=Paper=
{{Paper
|id=Vol-1416/Session2Paper2
|storemode=property
|title=Uncovering Sustainability Requirements: An Exploratory Case Study in Canteens
|pdfUrl=https://ceur-ws.org/Vol-1416/Session2Paper2.pdf
|volume=Vol-1416
|dblpUrl=https://dblp.org/rec/conf/re/HuberHG15
}}
==Uncovering Sustainability Requirements: An Exploratory Case Study in Canteens==
https://ceur-ws.org/Vol-1416/Session2Paper2.pdf
Uncovering Sustainability Requirements:
An Exploratory Case Study in Canteens
Martina Z. Huber, Lorenz M. Hilty, Martin Glinz
Department of Informatics
University of Zurich, Switzerland
{mhuber, hilty, glinz}@ifi.uzh.ch
Abstract—Software systems are embedded into daily life and which is negative. For a more detailed classification of effects
as such have significant effects on the behavior and decision of software systems on sustainability see Berkhout & Hertin
making of both their users and the people affected by using these [5] or Hilty & Aebischer [4].
systems. Such effects can be positive or negative. Considering
them in requirements engineering (RE) is an important step Our dependency on software systems [6] and their ubiq-
towards sustainable development, as RE strongly influences the uity [7] within many societies has intensified the impact of
development and the evolution of a software system. enabling effects. Consequently, when striving for sustainable
So far, RE researchers have focused on requirements about development, we must take into account potential enabling
minimizing negative environmental effects. However, effects that effects when developing a software system, in order to avoid
are enabled by using a software system can also be posi-
tive. For example, a software system could motivate its users negative effects and leverage positive ones.
to take environment-friendly decisions. Corresponding require- The importance of considering sustainable development in
ments about such positive enabling effects have been far less software engineering has been emphasized by researchers
addressed. in the field [6], [8], [9]. Due to its early influence in the
In this paper, we present an exploratory case study where development process of a software system, requirements
we elicited requirements about positive enabling effects with
respect to environmental sustainability. The project we studied engineering (RE) is considered to have the biggest impact
is about extending an existing decision support system for meal on the eventual effects of a software system [10]. As such,
planning in canteens by game-based elements. The extended RE provides promising opportunities to affect the transition
system shall motivate meal planners who work in canteens to towards sustainable development significantly [2], [3], [10],
make environment-friendly choices. [6], [8].
Our contribution is an exemplar of concrete sustainability
requirements as well as insight into the nature of sustainability
Current RE contributions considering sustainable develop-
requirements about positive effects that are enabled by the use ment have focused on minimizing negative environmental
of a software system. effects and referred to corresponding requirements as quality
Index Terms—enabling effects; sustainability requirements; requirements, e.g. [3], [7], [1], [10]. Sustainability requirements
sustainable software systems; gamification; exploratory study; related to positive effects however, have not received much
attention and can not be limited to quality aspects.
I. I NTRODUCTION Our goal is to better understand sustainability requirements
Any software system and its utilization have effects causing related to positive effects, specifically the ones that are enabled
environmental impacts [1], [2], [3]. We categorize these by end-users while using a software system. This paper
effects into two groups. One group relates to the direct describes an exploratory case study in which we elicit and
effects of running a software system, e.g., the energy and discuss such sustainability requirements for a decision support
resources required for powering and cooling the computing system (DSS) that is extended with game-based elements for
equipment the system runs on or connects to. Such effects are motivating meal planners to make choices that reduce the CO2
usually considered to be negative with regard to sustainable value of their meals.
development. The second group relates to enabling effects. The contribution of our study is twofold. Firstly, we present
These are effects that are enabled by using a software system. an exemplar of concrete sustainability requirements related to
Enabling effects can be the result of automated processes positive enabling effects. Secondly, we discuss these sustain-
and/or of human behavior that is influenced by using the ability requirements. The study reveals that when considering
software system. With regard to sustainable development, positive enabling effects (i) corresponding requirements include
enabling effects can be both positive and negative [4]. For new requirements and existing requirements that become
example, heating management software can reduce the energy more important, and (ii) treating these requirements as quality
consumption of a building significantly, which is positive. requirements is inadequate as we found functional requirements
An e-commerce system providing business-to-consumer and and constraints. The study further provides some indication
consumer-to-consumer sales services may influence human that (iii) corresponding requirements can be classified into
behavior positively by encouraging people to sell used goods requirements about integration, (meaningful) representation,
so that they get re-used, but it can also lead to over-consumption, and (fair) comparison, and that (iv) indirect stakeholders who
Copyright c 2015 for this paper by its authors. Copying permitted for private and academic purposes.
are no system users, but are affected by the use of the system emitted less CO2 than possible alternatives. To achieve this
and influence its success, are important. goal, the Compass Group (Switzerland) decided to order an
This paper is outlined as follows. We first provide back- extension of the existing DSS with game-based techniques
ground on the case study and related work on RE with regard for further processing the calculated LCA data and presenting
to sustainable development. In Sect. III we describe our research them to the meal planners in a recurrent report.
approach, the study design and the research questions. The 3) The Context of our Case Study: The case study that
results and threats to validity are presented in Sect. IV, followed we present in this paper contributes to the second part of
by a discussion in Sect. V and conclusion and future work in the CarbonFoodPrint project. In the framework of a research
Sect. VI. collaboration with Eaternity and the Compass Group (Switzer-
land), the sustainability requirements for the new system were
II. BACKGROUND AND E XISTING W ORK elicited and studied by the first author of this paper, together
In this section, we provide information about the case study with a graduate student. As the focus on the CO2 emission
as well as relevant background knowledge for readers less caused by the production and transportation of ingredients
familiar with RE in the context of sustainable development. was given, our study is confined to sustainability requirements
Further, we describe some earlier RE research which is relevant with respect to CO2 which, actually, is only one aspect of
in the context of our study. sustainable development.
