected by over 7,500 desk sensors, and the methodology of collecting data was defined by the SWS. The SWS automatically processes the sensor data and reports utilisation percentages. It reports utilisation percentages for desks with just sensor data or sensor data with the effect of the CDP TD that is being used at that particular site.

Study sites’ locations, industries and CDP TDs are introduced in Table 1. The studied companies were from energy, technology, manufacturing and financial industries. Their current CDP TDs ranged from 1.5 h to 12 h. The most common policy was 2 h followed by 12 h. The 12-h policy indicated that the desk must be cleaned at the end of the day, but the worker may keep the desk even though they would leave for a long meeting during the day.

The occupancy data were collected with commercially available wireless PIR presence sensors. These sensors were placed under each workstation at the studied sites. Each sensor measured occupancy continuously and returned either value 0 (no presence) or 1 (presence) and sent the occupancy status to a cloud system through a gateway in 1-min intervals if it detected occupancy. This occupancy data was then processed in the cloud and finally communicated to the user via an end user application. Figure 1 shows the data flow from sensors to the end user application.

All together there were 7,522 sensors included in the study. The average number of sensors per site was 752 sensors. The number of sensors per site is introduced in Figure 2. The least number of sensors was in site C (225 sensors) and the most in site G (1,183 sensors), and the number of sensors indicates the number of monitored workstations and therefore gives a rough estimation of the organisation’s size. Meeting rooms were excluded in this study.

The occupancy data were processed before the CDP TD filter was applied on top of the data. The data processing is demonstrated in Figure 3. In the first step, occupied spans with gaps less than 140 s were connected. This approach reduced noise caused by the person occupying the desk staying still or leaving the desk for a short time. After connecting the spans, all occupied intervals that were less than 80 s long were eliminated, which is visualised in the second step. These short occupancies were removed to neglect occupancies resulting from someone passing the desk or sitting at the desk for an insignificant duration. After these steps, the CDP TD filter could be applied on top of the data. For example, if the CDP TD was set to 120 min, the filter connected spans that were less than 120 min away from each other. In this study, three CDP TDs were examined: 0 min, 60 min and 120 min.

The occupancy status of the desks was communicated to the end user via an application. As mentioned, the desk was interpreted as occupied if the desk sensor sensed movement. It then communicated its status to the end user application and marked the table as reserved. Based on the CDP TD, the desk remained reserved even though the sensor did not recognise movement for the pre-set time. As an example, if a person sits at the desk and then leaves for a meeting, the desk was marked as reserved for an hour still, if that was the CDP delay time.

In this study, time-and-capacity-based utilisation was considered. In a time-based utilisation rate, the occupied time of a space is divided by the total time under analysis. In a capacity-based utilisation rate, the number of individuals using the space is divided by the capacity of the space. Finally, the time-and-capacity-based utilisation rate combines both time and capacity factors. The utilisation rates were calculated as an average utilisation per minute over the chosen hours for each desk. If a desk was occupied for 45 min and then unoccupied for 15 min during an hour, the utilisation would be 75% for that hour if no CDP delay time was applied. The final utilisation percentage was the average utilisation of the desks over a 130-day measurement period. The monitored utilisation was therefore long-term utilisation.

The results indicate that applying a CDP TD significantly influenced the calculated space utilisation, as detailed in Table 2. For instance, the average utilisation was 15.8% with a 0 min CDP TD, 25.5% with a 60 min CDP TD and 29.5% with a 120 min CDP TD. The increase between 0 and 60 min CDP TD (9.7 percentage points) was more pronounced than between 60 and 120 min (4 percentage points). With a 0 min CDP TD, utilisation ranged from 8.7% to 24.7%, indicating generally low active occupancy. In contrast, a 60 min CDP TD exhibited utilisation ranging from 14.8% to 39.9%, whereas a 120 min CDP TD demonstrated a range from 17.7% to 48.2%, which are notably higher. Although these figures may seem low when considering only active occupancy, they are consistent with reports from sources like Vergesense (2023). Moreover, the results are presented as average utilisation. If peak utilisation, defined as the highest momentary utilisation within a given timeframe, were reported, the figures would be considerably higher.

