Supporting User-Centered Requirements Elicitation from Lean

Personas: A UX Data Visualization-Based Approach

Maylon Macedo

, Gabriel V. Teixeira

, Ariel E. S. Campos

and Luciana Zaina

Federal University of S

ao Carlos (UFSCar), S

ao Paulo, Brazil

Keywords:

Lean Personas, User Experience, UX, UX Data, Requirement Elicitation, Information Visualization.

Abstract:

The interest in exploring User Experience (UX) data to support the requirement elicitation of interactive sys-

tems is not new. Although the literature discusses traditional methods to gather UX data (e.g., interviews and

surveys), personas have arisen as a more user-centered technique that presents information about users’ needs,

preferences, and characteristics about the application domain. Nonetheless, persona data is often represented

from qualitative data in textual format, which can insert difﬁculties in browsing and exploring a set of per-

sonas. This paper aims to present visualizations designed using Information Visualization (InfoVis) principles

that support the navigation and search-out UX data in a persona dataset. To aid in eliciting UX-related require-

ments, our visualizations are based on a funnel perspective that guides designers and developers to examine

data from an overview ﬁrst, then zoom and ﬁlter, to achieve the qualitative data in detail ﬁnally. We evaluated

with 20 participants concerning the interpretation of the visualizations. The results revealed that the partici-

pants, even those with little experience in requirement elicitation, could interpret and ﬁnd relevant UX data

from the visualizations.

1 INTRODUCTION

User experience (UX) is an important aspect of soft-

ware quality that affects software acceptance by users

(Kashﬁ et al., 2019). Although the literature provides

different deﬁnitions of UX, most of them state that

UX encompasses both the software functionalities

and its quality characteristics that are perceived by

end-users (Hassenzahl, 2018). Studies have discussed

the importance of including the discussion of UX

from the software conceptualization stages (i.e., re-

quirement elicitation, software ideation) (Kashﬁ et al.,

2019; Choma et al., 2022). Different requirements

elicitation techniques have emerged in the literature

(e.g., personas, user stories) (Faily and Lyle, 2013)

and the personas technique has been recognized as a

more focused user-centered approach to explore data

from end-users (Pinheiro et al., 2019). By creat-

ing ﬁctional characters that represent different groups

of end-users, the personas provide designers and de-

velopers with users’ data about their needs, experi-

ences, behaviors, and goals, hereinafter called UX

https://orcid.org/0000-0002-2686-1103

https://orcid.org/0000-0001-7589-3622

https://orcid.org/0000-0003-0857-4570

https://orcid.org/0000-0002-1736-544X

data (Billestrup et al., 2014; Pinheiro et al., 2019).

UX data contains information that allows mapping

users’ decisions and interactions with a product, en-

abling measurements and statements about the causes

and results of implementations in interactive systems

(Luther et al., 2020).

Personas often provide a rich amount of qualita-

tive UX data (Gothelf, 2012). Nonetheless, the tradi-

tional persona demands time and effort in user data

collection which becomes an expensive technique

(Billestrup et al., 2014). As an alternative, Gothelf

proposed the proto-persona also called lean persona,

an agile approach that allows designers/developers to

create concise personas based on their prior knowl-

edge about the target audience (Gothelf, 2012; Pin-

heiro et al., 2019).

On the one hand, designers and developers rec-

ognize the qualitative UX data from lean personas as

useful to elicit users’ needs and preferences in their

interaction (Isherwood, 2020). On the other hand,

they struggle to identify and extract meaningful in-

sights about UX from qualitative data in textual for-

mat which turns the elicitation of UX-related require-

ments into a complex task (Ye Lim Rhie and Yun,

2017). The extraction of requirements involving text

interpretation requires deﬁning meanings to provide

Macedo, M., Teixeira, G., Campos, A. and Zaina, L.

Supporting User-Centered Requirements Elicitation from Lean Personas: A UX Data Visualization-Based Approach.