A. The Case Study B. Requirements Engineering and Sustainability Requirements
1
This case study is part of the project CarbonFoodPrint Software systems are embedded in their environment which
initiated by Eaternity2 , a company that provides software- by nature is in a permanent process of change. As such,
supported solutions helping people to make their own eating developing and evolving a sustainable software system is an
habits climate-friendly. In this project Eaternity collaborates ongoing process. Typically, the decisions that shape a software
with the Swiss branch of the Compass Group 3 , a world-leading system are taken during RE [3].
food service company. 1) Sustainability: The term sustainability has been used
1) The Current State: The process of planning meals that in different contexts and overused for several purposes. To
is applied within the Compass Group (Switzerland) consists avoid any misunderstanding, we briefly define the terminology
of two phases: First, a small culinary development team in used in this paper. Based on the “Brundtland definition” [12],
the headquarters composes and develops meal propositions Christen proposes to conceptualize sustainability as an “attempt
for all canteens for the forthcoming months. Based on the to grant the right to a decent life to all living human beings
categories offered in a canteen (e.g. meat, veggie, low carb) without jeopardizing the opportunity to live decently in future”
the proposed meals are accordingly filtered and sent over ( [13], p. 2). As such he emphasizes that sustainability does not
an SAP-based system to the canteens. In the second phase, solely focus on future generations, but also on human beings
the meal planners working in the canteens adapt the meals living now. He also argues that sustainability is not limited to
based on specific canteen-related criteria, (e.g., their budget and sustaining aspects, but also addresses enabling aspects. By
customers’ preferences). A decision support system (DSS) is definition, sustainability is a global (temporal and spatial)
used to support the meal planners to plan meals by providing concept which makes it clear that no single technology can be
specific functions; in particular, the possibility of selecting sustainable in this sense. However, technology can support the
ingredients (e.g., tomatoes), meal components (e.g., tomato transition towards sustainable development [4]. Based on this
sauce or spaghetti) and whole meals (e.g., spaghetti with tomato notion of sustainability we define the following terms.
sauce) from a large recipe database, together with corresponding • A sustainable software system is a software system that
information on nutrition factors and costs. supports the transition towards sustainable development.
2) The Project Context: The overall goal of the CarbonFood- • A sustainability requirement is a requirement for a sus-
Print project is to motivate meal planners to select ingredients tainable software system which concerns sustainability.
whose production and transportation emitted less CO2 than • A positive enabling effect is an effect that is enabled by
possible alternatives. The project consists of two parts. The first using a software system and positively contributes to the
part focuses on the calculation of CO2 emissions by applying transition towards sustainable development.
life-cycle assessment (LCA) [11]. This is a specific technique to Different metaphorical descriptions of the roles of environ-
address the environmental aspects and potential environmental ment, society, and economy in sustainable development exist.
impacts such as use of resources and the environmental As the economic system is part of human society, which in
consequences of releases throughout a product’s life cycle. turn is part of the environment, we use the metaphor of nested
The second part of the project focuses on the utilization and circles, where economy is represented by the inner circle,
representation of these data to motivate the meal planners to society by the middle one and environment by the outer one.
select ingredients whose production and transportation have For more information see for example Hilty & Aebischer [4].
1 http://bit.ly/1L2Qzcc 2) State of Research: The high importance of addressing
2 http://eaternity.com the concept of sustainable development within the process
3 http://welcome.compass-group.ch/ of requirements elicitation has been recognized by the RE
community, e.g., [6] [8] [9]. However, as noted by Becker [14], in the context of a project for extending an existing software
this has not yet been transferred into practice successfully. system with game-based mechanics.
Prior RE research has mainly conceptualized sustainability Research Questions. From this goal we derived two re-
requirements as quality requirements (i.e., a sub-category of search questions:
specific quality requirements in the taxonomy introduced by RQ1: What is specific about requirements concerning positive
Glinz [15]). enabling effects?
Thereby, the focus was on goal modeling processes e.g. RQ2: How can game-based mechanics motivate positive
by regarding sustainability as a trade-off between business enabling effects when extending existing software systems?
goals [16], by using the idea of generic goal refinement as a
checklist for sustainability requirements [3], by treating them B. Study Design
similarly to conflicting goals of budget restrictions and quality We followed a “mixed methods” approach [23], consisting
improvements [17], by suggestion how to align the objective of of three sequential steps: a contextual inquiry, semi-structured
environmental sustainability with the other objectives [1], and interviews and an online questionnaire. We used this approach
by building upon different levels of impacts [18]. Further, for investigating our research questions from more than one
Roher & Richardson work on patterns for sustainability perspective, thus getting more thorough results. As mentioned
requirements [19]. in Sect. II.A, the tasks of eliciting the requirements and
conducting the study were both performed by the first author
C. Sustainability Requirements Related to Enabling Effects of this paper, together with a graduate student.
Most existing work on sustainability requirmements as 1) Study Setup: The process for all three steps and the
described above focus on direct effects (cf. the classification questions to be asked were elaborated by the first author of
presented in the introduction) and treat sustainability require- this paper with support from the graduate student, then reviewed
ments as goals or as specific quality requirements. Sustainability by a group of RE researchers and finally improved according
requirements related to enabling effects of a software system to the feedback received. The interviews and the questionnaire
are far less addressed both in research and – to our experience were both piloted with people neither specifically related to the
– also in industry. A possible explanation is that in most cases domain of RE nor to the one of sustainability. This approach
direct effects are directly connected to economic goals, whereas was chosen to make sure the questions are clear to participants
for enabling effects, such a relation is hard to establish in most who are unfamiliar with these domains [24].
cases. The study was carried out over a period of four months and
Nevertheless, as outlined by Wang [20], research shows included eight steps: (1) Preparing the contextual inquiry, (2)
that while considerations about sustainable development are carrying out the contextual inquiry and evaluating the results,
becoming more relevant in societies, positive environmental (3) designing the interview questions, (4) conducting pilot
effects positively impact the value chain and the image of interviews, (5) carrying out the interviews and evaluating the
a company. Consequently, requirements related to enabling results, (6) designing the questions following the guidelines for
effects of software systems increasingly gain relevance for creating a questionnaire by Kitchenham & Pfleeger [24], (7)
both companies and their stakeholders. publishing the questionnaire, (8) evaluating the questionnaire
results. Note that the results of the pilot interviews were not
III. R ESEARCH M ETHODOLOGY AND S TUDY D ESIGN included in the data analysis.
For choosing our research methodology, we considered the All steps of the study were conducted in German. Conse-
following facts: (i) We wanted to analyze and better understand quently, the questions and interviewee quotes reported in this
sustainability requirements with respect to positive enabling paper are our translations of the German originals.
