Content uploaded by Giovanni Betti
Author content
All content in this area was uploaded by Giovanni Betti on Jan 28, 2020
Content may be subject to copyright.
HENN Workplace Analytics
Giovanni Betti
(&)
, Saqib Aziz
(&)
, and Gili Ron
HENN, Alexanderstraße 7, 10178 Berlin, Germany
{giovanni.betti,saqib.aziz}@henn.com
Abstract. In this paper we present an experimental methodology for the
evaluation and comparison of indoor workplace qualities. We investigate how
factors such as the overall available spatial connectivity and visual perception in
conjunction with environmental variables such as natural daylight affect the
face-to-face communication potential in office spaces. We visualize these find-
ings through various interactive graphical maps. Subsequently we conclude our
finding by offering a multi-category profiling of the probed spaces, highlighting
potential spatial zones for the various modes of communication that we further
explore in this paper.
Keywords: Workplace environment Social Physics Social dynamics
Digital crafting Workplace qualities Allen curve Connectivity Visibility
Environmental qualities Daylight Factor Generative modelling
Architecture Space syntax
1 Introduction
1.1 Aim
The space that we inhabit inevitably shapes our social and cognitive behavior. In office
spaces therefore the spatial layout plays a significant role in creating a healthy and
productive workplace environment [1]. As illustrated in the annual report by the
Gensler research institute, workplace design and dynamics are in constant need for
improvement [2]. The main task hereby is to provide a workplace environment that is
focused on enabling richer, more fulfilling work environments for employees while
promoting productivity, thus generating a balance between spaces for different work
modes, for example enabling both individual work and collaboration spaces. The report
also highlights the increase of performance, if the spaces for collaboration are design
appropriately. In his book Social Physics [3] Prof. Pentland examines the social psy-
chology and the behavioural patterns of employees in their office environment. Using
empirical studies based on Big-Data collection over an array of professions he has
identified 3 overarching modes of communication and their relevance for productivity.
Pentland refers to them as Energy, Engagement and Exploration [4] and also promotes
the notion that a higher performance of productivity is obtained when a healthy balance
among those modes is in place. In recent years great leaps have been made regarding
the development of technology to monitor building physics and harvest data. These so
called Smart Buildings implement a wide range of cutting edge technology to evaluate
and optimize the performance of buildings [5,6]. While these deployments are
©Springer Nature Switzerland AG 2020
C. Gengnagel et al. (Eds.): DMSB 2019, Impact: Design With All Senses, pp. 636–647, 2020.
https://doi.org/10.1007/978-3-030-29829-6_49
giovanni.betti@henn.com
extremely beneficial in understanding the technical performance of buildings, a similar
scientific focus has to be oriented towards the hidden potentials of spatial programming
[7,8]. In our paper we implement computational strategies to illustrate the inherent
spatial potentials of office spaces for programmatic layout distribution.
2 Methodology
We first explain our methodology and simulation principles, applying them on 3 office
buildings. The case studies shown here are completed projects by the architecture firm
HENN. HENN is a family-owned enterprise in the third generation. The featured
buildings are representative projects for each generation. We then proceed by gener-
ating a multi-category profiling of the probed spaces and offer spatial zoning tactics
based on the modes of communication as described in the following chapters. The aim
is to compare the buildings based on their provisions and spatial programming of
interaction areas. In addition we try to derive an experiential profile for each building.
This process shall provide possible best practices, design strategies and layout sug-
gestions for future building designs.
2.1 Indoor Connectivity Score
In their publication the Organization and Architecture of innovation, Allen and Henn
[9] described how the probability of face-to-face communication depended on the
spatial layout of office spaces. Through an empirical study conducted on office
buildings they derived a 2D correlation, the so-called Henn-Allen curve. This maps the
weekly frequency of interaction between employees as a function of the distance of
their respective workstations. As illustrated in Fig. 1the Henn-Allen curve resembles
an asymptotic function.
Fig. 1. The Henn-Allen curve
HENN Workplace Analytics 637
giovanni.betti@henn.com
It is important to point out that in the same publication, communication in various
forms between peers, groups and departments/professions is identified as the most
important factor for the increase of productivity. “In an organization that relies on
creative solutions to problems, communication for inspiration is absolutely critical. It is
usually spontaneous…cross-disciplinary, cross-functional communication that allows
the development of unusual combinations of ideas that leads to imagination and cre-
ativity”[9]. This finding coincides with the research by Prof. Alex Pentland at the MIT.
In his extensive research about social dynamics, communication happens in three forms
described as Energy, Engagement and Exploration. Each of those modes of commu-
nications relate to different scales.