Figure 4 illustrates the changes in calculated utilisation between 0 and 120 min CDP TDs. The darker line represents utilisation with a 120 min CDP TD, while the lighter grey line represents 0 min. These specific TDs were chosen because 120 min was the most frequently applied CDP TD after the 12-h. The difference between the two is expressed in percentage points, with the highest absolute difference observed at location H (23.8 percentage points) and the lowest at site F (8 percentage points). The figure shows that when initial utilisation was low, the increase in utilisation was also modest. To provide a more comprehensive view, relative changes were also calculated for each site, as percentage points alone do not capture the full extent of the variation.

Figure 5 demonstrates the relative change in utilisation between 0 and 120 min CDP TDs, expressed as percentages. Each dot represents the relative change at a specific site, with the dashed lines indicating the largest difference (green line) and the smallest difference (red line). It is worthwhile to note that the sites responded differently to the introduction of CDP TDs. This variation suggests that CDP TD was influenced by differences in working cultures, office layouts and/or desk-sharing ratios. Nevertheless, the relative change was consistently high, ranging from 67.2% at site J to 111.5% at site E. On average, the calculated utilisation increased by 86.4%, indicating that utilisation nearly doubled when passive occupancy was included in the calculation. Neither the initial average utilisation nor the organisation’s size, as determined by the number of sensors, notably affected the relative increase when applying CDP TD. Although the specific reasons for the substantial variation in relative change cannot be determined from this data, the results clearly demonstrate that introducing CDP TD significantly affects calculated utilisation.

In conclusion, CDP TD significantly affected space utilisation. Even if individuals adhered to the CDP by cleaning their desks after the 2-h period, the space utilisation would nearly double. However, it is likely that not everyone consistently remembers to do so, potentially increasing the impact of passive utilisation even further. Unfortunately, this cannot be confirmed with the current data.

The concept of a CDP brings a third state to the traditional binary state of desks when considering utilisation. Traditionally, desks have been categorised as either occupied or unoccupied. However, under a CDP, a third state known as “passive occupancy” is introduced, referring to unoccupied desks with personal belongings left on them. The findings demonstrate that the implementation of a CDP TD considerably affects office space utilisation calculations, which has a substantial impact on the amount of office space companies choose to retain during space re-evaluations. The results highlight two key points. First, there is a notable difference in utilisation rates between scenarios without a CDP and those with a 120-min CDP TD. The introduction of the CDP filter on top of the utilisation data increased utilisation rates by 67.2%–111.5%. Second, these companies had relatively low average utilisation rates, which indicates that they might have excess space. One potential explanation for these relatively low utilisations is that the study considered average utilisation, which smooths out fluctuations from days with exceptionally high or low occupancy.

The results suggest that if these companies were to reduce the number of desks, considering passive occupancy would reveal a different picture of the excess space. For example, if company A only examined their active occupancy, they could conclude that on average they have 798 unused desks from a pool of 921. If they add a 120 min CDP TD, the average number of unused desks decreases to 677, which is substantially less. These numbers do not imply that the company should reduce the number of desks from 921 to 677 because, as mentioned, this calculation is based on average utilisation, and there are days with higher occupancy. However, this data can help the organisation to make decisions about their space needs based on utilisation rate and how popular certain areas are (Einola et al., 2023).

The sensor data does not show whether CDP guidelines were followed, and the behaviour of people may impact the accuracy of adding the CDP filter in two ways. First, if workers complied with the rule and cleaned their desk every time, they planned to be away for more than CDP TD, the results would overestimate the impact of CDP TD. This is because the desk would still be considered occupied after the person left, for the duration of the CDP TD. Second, if employees did not comply with the rule, they may have left their belongings on the desk, and the SWS would still mark the desk as unoccupied after the CDP TD expired. In this case, the impact of passive occupancy would be underestimated. The previous literature on CDP is not sufficient to determine how well people adhere to this policy. For example, Babapour Chafi and Rolfö (2019) found out that the workers complied with the rule, whereas Søiland (2021) stated that it was resisted.