effects, (ii) there is little knowledge available about this kind 2) Selection of Participants: The Compass Group (Switzer-
of requirements, (iii) we had the opportunity to study a real land) selected the participants for the contextual inquiry and
industrial project in this context. Given this situation, we chose the interviews. However, we could provide our criteria for the
an exploratory case study as our research methodology. Such a selection process. The following of our criteria were accepted
study enables an in-depth investigation of a phenomenon in its and applied by the company: (1) all participants are responsible
context [21] and is specifically suitable when little knowledge for the meal planning process and as such are direct end-users
about the subject is available [22]. Further, the results of an of the DSS; (2) the contextual inquiriy is conducted in two
exploratory study form the basis for both theory generation [21] sessions, one in the headquarters of the company and one in
and constructive solution design. a canteen; (3) at least 15 meal planners are selected for the
interviews; (4) the group of participants is heterogeneous with
A. Research Goal and Research Questions regard to canteen size, region, and work sector (“Business &
According to our research plan and the given project context, Industry” (B&I) and “Eductation” (Edu)). We did not include
we formulated our research goal as follows. gender, age or nationality into our selection criteria, since we
Goal. Analyze sustainability requirements for the purpose considered them as irrelevant for the purpose of the study.
of developing sustainable software systems with respect to Eventually, 19 meal planners working in different canteens
positive enabling effects from the viewpoint of the end-users participated in the semi-structured interviews.
The URL of the online questionnaire was sent to all student and then discussed between them in a retrospection
meal planners (about 150). 67 participants finished the on- session. As mentioned above, the interview questions as well
line questionnaire. However, seven of them did not answer as the interviews were piloted before conducting the actual
the question about current reasons for changing ingredients. interviews.
The questionnaire results presented here come from the 60 5) Online Questionnaire: We used the results from the
meal planners who answered all questions (this includes all interviews as a basis for designing the online questionnaire.
participants of the interviews and the contextual inquiry). Our goal was to elicit quantitative information about important
3) Contextual Inquiry: To understand the current situation, aspects of sustainability requirements from a sample of involved
i.e., how meal planners currently work and apply the existing people which is larger than the number of meal planners
DSS, we conducted a contextual inquiry in two sessions: interviewed. We exploited the majority of the interview
the first one at the headquarters of the Compass Group questions in the online questionnaire, omitting the ones that
(Switzerland) and the second one in one of the canteens. focus on the end-user’s attitude towards the project. Table I
Contextual inquiry [25] is an elicitation technique that studies shows four representative questions (QQ-1–QQ-4) that we
stakeholders in the field, bringing the requirements engineer in further analyze in this study. Semantic differential scales [29]
contact with the stakeholders in their real work environment. were applied to evaluate the participants’ attitude. This type
Thereby, the requirements engineer takes a role similar to of scale is similar to the Likert scale [30] with the benefit of
the one of an apprentice, asking questions while observing revealing both the direction and the intensity of each opinion.
the work process. By allowing requirements engineers and For questions about familiar topics we used an even scale
stakeholder representatives to work together and to share (four point), for questions where we expected less or non-
insights, a contextual inquiry enables a full understanding familiarity, we used an odd scale (three or five point, including a
of the work practices in the specific work environment. As neutral point). This approach is generally suggested for defining
a contextual inquiry is exploratory and open-ended, we did the number of alternatives given in ordinal scales [31]. The
not prepare questions beforehand, with the exception of some link to the questionnaire was published over the intranet of
ice-breaking questions for starting the inquiry sessions. the Compass Group (Switzerland) together with background
4) Semi-Structured Interviews: The questions for the inter- information about the project, who we are, the goal of the
views were grouped into three parts: demographics and current study and criteria for participation.
work routines, usability, and motivation. While the questions Technically, we used an online questionnaire tool5 for creat-
about the first two parts build on the results of the contextual ing the questionnaire. The full set of questionnaire questions
inquiry, the ones about motivation regarding the game-based (in German) and the questionnaire design are available at 6 .
aspects refer to the results from our previous research on C. Collecting the Data
requirements for game-based approaches motivating sustainable
consumption [26]. Four representative interview questions are Data collection started in July 2014 with the two contextual
given in Table I. IQ-1 is from part one about current work inquiry sessions and ended in October 2014 when the online
routines, IQ-2 is from part two about usability, and IQ-3 as questionnaire was online for two weeks.
well as I-Q4 are from part three about motivation, however, The full data set has a size of 81, comprising the data from
I-Q4 also affects part two. The full set of interview questions two contextual inquiry sessions, 19 interviews and 60 fully
(in German) is available at 4 . completed questionnaires. It turned out that the first contextual
The interviews lasted between 25 and 45 minutes on average. inquiry session had primarily served for making the researchers
The time difference can be explained by our approach of familiar with the context of the project, so we excluded it from
conducting semi-structured interviews with a mix of open- the data set. Further, one interview was not recorded due to a
ended and specific questions. We did this for enabling the technical problem which we realized only after the interview
elicitation of both foreseen and unexpected information [27]. was finished. Hence, we also excluded that interview from
This was specifically relevant since the domain knowledge the data set. All interviewees as well as the participant of the
of favorabale enabling effects is still immature in RE. It also contextual inquiry also answered the questions in the online
helped to build a positive rapport with the interviewees [28]. questionnaire. So we have a total of 79 data points from a total
All interviews were conducted over Skype by calling the of 60 participants for analysis. With respect to the canteen
interviewees on their business phones. This approach was sectors, i.e., Business & Industry (B&I) vs. Education (Edu),
chosen because the interviewees were distributed over the the data are distributed as follows: (i) contextual inquiry session:
whole country. So visiting them all would have been too costly zero B&I, one Edu; (ii) interviews: fourteen B&I, four Edu;
and also not possible in the timeframe given for the interviews. (iii) online questionnaire: fifty B&I, ten Edu.
Half of all interviews were conducted by the first author of D. Participant Demographics
this paper, the other half by her graduate student. To align Some demographic information about the participants is
the interview styles and reduce observer bias, the first two summarized in Table II. 83% of all participants work in
interviews were conducted jointly by the first author and her
5 http://ww3.unipark.de
4 http://bit.ly/1QyuR3i 6 http://bit.ly/1FiFKdO
TABLE I
I NTERVIEW Q UESTIONS (IQ-) AND Q UESTIONNAIRE Q UESTIONS (QQ-)
Identifier Original Question (English translation)
Question Id1
IQ-1 2.d According to what criteria do you compose a meal?