Energy refers to the magnitude of formal and informal communication between
team members; Engagement refers to the degree of interaction between team members
and their contribution to the group’s discussion; Exploration refects the extent to which
team members engage with other teams working on related initiatives and report their
findings to the group (Fig. 2).
2.2 Indoor Connectivity Score
The probed office spaces in our case studies feature the Osram building (Munich,
1965), the BMW Project House (Munich, 2004) and the Merck Innovation Center
(Darmstadt, 2018). Before conducting any simulation a fair amount of effort went into
the preparation of the individual CAD files to generate a consistent input format for the
simulation process. Most noteworthy being the generation of a 3-dimiensional analysis
meshes representing the accessible circulation area including stairways and lifts
(Fig. 3).
A network of curves is then extracted using the individual mesh faces. In this
process not only the face boundaries are extracted but also diagonal connections of the
mesh vertices to the mesh centroid. This process results in a large set of unique curve
segments that represent the walkable paths in the overall space. The connectivity score
aims to map the overall distance from the centroid of each individual mesh cell to all of
the other mesh cells. To calculate this we use a shortest walk method implemented in
Fig. 2. Modes of communicaction according to Pentland
638 G. Betti et al.
giovanni.betti@henn.com
the algorithmic modelling editor Grasshopper. The shortest walk component calculates
the shortest length path from a starting point to a set of destinations. Along the path for
each curve segment it evaluates the list of all the immediate intersecting curves and
their length. Through this iterative method it can render the next shortest curve segment
length to follow towards the destinations. We used this method to analyze walking
distances in our models. Simple slope inclination enabled us to estimate if the move-
ment is occurring on a horizontal plane (floor), an inclined one (stairs) or through a lift
(vertical). As Fig. 4highlights we allocate different walking speeds and delays to the
each conditions. This causes the calculation of the shortest walk component to consider
also the 3 different mobility conditions and based on those to choose the shortest path
between start and end location. The division of the velocity to the distance results in
time values as an output parameter.
In order to set the resulting time values into a relation to the Allen-Curve we remap
the time values to a defined range, calculated by the maximum communication distance
of 80 m divided by the averaged horizontal walking speed of 1.47 m/s, converting the
distances in time intervals. We then parametrize the Allen-Curve into our workflow and
Fig. 3. Size difference of the 3 building and extracted analysis meshes
Fig. 4. Accessibility analysis
HENN Workplace Analytics 639
giovanni.betti@henn.com
project the new distance values onto the 2D-graph. The process results in a percentage
value identifying the frequency of communication potential depending on travel time.
To ensure that the size of the analysis mesh does not distort the results, we also
multiply the connectivity values for each mesh cell to its mesh area. The sum of all of
the calculated distance values of the mesh cells are then divided by a maximum 3D-
connectivity value. This value describes the possible maximum connectivity score on
an unobstructed plane (Fig. 5).
Through this methodology we are able to generate a magnitude of different output
maps shown in Fig. 6. The main map that we produce illustrates the holistic spatial
connectivity score of the analysis mesh. This visually renders the areas with the highest
to lowest potential for interaction based on the space reachable around them.
2.3 Indoor Visibility Score
The spatial connectivity is only one side of the coin when aiming to determine the
probability of interaction. We also have to factor in the visual perceivable volume at
any given location of an office space. We are more encouraged to interact with
something if we actually see it. In order to evaluate the 3-dimensional human field of
view, we project an array of 3D vectors from eye level from the centroid of each face in
our analysis mesh. By looking at the physiology of human vision, characterized by
central and peripheral vision, we weight our ray smapling as shown in Fig. 7.
Assuming that the human field of view indoors is mainly focused horizontally we
project more rays horizontally. The resolution of the spherical isovist rays can be
controlled as also shown in Fig. 7. For each mesh centroid the isovist rays are flinged
and then are searched for an intersection with any 3D obstructions of the evaluated 3D
space, such as solid walls. Transparent elements are currently not defined as
obstructions.
Fig. 5. Accessibility analysis
640 G. Betti et al.
giovanni.betti@henn.com
After the collision test between the rays and the obstruction we generate a list of the
ray distances to the collision points. Due to the fact that the quality of visual perception
also decreases by the distance, we implemented an experiment to test the boundary of
Fig. 6. Accessibility analysis
Fig. 7. Human field of view and isovist sampling
HENN Workplace Analytics 641
giovanni.betti@henn.com
visual decay as presented by Luftus and Harley [10]. The experiment aims to identify a
marginal distance threshold from where a known face is no longer recognizable. In the
experiment a celebrity’s picture was printed in different sizes, to simulate perception at
varied distances. The percentage of the relative face identification energy is similar to
the Henn-Allen curve and drops in an exponential fashion. After a distance of 45 m the
identification energy as stated by Luftus and Harely is equal to 0%. In the U.S. this
experiment serves to evaluate suspect identification testimonies by witnesses. If a
witness sees a suspect at 50 m distance the probability of the identification has to be
rendered to the findings of the experiment, hence the probability of actual face
recognition is equal to 0% [10]. In office spaces the relationship or identification energy
also relies on different patterns of recognition, due to the fact that the employees also
identify their peers by familiar clothing or physical unique features. For our simulation
we solely include the relative face identification energy to evaluate the visibility factor.