DOI: 10.5220/0012567400003690

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 26th International Conference on Enterprise Information Systems (ICEIS 2024) - Volume 2, pages 461-472

ISBN: 978-989-758-692-7; ISSN: 2184-4992

461

different values to the raw data (Ye Lim Rhie and

Yun, 2017). Without deﬁning meanings, the analysis

depends on the judgment of each researcher, and the

result may vary in relation to knowledge and experi-

ence about UX (Ye Lim Rhie and Yun, 2017; Desai,

2019).

Visual data representations can guide designers

and developers during the requirement elicitation

(Kashﬁ et al., 2019) on the exploration of textual data

which reduces the effort to obtain information based

on patterns and emergent data (Desai, 2019). Con-

sidering the problem of browsing and exploring the

textual UX data provided by lean personas, this pa-

per presents three visualizations that deliver UX data

that are obtained from lean personas previously elab-

orated. These visualizations lead designers and devel-

opers to explore the lean persona data gradually rang-

ing from the data overview to data in detail taking into

account Shneiderman’s mantra (Shneiderman, 1996).

This paper is organized as follows. section 2

presents the fundamental background that underlies

the design and development of the visualizations and

the related work. We explain all activities made in

our systematic process to design the visualizations,

including the visualizations’ rationale and character-

istics, in section 3. In subsection 3.1 we explain the

dataset used as a baseline to design the visualiza-

tions. We discuss the study validity and present the

planning, execution, and analysis of the visualizations

evaluation in section 4. The analysis of the evalua-

tion results and our ﬁndings about the use of visual-

izations to get information by UX data are presented

in section 5. In section 6 we presented conclusions

and suggestions for future work.

2 BACKGROUND

2.1 Requirements Elicitation and

Personas

Software professionals have a common understand-

ing that a good process of requirements elicitation

promotes support for deﬁning requirements (Orme

and Panach, 2013). Software requirements have two

main classiﬁcations which are named functional and

non-functional. Quality requirements like UX re-

quirements ﬁt into the deﬁnition of non-functional

and involve understanding system tasks, user expecta-

tions, ease of use, and user satisfaction (Orme

no and

Panach, 2013).

As mentioned before (see Section 1), personas

have been used as an artifact to support UX-related

requirements elicitation due to the details about users’

needs and characteristics these artifacts provide to de-

velopers and designers (Faily and Lyle, 2013; Teix-

eira. and Zaina., 2022). However, the construction

of traditional personas requires efforts of time and

a great amount of data collection (Billestrup et al.,

2014). Lean personas is a technique that impresses

more agility to the personas creation process by con-

sidering the prior knowledge of the software practi-

tioners about the target audience (Gothelf, 2012).

Pinheiro et al. (2019) extended Gothelf’s lean

persona and proposed a four-quadrant template that

contained guideline questions to guide the lean per-

sona artifact construction. The four quadrants orga-

nize the lean persona information as follows: Q1 -

Demographic data; Q2 - Objectives and needs; Q3 -

Behaviors and preferences; and Q4 - Difﬁculties. The

four quadrants of the template and the guideline ques-

tions enhance the description of users’ information

while keeping the characteristic of being a lean ar-

tifact (Teixeira. and Zaina., 2022). To automatize the

creation of lean personas based on the four-quadrant

template, Teixeira. and Zaina. (2022) developed the

Lean Persona+ tool.

The results about the effectiveness and efﬁciency

of adopting personas artifacts during requirements

identiﬁcation were presented by Salminen et al.

(2020). The authors compared the use of personas

with requirements elicitation conducted using data

from an analytics tool (Salminen et al., 2020). The

results revealed that the developers spent more time

identifying the requirements that were caused by the

need to interpret a great amount of data from the an-

alytics tool. On the other hand, the personas artifacts

presented end-user data in a straight way which makes

the developers work faster (Salminen et al., 2020).

2.2 UX Data Visualization

Techniques such as lean persona produce qualitative

data about user characteristics, needs, preferences,

and experiences (Gothelf, 2012). This qualitative UX

data can aid designers and developers in setting prior-

ities for designing a better user experience during the

requirement elicitation process (Luther et al., 2020).