IQ-2 3.a Can you imagine that the idea of reducing the CO2 value of meals due to the choice of ingredients is realizable in your
canteen?
IQ-3 4.a Are you additionally to the CO2 value interested in the following information? i.What the CO2 value means e.g. how many
kilometers driven by car do correspond to it? i.1 Are you further interested in other representations? If so in which?
IQ-4 6 Do you believe that the CO2 values of different canteens can be compared with each other in such a report?
QQ-1 4.1.1 What does motivate you to change a meal? Formulated in a clearer way, how strong does one of your changes affecting a
meal component or ingredient depend on the following criteria? (costs, variety, customer preference, season, compliance,
environmental aspects)
QQ-2 5.3 How should the CO2 emission value of your meals additionally be represented in order to raise your interest? (number of
kilometers a mean of transportation has to make; amount of energy an ordinary private household has to use; number of days
an ordinary private household has to be heated in order to emit the same amount of CO2 ; number of trees that are needed to
compensate the same amount of CO2 );
QQ-3 7.1.1 How good do you think is the following information? Segmentation of the CO2 based on the process steps; meal components;
origin of ingredients
QQ-4 6.1 How strong have the following factors to be considered in order you perceive the comparison of the CO2 emission of different
canteens to be fair? (size of the canteen according to the number of cooked meals; the number of employees working in the
kitchen; the number of meal categories served in a canteen; the customer preferences)
“IQ” refers to the interview questions and “QQ” to the questionnaire questions. The column “Original Question Id” refers to the corresponding original question
id of the interview or the questionnaire, available under 1 : http://www.ifi.uzh.ch/rerg/research/stakeholderengagement/garuso
TABLE II into sustainability requirements for positive enabling effects
OVERVIEW OF ALL PARTICIPANTS WHOSE DATA WAS EVALUATED with regard to sustainable development.
1) Contextual Inquiry: Information from the contextual
Attributes / Domain Business & Education
Industry
inquiry sessions was first structured to identify main work
processes which then were used for defining the interview
Percentage of all participants 83% 17%
questions. In the presentation of the results in the next section,
Average # of years in position 7 5 the data from the contextual inquiry is not analyzed separately,
Median # of years in position 5 3 but presented together with the data from the interviews.
Average # of meal categories 3 4 2) Semi-Structured Interviews: To be able to better evaluate
the data from the interviews we structured the data by first
< 150 meals/day 42% 20%
150-499 meals/day 42% 40%
transcribing and then coding them according to the process
500-1999 meals/day 16% 40%
described by Seaman [33]. The corresponding codes are listed
in Table III. The results of this analysis were used to define
the questions and structure of the online questionnaire.
3) Online Questionnaire: As it is common in exploratory
B&I canteens and 17% in Edu canteens. On average, the
studies, we visually analyzed the quantitative data from the
participants have more than five years experience in their
questionnaires, choosing divergent stacked bar charts.
position. Regarding meal variety and number, participants
working in B&I canteens have three meal categories on IV. R ESULTS
average (e.g., meal with meat, vegetarian, low-carb), while In this section we present our results. As usual when
participants working in Edu canteens have four. Around 40% presenting qualitative results, the data are complemented with
of all participants from both sectors work in canteens that quotations from the interviewees. The quotations are written
produce between 150 and 499 meals per day. 42% of the B&I in italic, the interviewee who stated the quote is indicated
canteens and 20% of the Edu canteens produce less than 150 in brackets by “I-” followed by the number of the interview.
meals per day, and more than 500 meals are daily produced in We coded relevant information in the qualitative results to
16% of the B&I canteens and in 40% of the Edu canteens. structure and quantify them; these codes are underlined and in
brackets. Table III lists the sixteen codes that we used, together
E. Analyzing the Data with the frequency of their appearance in the interviews. The
All data is exploratory, which means we did not pre-specify questionnaire results for QQ-1–QQ-4 (cf. Table I) are visualized
a hypothesis as it is done in a confirmatory analysis. The in Figures 1-4 in the same order as the corresponding questions.
end-product of exploratory data analysis is rather suggesting We present the results grouped by topics. Within each topic,
patterns for further studies and providing hypothetical insight we first present the results from the interviews regarding this
into these patterns instead of statistical figures [32]. As such topic, and then the corresponding results from the online
we did not apply any statistical tests, but provide first insights questionnaire.
A. Requirements Looked through the Sustainability Lens TABLE III
OVERVIEW OF I NTERVIEW R ESULTS
The first aspect we investigated was whether and how the
current set of requirements changes from the end-users’ perspec- Codes Used to Quantify the Qualitative Re- Frequency Origin
tive when adding the dimension of sustainable development to sults (Semantic:[effect_causedBy_inSystem]) Num- Per- IQ Spont.
ber cent
the domain context (RQ1). We first focused on actually existing
requirements causing the participants to change ingredients in changeIngredients_customers_current 13 72 X
their current work process. Second, we elicited requirements changeIngredients_costs_current 12 67 X
that the participants perceive as important if they had to select changeIngredients_variety_current 6 33 X
ingredients with respect to CO2 emission. This subsection successCO2 Red_changingIngredients_new 13 72 X
highlights the aggregated results related to this aspect. constraintCO2 Red_costs_new 9 50 X
constraintCO2 Red_customers_new 9 50 X
1) Current Requirements Motivating Change: We specif-
constraintCO2 Red_space_new 2 11 X
ically asked the interviewees about reasons for changing
successCO2 Red_integratedInWorkProcess_new 13 72 X
ingredients in the proposed meals they get from the culinary
successCO2 Red_alternativesShown_new 6 33 X
development team (IQ-1). The three criteria mostly mentioned
are customers’ preferences, costs and variety. successCO2 Red_CO2 Rep_new 17 94 X
As indicated in Table III for thirteen interviewees successCO2 Red_CO2 RepGraphically_new 5 28 X
customers’ preferences (changeIngredients_customers_current) successCO2 Red_CO2 PerOrigin_new 2 11 X
are a strong reason to change ingredients, twelve successCO2 Red_CO2 PerProcessSteps_new 3 17 X
emphasized the relevance of cost restrictions successCO2 Red_CO2 PerComponents_new 4 22 X
(changeIngredients_costs_current) and six mentioned the comparisonCO2 Possible_gamification_new 13 72 X
variety (changeIngredients_variety_current) of their meals as comparisonCO2 Possible_numberOfMeals_new 8 44 X
an important reason. Interviewee I-8, for example, highlighted The codes are presented in the order that they appear in the paper; the upper
both variety and customers’ preferences: “I do it according to group is referred to in Sect. IV.A, the lower one in Sect. IV.B. “Red” stands
for “Reduction”, “Rep” for “Represented”. In column “Origin”, “IQ” indicates
the following criteria, such that there is variety. Theoretically, codes that stem from explicitly asked interview questions, while “Spont”
pork is the cheapest meat we can get, but I nevertheless look indicates codes found in information spontaneously raised by interviewees.