Although CDP can be considered a tool to ensure that employees do not reserve desks they are not planning to use, there are also potential adverse impacts associated with implementing the policy. One of these adverse effects was that employees might waste time if they need to clean their desk before every meeting (Babapour Chafi and Rolfö, 2019). As the results suggest, there was little difference between 60-min and 120-min CDP TDs in terms of utilisation rate. This implies that if a CDP is implemented, it could be longer without significantly impacting the utilisation rate. For employees, this longer CDP TD could be more pleasant than a shorter one. Other adverse aspects of CDP were lack of personalisation (Göçer et al., 2018; Van Marrewijk and Van den Ende, 2018). This lack of personalisation may affect how well the CDP is adhered to, potentially impacting utilisation rates if employees do not clean their desks as required.

Although the implications of this study were clear, there were some limitations. First, this study uses only one method of collecting the data, which was in this case desk-mounted PIR sensors that detect if a certain workstation is occupied. However, there are many ways to collect this data, and the results would become more generalised if other collection methods were also studied. Another limitation of the study is that this article has followed the sensor data processing techniques of the SWS platform which was being used in the 10 offices. Although, the utilisation calculation has been studied and explained in detail in this article, it may also be beneficial to study if there is a more robust way to process the data. One of the limitations of this method was that it included the CDP TD after every occupied span of more than 80 s. This means that there are also spans that were considered passively occupied but were not. Nevertheless, this reflects how the availability of the desks was informed to the occupants in the end user application.

Finally, there are office buildings all over the globe, and the findings apply to all office buildings; however, the detailed results of this study are confined to the three Nordic countries from where the data was collected. Thus, the implications should be tested in other regions and especially in countries where desk-sharing is a more acceptable strategy, as in those countries the behaviour of the employees will be more mature than in other countries.

Office building utilisation is changing, and organisations that accurately assess their space needs can reduce costs and environmental impact while maintaining employee productivity and well-being. Technologies such as presence sensors offer valuable insights into current space use and support space optimisation decisions. However, these sensors primarily capture active occupancy and neglect passive occupancy, where a desk is unoccupied by its “owner” but also unavailable to others.

This study introduced the concept of CDP TD and explained its role in calculating office space utilisation using presence sensor data. The findings demonstrate that excluding passive occupancy results in underestimations of workspace availability, and adding a CDP TD almost doubled the utilisation rate compared to traditional calculations. This underscores the need for more nuanced methods to assess utilisation when using sensor data.

The strong organisational inertia of adopting desk-sharing practices has been described by past scholars, and this should not be underestimated (Dooley, 2017). Robust, trustworthy and accurate data is needed to make these changes for them to be fully implemented throughout the organisation. One potential point of failure to reap the benefits of desk-sharing would be for the space cutting to be too high or too low in the first office buildings that it is applied in, and this would demotivate the managers of the remaining building stock in that organisation to implement the change. They may be demotivated in the situation where not enough space has been cut, as in this case the cost savings would not be significant enough. On the other hand, they would also be discouraged in a situation where too much space has been cut and the new approach has been received negatively by employees.

The results of this study indicate that passive occupancy should be considered when estimating office utilisation based on presence sensor data. However, in the future, using data collection other than desk-mounted sensors could help to validate the robustness of the results from this study. Moreover, observational studies could provide further insights into how well CDP is adhered to and the prevalence of passive occupancy. These steps would strengthen the understanding of utilisation and enhance the robustness of office space optimisation strategies based on sensor data.

This work was supported by Center for Electrical Engineering and Energy Efficiency (STEK), Electrical Contractors’ Association of Finland (STUL), Finnish Electrotechnical Trade Association (STK), K.V. Lindholm Foundation, Kiinteistösäätiö, Finnish Association of Civil Engineers (RIL), Granlund, Helvar, Ramboll and Caverion.