Besides, it can bring insights about why users need

some software features, what software features are

suitable to fulﬁll users’ needs and preferences, and

how users can interact with the features (Hassenzahl,

2018). However, the alignment on the meaning of

UX can insert difﬁculties in using the UX qualitative

data to requirement elicitation (Choma et al., 2022;

Ye Lim Rhie and Yun, 2017).

Different UX deﬁnitions are found in the litera-

ture. Hassenzahl provides a content-model deﬁnition

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

462

that categorizes UX into three levels based on inter-

action action (Hassenzahl, 2018). The why level en-

compasses the motivations and needs that lead an in-

dividual to use the product. The what level refers to

product features that fulﬁll users’ needs. The how

level is about user activity on the product to inter-

act with the functionalities. When the user reaches

their goal (i.e., why level) through successful use of

the concrete actions (i.e., how level) of product func-

tions (i.e., what level) a feeling of wellbeing is awak-

ened (Hassenzahl, 2018). Hassenzahl’s content model

provides a perspective of the UX information linked

to users’ interaction with the product (what and how

levels) and with the motivations and needs for having

these interactions (why levels). Each level of Hassen-

zahl’s content model focuses on information related

to UX which facilitates the levels of association with

UX qualitative data (Zaina et al., 2021).

Nonetheless, qualitative data provided from lean

personas are based on text data which turns the data

exploration for requirement elicitation harder (Slone,

2009). The visualization of data in plain text for-

mat does not assist users to infer patterns and ﬁnd

trends or outliers (Ware, 2012). The adoption of vi-

sual representations can assist the navigation of quali-

tative data guiding professionals to explore and inter-

pret UX data instead of relying on statements of indi-

viduals that analyzed the results (Slone, 2009; Ware,

2012).

The elaboration of visual representations can be

supported by Information Visualization (i.e., InfoVis)

techniques (Munzner, 2014; Ware, 2012). InfoVis of-

fers methods and techniques to enhance human per-

ception in the recognition and interpretation of data

(Ware, 2012). Concerning Human-Computer Interac-

tion (HCI), InfoVis aims to build visualizations that

are more adherent to the needs of a target audience

and prevent users from employing excessive efforts

to interpret data (Ware, 2012; Munzner, 2014).

2.3 Related Work

As far as we know, there is no previous work that pro-

poses visualizations to support the use of data from

personas. The work of personas related to require-

ment elicitation mostly discusses the potential of un-

covering user-centered data available from the per-

sonas (Pinheiro et al., 2019). Lean Persona+ tool de-

veloped by Teixeira. and Zaina. (2022) allows the

visualization of personas data; however, the visual-

ization is available from its template format. Consid-

ering that our approach is based on visual representa-

tions, we explored studies that provide contributions

to the visualizing UX data.

oro et al. (2014) proposed a bar chart view for

visualizing gaze-tracking data focused on testing dy-

namic web applications. An assessment of the us-

ability of the visualization showed that users want to

visualize the data in other formats, for instance heat

map graphs. To view website navigation data, four

visualizations (i.e., Arc Diagram, Word Tree, Sankey

Diagram, and Node-Link) were proposed by Buono

et al. (2020) to support novice evaluators in usabil-

ity testing. An evaluation by applying the System

Usability Scale questionnaire (SUS) with ﬁfteen par-

ticipants revealed that the Sankey Diagram was the

visualization most appreciated by the evaluators for

showing navigation data quickly and objectively. To

support the analysis of the results of usability and

UX assessments based on user tasks, Bernhaupt et al.

(2020) has created three visualizations. These visu-

alizations were built by applying the Shneiderman’s

mantra (1996) (i.e., overview ﬁrst, then zoom and ﬁl-

ter, details on-demand) with two levels of granularity:

overview and detailed.

Considering visualizations of qualitative data,

Bakiu and Guzman (2017) proposed an approach to

automatically detect software feature issues, usabil-

ity, and UX dimensions (e.g., satisfaction, comfort,

and motivation), and sentiment polarities (i.e., pos-

itive, negative, or neutral) according to user ratings

in apps reviews. Similar to Bernhaupt et al. (2020)

work, Shneiderman’s mantra (1996) was applied in

building the views, visualizing the results at two lev-

els of granularity: overview and detailed. The authors

do not report an assessment of the visualizations.