that it is only served once a week. Also a little bit because of
our Muslims (...)”. Cost constraints (constraintCO2 Red_costs_new) were stated
In the online questionnaire, we further explored these results by nine interviewees to challenge the successful consideration
by asking the participants to rate the importance of costs, of CO2 emission when selecting ingredients. Three of them
variety, customers’ preferences, together with the seasons, regard cost restrictions as a major criteria for a project failure.
compliance with suppliers, and the environment as a reason The rest of them is not that strict, however they mentioned
for them to change the ingredients (QQ-1). Participants could that local and organic products usually are more expensive
rate the importance on a semantic differential scale of four than non-organic products or products from farther away. They
criteria: “Plays no role at all”, “Plays a minor role”, “Certainly highlighted that if their budget remains the same, customers
plays a role”, “This criterion is one of the most important had to pay the cost difference. Interviewee I-17 considered the
reasons for change”. The results shown in Fig. 1 support what relevance of a company-wide change: “Well it [the success]
we found in the interviews: Customers’ preferences and costs depends on what kind of food it is. In my opinion, if costs
are the two most important reasons to change ingredients in raise, we have to pass these costs on to our customers in order
both Edu and B&I canteens. Variety and season come next. to stay in the green zone [with the costs]. (...) Actually, the
Participants working in Edu canteens rate these two criteria as whole company should have to participate.”. I-14 stated: “(...)
equally important, while participants working in B&I canteens We have to get the cheapest products in order to fulfill the
rate season to be slightly more important than variety. terms of the company and meet the demands of the customers
2) Considering Sustainability Requirements: When asked (...) the customers have to rethink because, if I buy a regional
the interviewees whether they can imagine reducing the CO2 product, this has immediate influence on my costs. ”.
emission of their meals by selecting specific ingredients (IQ- Customers’ preferences are further regarded as a
2), thirteen were positive about having the potential to do constraint that challenges the success of the project
so (successCO2 Red_changingIngredients_new). However, the (constraintCO2 Red_customers_new) by nine interviewees.
interviewees agreed that this goal is only achievable when Seven of them can imagine to possibly manage this challenge
considering certain constraints. The results described below by involving the customers (e.g., by explaining the effects). The
show that (i) costs and customers’ preferences are perceived as other two do not think that customers will change their eating
the two most relevant constraints if the context of sustainable habits or pay more for climate-friendly food. I-14 stated for
development is added; (ii) the influence of these two constraints example: “Well, the cooks would like to consider this [the CO2
becomes stronger in this context and new constraints become emission related to the ingredients]. It has been a nonsense to
relevant; (iii) an integration of the game-based extension into offer tomatoes in December. However, it is a fact that not us,
the existing DSS is important. but the customers do rule the market (...) In my opinion the
Customer preference
Canteens
Canteens #trees for compensation
Cost
-E d u
#heating days in a household
-E d u
Variety
Season #kilometers by any mean of transportation
Compliance
Environment Eletricity used in a private household
Customer preference 0 20 40 60 80 100 120 140 160 180 200
#trees for compensation
-B u s i n e s s
Canteens
-B u s i n e s s
Canteens
Cost
Variety #heating days in a household
Season #kilometers by any mean of transportation
Compliance
Environment % of respondents Eletricity used in a private household % of respondents
100 80 60 40 20 0 20 40 60 80 100 100 80 60 40 20 0 20 40 60 80 100
Plays no role at all Plays a minor role Certainly plays a role This makes absolutely no sense This seams not very meaningful I cannot answer this question
This criterium is one of the most important reasons for change I regard this as rather meaningful Absolutely, this is absolutely meaningful
Fig. 1. Evaluation of reasons to change ingredients in the current work process Fig. 2. Evaluation of alternatives for representing the CO2 value
awareness of the customers is not yet there.” (...) because we do not know what consumes how much [CO2 ].
The size of the work space was highlighted by two inter- For example, when I drive home with my car, I know I produce
viewees (constraintCO2 Red_space_new) as a major constraint. CO2 , but I do not know how much. And this is the reason why
Little space in the kitchen affects work processes and as such we don’t know what is a good value and what is a bad one.”
the choice of ingredients. I-12 put it this way: “Our restaurantsFive interviewees highlighted the relevance of using graphics
are not all equipped in the same way (...) Some [of us] really and pictures (successCO2 Red_CO2 RepGraphically_new). I-14
must additionally plan their meals based on the available space explained it this way, “(...) something like ’[with this amount
[to process the food] and the storage possibilities and then of CO2 ] you could have driven from here to there with a truck’
decide whether they buy fresh broccoli or frozen broccoli.” and then showing something similar to Google maps. This
The proper integration (successCO2 Red_integratedIn- way, it becomes visible right away that I could have driven to
WorkProcess_new) of the game-based extension in the Marseilles.”
existing DSS is regarded as highly important by thirteen
The questionnaire results underline these results. Together
interviewees. For example, six of them said that it is highly
with the given example on the amount of kilometers we
important to have immediate access to alternative ingredients
provided three more representation options to rate for (QQ-2):
with a lower CO2 emission value while planning meals over
the number of trees needed to compensate the amount of CO2
the system (successCO2 Red_alternativesShown_new). I-2
emitted, the used heating energy, and the used electricity in
emphasized: “(...) there really must be alternatives with which
an average household (Fig. 2). These four options could be
the CO2 emission value can be reduced. Showing a direct
rated on a semantic differential scale of 5 criteria: “This makes
alternative in the sense of ‘tomatoes from Italy instead of
absolutely no sense”, “This seams not very meaningful”, “I
cherry tomatoes from overseas’ should be possible.” In this
can’t answer this question”, “I regard this as rather meaningful”,
context, I-7 highlighted time pressure as a reason: "When
and “Absolutely, I think this is absolutely meaningful”. The
we get the [suggested] meal plan, it [the system] has to be
results show that the number of trees needed for compensation
ready with the CO2 emission values. Because we cannot
and the number of kilometers that can be driven are similarly
search extensively for this information on the lists of the
perceived as highly meaningful for representing the amount of
vegetable, meat, or fish suppliers (...) this [information] has to
CO2 emissions. Participants from both sectors were much less
be integrated and then we can work.”