In a similar fashion as the connectivity score we map the isovist ray lengths to a 2D
graph representing the findings by Luftus and Hareley. The sum of the mapped dis-
tances is then divided by the sum of a maximum possible spherical ray lengths con-
trolled by the resolution. The maximum is set to 45 m referencing Luftus & Hareley.
The iterative process is likewise conducted for each mesh centroid in the evaluated
office space and finally mapped onto the initial analysis mesh (Fig. 8).
2.4 Indoor Environmental Score
The workplace quality in office space that are received as pleasant are also largely
defined by the prevailing environmental conditions, such as access to natural daylight,
ventilation, temperature regelation or suitable acoustic attributes [11]. Although all of
the mentioned factors are important to form a pleasant indoor microclimate, we limited
the environmental simulation to the accessibility to natural daylight. The used method
Fig. 8. The ability to recognize familiar faces decays with distance
642 G. Betti et al.
giovanni.betti@henn.com
for this simulation is the Daylight Factor (DF). The Daylight Factor is a long-
established daylight metric, formulated as a simple ratio of internal illuminance to
unobstructed horizontal illuminance under standard CIE overcast sky conditions. It is
expressed as a percentage, and therefore there is no consideration of absolute values.
Luminance distribution of the standard CIE overcast sky is rotationally symmetrical
about the vertical axis and it does not include sun. The Daylight simulation is then also
mapped likewise to the analysis map.
3 Multi-Variable Map
In order to visualize the connectivity, visibility and daylight accessibility score, we
have generated a RGB channel based color mapping strategy. Harnessing the data for
each simulation we generate 3 individual colorized meshes. The color hereby always
ranges from black, referring to the minimum values transforming to the respective RGB
color, indicating the mid-range to white equivalent to the max values in the simulation
data set, as seen in Fig. 9.
–Red color range = Visibility Score
–Green color Range = Connectivity Score
–Blue color range = Natural Daylight Accessibility Score
This allows us to then filter the allocated RGB color scheme and monitor the
frequency of occurrence for each individual mesh face in the analysis mesh. We can
characterize the prevailing condition and define a percentual relationship of occurrence
of the 3 different conditions. We proceed by generating automated graphs, illustrating
the overall percentage of occurrence of the 3 scores as shown in Fig. 9. We can so
Fig. 9. Multi-Variable Map: 3-Category Analysis (Analysis mesh sizes rescaled for diagram)
HENN Workplace Analytics 643
giovanni.betti@henn.com
visualize the character of the office space under consideration through our criteria. This
allows to get a better understanding of the overall performance and programmatic
tendency of the office spaces. A low connectivity and visibility score indicates a high
probability for an arrayed of small and enclosed office room arrangements.
Building on the notion of generating a more discrete usage profile for the probed
space, we refined the analysis process of the RGB color scheme as seen in Fig. 10. The
scheme here is dissolved into the full color spectrum also ranging from black to white.
In order to do that we take the initial RGB colors and allocate them to a 3D vector. This
is achieved by equating the RGB ranges to the spatial axial dimensions of X, Y, Z.
With this method we are also able to measure the intensity of the overlapping vector
inputs. If the resulting 3D vector is in a darker range the intensity for all values
combined is very low. This means that the occurrence of all 3 criteria is poor in the
respective area. If the color scheme is mainly in the white color range then all criteria
are met with a high intensity. This process also introduces a wider color scheme.
Allowing to identify overlapping situation for the conditions. One can know identify an
area that features both a high connectivity and visibility score for example colored in
yellow. Through this filtering method we generate a 6-Categorie Multi-Variable Map.
As shown in Fig. 11 we have mapped the 6 Categories to the original analysis
meshes. We can then also generate automated graphs outlining the percentual rela-
tionship or frequency of occurrence of each mesh face.
Fig. 10. Multi-Variable Map: 6-Category Analysis
644 G. Betti et al.
giovanni.betti@henn.com
4 Experiential Map
The Experiential Map aims to zoom in on to the human level of perception. We
visualize what a human or vicariously an agent would experience, if it would move
through the space from the entrance to a designated workspace. What spatial qualities
for potential interaction and communication are offered on its way. Based on our 6
Categories Analysis we visualize the perceived or experienced sequence of zones
unrolled in various graphical maps (Fig 12).