3 UX DATA VISUALIZATION

DESIGN

To design the visualizations, we followed two ap-

proaches: the Design Activity framework and the

Nested Model for Visualization Design proposed by

McKenna et al. (2014) and Munzner (2014), respec-

tively. Design Activity Framework (DAF) encom-

passes a process of four activities to support the de-

sign of visualizations: Understand - encourage de-

signers to adopt a user-centered design thinking about

requirements, goals, and opportunities of visualiza-

tions; Ideate - motivate the brainstorm of ideas and

thus the selection of one that best fulﬁlls the needs

discovered in the understand activity; Make - proto-

type the selected idea in a visualization (or a set of

visualizations), and Deploy - deploy an effective visu-

alization system which can be explored and evaluated

by end-users (McKenna et al., 2014).

Nested Model for Visualization Design (i.e., in

Supporting User-Centered Requirements Elicitation from Lean Personas: A UX Data Visualization-Based Approach

463

short, named Nested Model) is a well-known ap-

proach to support the evaluation and design of visual

representations (Munzner, 2014). It contains four lev-

els that cover (see Figure 1): (i) Domain Situation -

which guides designers in understanding the proposal

of the visualization; (ii) Data/task abstraction – which

helps designers ﬁnd which data and its formats are

used to do a task; (iii) Visual encoding/Interaction id-

iom – encourage designers to consider the visual rep-

resentations and approaches to show the data for users

(i.e., idiom of visualization); and (iv) Algorithm – de-

velop the visualization (Munzner, 2014). The idiom

of visualization can be perceived by chart models, in-

teraction features (e.g., zoom and pinch), and visual-

ization systems with the information presented in dif-

ferent degrees of granularity (Munzner, 2014).

We decided to adopt both approaches because

they present a complementary perspective (see Fig-

ure 1). Design Activity Framework (McKenna et al.,

2014) supporting us in applying the good practices

presented in the Nested Model for creating the visu-

alizations. We applied this complementary view for

each visualization we have proposed. Taking into ac-

count the complementary view of the approaches, we

proceeded to design the visualizations based on two

steps, i.e., preparation of lean personas dataset, and

visualizations elaborations. Each step is presented in

the next sections.

Figure 1: The complementary view of DAF and Nested

Model approaches - DAF in grey boxes and the Nested

Mode in colors - adapted from (McKenna et al., 2014) and

(Munzner, 2014), respectively.

Considering the complementary view of

McKenna et al. (2014) and Munzner (2014)

model (see Figure 1), we conducted the understand

step about the domain situation, i.e., how the visual-

izations could minimize the designers and developers

efforts to browse the qualitative data of lean personas.

Taking into account the great amount of qualitative

data, we decided to adopt a funnel perspective to

visualize the UX data as used in (Bakiu and Guzman,

2017; Bernhaupt et al., 2020). In both works, the

authors applied two levels of the Shneiderman’s

mantra (1996) (i.e., overview and detailed). In our

proposal, we planned to design three visualizations

to disclose the UX data gradually throughout the

three levels (i.e., overview, zoom and ﬁlter, and

details). After that, we searched for a lean personas

dataset that could be used in the visualization con-

struction. Besides, we conducted the understand of

the data available to then ideate how this data could

support the conception of the three visualizations

(see data/task abstraction level in Figure 1) as we

described in the next section.

3.1 Lean Personas Dataset Model

Our ﬁrst activity was to obtain the persona dataset.

We contacted the Lean Persona+ tool authors because

we considered they would have a dataset of virtual

data (Teixeira. and Zaina., 2022). We asked for ac-

cess to a dataset with the lean persona artifacts and

they provided a dataset containing 23 lean personas

speciﬁcations in the four quadrants template (see the

quadrants template explanation in section 2). Accord-

ing to the authors, this dataset was produced from the

studies conducted to evaluate the Lean Persona+ tool

(Teixeira. and Zaina., 2022). The dataset encom-

passed lean personas that represented the target users

of tourism mobile apps (e.g., TripAdvisor

). Each

lean persona is composed by quadrants, questions,

and answers which are the base elements of the lean

persona template (see an example of a persona’s data

in the Figure 2-III) (Pinheiro et al., 2019).