interested in a representation by the energy consumption in a
B. Game-based Mechanics for Positive Enabling Effects household of both electricity and heating.
The second aspect we investigated was about how game- Segmentation of the CO2 value was raised by four intervie-
based mechanics have to be applied to motivate positive wees. Two of them wanted a segmentation based on the origin
enabling effects in the given context (RQ2). Our focus was on of ingredients (successCO2 Red_CO2 PerOrigin_new) e.g., the
a meaningful representation of the CO2 emission value and on CO 2 value of a tomato from Greece compared to one from Italy.
relevant factors for comparing the values of different canteens Three interviewees were interested in segmenting the CO2 value
with each other. In this sub-section we present the aggregated according to the two process steps of production and trans-
results. portation (successCO2 Red_CO2 PerProcessSteps_new), and all
We asked the interviewees if the CO2 emission shall be four found it relevant to segment the meals with regard to
represented (successCO2 Red_CO2 Rep_new) by more familiar their components (successCO2 Red_CO2 PerComponents_new),
measures (IQ-3) and provided the number of kilometers driven e.g., the two components pasta and pasta sauce. I-14 made
(i.e., how many kilometers one could drive for emitting the same this very clear: “When you do something like this [the project
amount of CO2 ) as an example. Seventeen interviewees agreed as a whole], I think it is good that we have a learning effect,
on the importance of a meaningful representation. I-12 added something like ’Ahaaaaaa, there it is, this shrimp has messed
that such a value would even be more meaningful if represented up our whole meal statistics (...) This way rethinking happens.
by a journey. “It would be good if you could say, we have We are a bit practice-oriented.”
saved that amount of CO2 , this is enough to travel from here to These results were exploited further in the questionnaire by
Moscow or simply, you could travel that far with this amount asking to rate the segmentation factors origin, meal components,
Canteens
Origin of ingredients Customer preferences
Canteens
Location of a canteen
-E d u
-E d u
Meal components #offered meal categories
#employees working in the kitchen
Production & Transportation
#cooked meals
0 20 40 60 80 100 120 140 160 180 200 220
Origin of ingredients Customer preferences 0 10 20 30 40 50 60 70 80 90
-B u s i n e s s
Canteens
-B u s i n e s s
Canteens
finde ich gar nicht gut finde ich eher schlecht finde ich eher gut Auf jeden Fall, finde ich super. Location of a canteen
Meal components
#offered meal categories
Production & Transportation #employees working in the kitchen
% of respondents
100 80 60 40 20 0 20 40 60 80 100
#cooked meals % of respondents
0 10 20 30 40 50 60 70 80 90
I don't like this at all I think this is rather bad
I think this is rather good Absolutely, I think this is awesome Plays no role at all !! Plays a significant role
Fig. 3. Evaluation of segmentation factors Fig. 4. Importance of different factors regarding comparison
and process steps (QQ-3) on a semantic differential scale of for selection bias when analyzing the interviews. Answering
four criteria: “I don’t like this at all”, “I think this is rather the online questionnaire was voluntary, which may cause the
bad”, “I think this is rather good”, “Absolutely, I think this is results to be biased due to highly supportive or highly non-
awesome”. The results show that the majority of both sectors supportive participants. Since the subject of the study was
is interested in comparing ingredients based on their origin equally interesting for both supporters and opponents, we do
(Fig. 3). not believe that the outcome of our study is affected by this
As game-based techniques typically involve comparisons, we potential threat. We mitigated potential maturation problems by
specifically asked whether the CO2 value of different canteens scheduling the interviews during working hours and limiting
can be compared with each other (IQ-4); thirteen interviewees their duration to 20-40 minutes.
said yes (comparisonCO2 Possible_gamification_new). Eight of
External Validity refers to the extent of being able to
them stated that the most important condition to consider for
generalize the results. The biggest limitation to the external
enabling comparisons between canteens is the number of meals
validity of our study is the fact that (i) the study was conducted
produced (comparisonCO2 Possible_numberOfMeals_new).
in a single company and (ii) the project was limited to positive
I-17 got very specific: “You cannot compare a small canteen
enabling effects with respect to CO2 emission. However, the
with a large one (...) If I compare two small canteens with
setting of our study is typical for systems in the service domain:
each other that sell on average approximately sixty to eighty
employees of a service company use a software-based system as
meals per day, then I think this is comparable. (...) I think here
end-users to provide services to the customers of the company,
we should differentiate such that we only compare very similar
while the customers’ preferences as well as cost considerations
canteens, also with respect to the location.”
have a major influence on the system’s requirements. Further,
These interview results are supported by the results from
reducing CO2 emission is a problem that, in our opinion, can
the questionnaire (Fig. 4). The participants could rate the
be considered to be representative for the whole domain of
importance of number of meals, of employees working in the
problems considering sustainable development. Based on these
kitchen, of meal categories offered, location of the canteen,
two reasons, we argue that our results are generalizable at least
customers’ preferences for comparison (QQ-4), on a semantic
to some extent to positive enabling effects about sustainability
differential scale of four with the two opposite criteria: “Plays
in general for systems in the service domain.
no role at all” and “Plays a significant role”. The majority of
participants from B&I canteens selected the number of meals, Construct Validity describes our ability to measure what
followed by customers’ preferences to enable fair comparison, we actually intend to measure. All participants had the
while the majority of participants form Edu canteens selected same tasks. However, we believe to have reduced mono-
customers’ preference as the top criterion, followed by the operation bias by including different work locations and work
number of meals. domains (Business & Industry, Education) into our study. We
have minimized mono-method bias by using methodological
C. Threats to Validity triangulation (contextual inquiry, interview, questionnaire). To
We discuss the threats to validity using the usual four avoid evaluation stress, we assured all participants that their
categories: internal, external, construct, and conclusion validity data were treated confidentially and evaluated for research
[34]. purposes only.