Fig. 11. Multi-Variable Map: 6-Category Analysis (Analysis mesh sizes rescaled for diagram)
Fig. 12. Experiential analysis
HENN Workplace Analytics 645
giovanni.betti@henn.com
5 Results and Conclusion
The results of the analyzed buildings where carefully reviewed together with the
designers. Besides obvious accordance between the results shown in the 3D heat maps
and the intuitive spatial understanding of the designers, there were also considerable
deviations. For instance the Merck building was characterized as a visual well con-
nected and well-lit building, but the spatial connectivity scores were lower than
expected. This is due to the reachable mass area within the defined range. This does not
necessary mean that the spatial connectivity is by any means ineffective, but just
characterized the functionality of the layout. The Osram building for instances scores
very well in connectivity and visibility but lacks the accessibility to natural daylight.
The results simply reflect the geometry of the building. Osram is one of the first open-
plan office introduced in Germany. Featuring deep open layout spaces on each level.
Therefor if only evaluated on the mass area the visual and spatial connectivity scores
are more prominent than in the Merck building. But if also considering the daylight
factor then the Merck building performs much better.
Overall the methodology helped the designers to visualize the present programming
of the space. They could identify areas that functioned exactly as planned but also
could highlight problematic areas or spaces with hidden potential that could be acti-
vated easily. Subsequently they could also identify typologies that might be better
suited for larger collaborative workspaces to more enclosed and focused office room
arrangements. In this paper we have only introduced an area based characterization of
the office spaces. A more human-centric approach would perhaps mirror this process
more realistically. Mapping the scores not to the total space but for instance to the
individual workstations. We are also aware that in a further iteration our results have to
be compared against real live data that must be collected on-site reflecting the actual
behavioral patterns of the employees. In addition on-site questionnaires evaluating the
sensed space experienced by the employees would further benefit the solidification of
our results.
Afirst attempt towards this goal was pursued by a on-site experiment. The authors
visited the Merck building and used a first person video recording process to navigate
through the entire building. Hereby capturing the experienced spatial qualities. The
recorded movement then got overlayed with a identical path simulated on the corre-
sponding Multi-Variable-Heatmap via a agent. Both the real and digital movement are
merged in a resulting video sequence. In a next step a subject group will compare the
digital findings with the on-site recordings to categorize the observed accordances and
differences between the physical and digtal experienced space through surveys
(Fig. 13).
646 G. Betti et al.
giovanni.betti@henn.com
References
1. Frontczak, M., Schiavon, S., Goins, J., Arens, E., Zhang, H., Wargocki, P.: Quantitative
relationships between occupant satisfaction and satisfaction aspects of indoor environmental
quality and building design. Indoor Air J. 22(2), 119–131 (2012). https://doi.org/10.1111/j.
1600-0668.2011.00745.x
2. Gensler Research Institute: Germany Workplace Survey (2019). gensler.com/research
3. Pentland, A.: Social Physics: How Good Ideas Spread - the Lessons from a New Science.
Penguin, London (2014)
4. Pentland, A.: The new science of building great teams. Harvard Bus. Rev. 90(4), 60–69
(2012)
5. Boston Consulting Group: The edge: creating the world’s most sustainable and most
connected office building by integrating smart technologies and collaborating with suppliers.
Case Study, World Economic Forum, Geneva (2017)
6. Randall, T.: The Smartest building in the world: inside the connected future of architecture.
Case Study, Bloomberg Businessweek (2015)
7. Anderson, C., Bailey, C., Heumann, A., Davis, D.: Augmented space planning: using
procedural generation to automate desk layouts. Int. J. Archit. Comput. 16(2), 164–177
(2018)
8. Manyika, J., Chui, M., Miremadi, M., et al.: A Future that Works: Automation, Employment,
and Productivity. McKinsey Global Institute, New York (2017)
9. Allen, T.J., Henn, G.: The Organization and Architecture of Innovation: Managing the Flow
of Technology. Routledge (2006)
10. Loftus, G.R., Harley, E.M.: Why it is easier to identify someone closer than far away?
Psychon. Bull. Rev. 12(1), 43–65 (2005)
11. Sundstrom, E., Town, J.P., Rice, R.W., Osborn, D.P., Brill, M.: Office noise, satisfaction and
performance. Environ. Behav. 26(2), 195–222 (1994)
Fig. 13. Video sequence overlaying real footage and digital simulation while navigating through
the Merck building
HENN Workplace Analytics 647
giovanni.betti@henn.com