Internal Validity reflects the relationship between cause and Conclusion Validity is concerned with drawing correct
effect. The internal validity of our study is limited due to its very conclusions based on our observations. The first author was
nature: in an exploratory case study in a real world project, involved in designing the study and executing the elicitation
many potentially confounding factors cannot be controlled. techniques, which could potentially cause observer bias. For
A potential threat is the fact that the persons participating mitigating this threat, we used methodological and observer
in the contextual inquiry as well as the interviewees were triangulation and reviewed the structure and questions of all
selected by the Compass Group (Switzerland). We do not three elicitation methods with a group of experienced RE
consider this a major threat because we provided the company researchers. By conducting several pilot studies we strengthened
with our selection criteria and also did not find any evidence the quality of our wording. Further, we encouraged the
interviewees to ask for clarification if something was unclear. end-users have to perceive the representations of sustainability
Therefore, we do not consider measure reliability as a major goals to be meaningful as well as comparisons to be fair.
threat. As mentioned above, the interviewees emphasize the need
to integrate the information about CO2 emission values into
V. D ISCUSSION the DSS and want to be able to immediately access a list of
In this section, we discuss the results of our study with alternative ingredients with respect to their CO2 footprint.
respect to our two research questions and present some key The requirement of representing the CO2 values in a mean-
findings. ingful way is perceived as relevant by 94% of all interviewees.
Metaphorical representations such as number of trees needed
A. RQ1: What is specific about requirements concerning to compensate the CO2 emission or number of kilometers made
positive enabling effects? by a mean of transportation until the same amount of CO2 is
The results of our study show that the requirements do emitted are preferred over more abstract representations such as
change when sustainability comes into play. In our study, when energy equivalents. Interviewees also mentioned the importance
extending the existing DSS for meal planning with game- of concrete and visual representations e.g a concrete example
based mechanics for motivating environment-friendly choices for a trip from city A to city B, preferably indicated on a map.
to achieve a reduction of CO2 emissions, we mainly found three A proper segmentation of the displayed information is also
kinds of changes in requirements: new constraints, existing important. In our study, most participants favored the origin
constraints that become more important, and new functional of ingredients as segmentation criterion.
requirements. As discussed in our previous research [26], comparison
For example, the size of the kitchen is a new constraint that of individual achievements is an important game mechanic
the interviewees only perceive when the system is used in the for motivating sustainable actions. The results of our study
context of positive enabling effects for achieving a reduction underline the importance of considering the work context to
of CO2 emission. The smaller the kitchen, the less fresh food enable fair comparison when comparing the CO2 footprint of
can be processed, which means that the meal planners have to different canteens with each other. The relevance of customer
go for more frozen or pre-processed food. This, in turn, has preferences as a factor for enabling fair comparison again
an effect on the CO2 footprint of the meals. underlines the need for taking the needs of indirect stakeholders
Participants from both Edu and B&I canteens identified into account.
customer preferences and cost restrictions as the most important When analyzing the interview data with respect to the codes
constraints. The interviewees emphasized their fear about that originate from information provided spontaneously by the
increased cost constraints in the context of positive enabling interviewees (see Table III), we found that this information (and
effects due to higher prices for local and organic food. Further, hence, the corresponding sustainability requirements) can be
they are afraid of loosing customers by excluding off-season grouped into three categories: (a) integration of sustainability
products, which results in a smaller variety of food. That means information into the current system and work process ( i.e., by
that the existing constraints of customer preferences and cost properly extending the current software system instead of just
restrictions become more important in a sustainability context. adding a new, separate module), (b) meaningful representation
72 percent of all interviewees emphasize the need to integrate of the addressed sustainability aspect (CO2 emission in our
the information about CO2 emission values into the DSS. 33 study), and (c) fair comparison of the achievements of the
percent specifically want to be able to immediately access a list addressed users (meal planners in different canteens in our
of alternative ingredients with respect to their CO2 footprint. study). Although more research is necessary to establish
These are new functional requirements. the generalizability of this finding, our study provides some
Consequently, treating sustainability requirements as a sub- evidence that a classification of sustainability requirements with
category of specific quality requirements (cf. Section II-B2) respect to positive enabling effects into the classes integration,
turns out to be inadequate. Requirements concerning positive (meaningful) representation and (fair) comparison makes sense.
enabling effects can be functional requirements or constraints.
C. Key Findings
The growing importance of customer issues in the context
of positive enabling effects such as customer preferences and In summary, we draw five key findings about sustainability
cost restrictions also provides evidence that in the context of requirements regarding favorable enabling effects from the
sustainability requirements, there is a strong need for taking results of our study. In the context of positive enabling effects
into account also the indirect stakeholders of a system (i.e., regarding sustainable development:
those who are not end-users of the system, but are affected by • Requirements for a software system do change when
its use, e.g., customers). sustainable development is considered.
• We find both new requirements and existing requirements
B. RQ2: How can game-based mechanics motivate positive that become more important.
enabling effects when extending existing software systems? • Considering sustainability requirements to be a subset
The results show that it is important to integrate the game- of quality requirements is inadequate. We also found
based mechanics directly into the underlying system and that functional requirements and constraints.
• Game-based mechanics need to be integrated directly into [9] N. Amsel, Z. Ibrahim, A. Malik, and B. Tomlinson, “Toward sustainable
the underlying system. software engineering,” in 33rd International Conference on Software
Engineering. ACM, 2011, pp. 976–979.
• Meaningful representations of the sustainability aspect as [10] A. Raturi, B. Penzenstadler, B. Tomlinson, and D. Richardson, “Devel-
well as fair comparison are important. oping a sustainability non-functional requirements framework,” in 3rd
International Workshop on Green and Sustainable Software. ACM,
Further we have two findings where our study provides some 2014, pp. 1–8.
evidence, but further research is necessary. [11] “ISO 14040:2006. Environmental management – Life cycle assessment –
• Sustainability requirements can be classified into three
Principles and framework,” International Organization for Standardiza-
tion, 2006.
classes: integration, (meaningful) representation and (fair) [12] UN World Commission on Environment and Development, “Our common
comparison. future,” pp. 1–300, 1987.
• Indirect stakeholders, i.e., those affected by the use of
[13] M. Christen, “A theory of the good for a conception of sustainability,” in
16th Annual International Sustainable Development Research Conference,
the deployed system, should be involved when eliciting 2010.
sustainability requirements. [14] C. Becker, “Sustainability and longevity: Two sides of the same quality?”
in 3rd International Workshop on RE4SuSy, vol. 1216. CEUR-WS,
VI. C ONCLUSION AND F UTURE W ORK 2014.
[15] M. Glinz, “On Non-Functional Requirements,” in 15th IEEE International
We reported on the results of elicited sustainability require- Requirements Engineering Conference. IEEE, 2007, pp. 21–26.
ments regarding positive enabling effects. The study includes [16] J. Cabot, S. Easterbrook, J. Horkoff, L. Lessard, S. Liaskos, and J. Mazón,
“Integrating sustainability in decision-making processes: A modelling
78 data points from 60 participants working in 60 different strategy,” in 31st International Conference on Software Engineering.
canteens. Our main contribution is to reveal differences of IEEE, 2009, pp. 207–210.
such sustainability requirements compared to requirements [17] B. Penzenstadler and H. Femmer, “A generic model for sustainability
with process-and product-specific instances,” in Workshop on Green in/by
in traditional settings, as well as important requirements to Software Engineering (GIBSE 2013). ACM, 2013, pp. 3–8.
consider in a context of sustainability requirements for favorable [18] C. Bomfim, W. Nunes, L. Duboc, and M. Schots, “Modelling sustainabil-
enabling effects. Moreover, we found evidence that indirect ity in a procurement system: An experience report,” in 22nd International
Requirements Engineering Conference. IEEE, 2014, pp. 402–411.
stakeholders are important in this context and a possible [19] K. Roher and D. Richardson, “Sustainability requirement patterns,” in
classification of sustainability requirements. 3rd International Workshop on Requirements Patterns. IEEE, 2013, pp.
In our future work we will further exploit the findings of 8–11.
[20] X. Wang, “A comprehensive decision making model for the evaluation
this study. In particular, we plan to investigate the elicitation of green operations initiatives,” Technological Forecasting and Social
of sustainability requirements from indirect stakeholders who Change, vol. 95, pp. 191–207, 2015.
are outside of organizational reach. [21] I. Benbasat, D. K. Goldstein, and M. Mead, “The case research strategy
in studies of information systems,” MIS Quarterly, vol. 11, pp. 369–386,
ACKNOWLEDGMENT 1987.
[22] R. B. Svensson, T. Olsson, and B. Regnell, “An investigation of how
The authors thank Eaternity for the opportunity to con- quality requirements are specified in industrial practice,” Information
duct this study and all employees of the Compass Group and Software Technology, vol. 55, no. 7, pp. 1224–1236, 2013.
[23] S. Easterbrook, J. Singer, M.-A. Storey, and D. Damian, “Selecting
(Switzerland) who participated in the study. This research was empirical methods for software engineering research,” in Guide to
partially funded by Gebert Rüf Stiftung under grant number Advanced Empirical Software Engineering. Springer, 2008, pp. 285–311.
GRS-023/13. [24] B. A. Kitchenham and S. L. Pfleeger, “Personal opinion surveys,” in
Guide to Advanced Empirical Software Engineering. Springer, 2008,
R EFERENCES pp. 63–92.
[25] H. Beyer and K. Holtzblatt, Contextual design: defining customer-
[1] B. Penzenstadler, “Infusing green: Requirements engineering for green centered systems. Morgan Kaufmann, 1998.
in and through software systems,” in 3rd International Workshop on [26] M. Z. Huber and L. M. Hilty, “Gamification and sustainable consump-
RE4SuSy, vol. 1216. CEUR-WS, 2014, pp. 44–53. tion: overcoming the limitations of persuasive technologies,” in ICT
[2] M. Mahaux and C. Canon, “Integrating the complexity of sustainability Innovations for Sustainability. Springer, 2015, pp. 367–385.
in requirements engineering,” in First International Workshop on [27] C. B. Seaman, “Qualitative methods in empirical studies of software
Requirements for Sustainable Systems, 2012. engineering,” IEEE Transactions on Software Engineering, vol. 25, no. 4,
[3] M. Mahaux, P. Heymans, and G. Saval, “Discovering sustainability pp. 557–572, 1999.
requirements: An experience report,” in Requirements Engineering: [28] J. Singer, S. E. Sim, and T. C. Lethbridge, “Software engineering data
Foundation for Software Quality. Springer, 2011, pp. 19–33. collection for field studies,” in Guide to Advanced Empirical Software
[4] L. M. Hilty and B. Aebischer, “ICT for sustainability: An emerging Engineering. Springer, 2008, pp. 9–34.
research field,” in ICT Innovations for Sustainability. Springer, 2015, [29] C. E. Osgood, G. J. Suci, and P. H. Tannenbaum, The measurement of
pp. 3–36. meaning. University of Illinois Press, 1964.
[5] F. Berkhout and J. Hertin, “Impacts of information and communication [30] R. Likert, “A method of constructing an attitude scale,” in Scaling: A
technologies on environmental sustainability: Speculations and evidence,” Sourcebook for Behavioral Scientists, 1974, pp. 233–243.
Report to the OECD, vol. 21, 2001. [31] E. P. Cox III, “The optimal number of response alternatives for a scale:
[6] T. Johann and W. Maalej, “The social dimension of sustainability in A review,” Journal of Marketing Research, vol. 17, pp. 407–422, 1980.
requirements engineering,” in 2nd International Workshop on RE4SuSy, [32] C. H. Yu, “Abduction? deduction? induction? is there a logic of
vol. 995. CEUR-WS, 2013. exploratory data analysis?” in Annual Meeting of American Educational
[7] B. Penzenstadler, A. Raturi, D. Richardson, and B. Tomlinson, “Safety, Research Association. ERIC Document Reproduction Service No. ED
security, now sustainability: the nonfunctional requirement for the 21st 376 173, 1994.
century,” IEEE Software, vol. 31, no. 3, pp. 40–47, 2014. [33] C. B. Seaman, “Qualitative methods,” in Guide to Advanced Empirical
[8] K. Roher and D. Richardson, “A proposed recommender system for Software Engineering. Springer, 2008, pp. 35–62.
eliciting software sustainability requirements,” in 2nd International [34] C. Wohlin, P. Runeson, M. Höst, M. C. Ohlsson, B. Regnell, and
Workshop on User Evaluations for Software Engineering Researchers. A. Wesslén, Experimentation in Software Engineering. Springer
IEEE, 2013, pp. 16–19. (originally published by Kluwer in 2000), 2012.