3.2 Labeling Lean Personas Dataset

Besides the feature of creating lean persona artifacts,

the Lean Persona+ tool provides a labeling tool. The

labeling tool allows the activity of marking chunks of

text available in the quadrant which are part of lean

personas descriptions. The labeling tool provides la-

bels based on the levels proposed by Hassenzahl (i.e.,

why, what, and how, see details in Section 2.2). With

these labels, we can conduct the labeling of chunks of

qualitative data to identify UX-relevant excerpts pre-

sented in the quadrants of lean personas. Different

labels could be assigned to the same chunks of text.

In a conversation with the authors tool, we have per-

mission to use the labeling tool to prepare the dataset

for being used in our visualization proposals. Three

experts with (i) knowledge of Hassenzahl’s content

https://www.tripadvisor.com.br/

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

464

model; and (ii) academic and practical experience

in Human-Computer Interaction and in Requirement

Engineering participated in the labeling activity. The

proﬁle of experts was (i) one professor with 15+ years

of experience; (ii) one PhD candidate with 5+ years of

experience in the software industry; and (iii) one ju-

nior researcher with 1+ years who also works in the

industry as a developer. After labeling 23 lean per-

sonas available in the dataset, we produced 165 labels

in total (i.e., 55 why, 77 what, 33 how).

3.3 Design the Visualizations

Taking into account the lean personas data labeled

with Hassenzahl’s content model levels, we moved to

ideate and make the visual encoding level with the

participation of the three UX experts that labeled the

lean persona dataset (see experts proﬁle in Section

3.2). These activities were performed for each visual-

ization that was associated with the three mantra lev-

els (i.e., overview, zoom and ﬁlter, and details).

In sequence, the make and deploy activities were

carried out for each visualization creation to complete

the algorithm level (see Figure 1). To provide a func-

tional version of the visualizations, we developed a

web application. It is worth noticing that although

we used a predeﬁned lean persona dataset, our visual-

izations proposal are extensible to other datasets that

follow the data structure presented in subsection 3.1.

Our three visualizations intend to orchestrate a

gradual navigation to the UX data of lean personas.

In this sense, the following description illustrates how

the visualizations are connected one each other. By

interacting with the overview visualization designers

and developers can ﬁrst ﬁlter the dataset by choosing

one of the labels (i.e. why, what, and how) to see the

personas that have information about that selected la-

bel. With the zoom and ﬁlter visualization they see

the quantitative data related to each persona-quadrant

relationship and can select one persona to be exam-

ined in detail. Details visualization shows the text

data and the labels assigned to each lean persona that

highlight the role of that chunk of text in providing

UX-relevant information. Such information can be-

come a UX-related requirement or trigger insights to

search for other requirements.

The following sections discuss each visualization

design. For each visualization, we will provide infor-

mation about our rationale for ideation that visualiza-

tion, the visual representation format selected and its

respective characteristics, and the chart and its details

that implemented that visualization.

3.3.1 Visualization to Overview the UX Data

Considering the organization of lean persona artifacts

data described in subsection 3.1, we noticed that there

is no direct way (i.e., without performing calcula-

tions) to visualize the quantitative data about the la-

bels (i.e., UX data). Primary navigation through lean

persona artifacts is performed without an overview by

selecting a particular persona in a tabular view to see

all data.

Rationale: An overview visualization avoids the

need to open each persona to observe the existence of

labels. This type of visualization (i.e. overview) helps

users see the scope of data and choose what informa-

tion they need to see in detail (Shneiderman, 1996;

Munzner, 2014). By looking at the labeling distribu-

tion, designers and developers can concentrate their

attention on the lean personas that may contain UX

data that they want to look for guided by the label that

abstracts the meaning of the data. Developers and de-

signers can take advantage of information such as: if

there are labels; the number of labels, and what kind

of existing labels.

Visual representation characteristics: In the

Sankey chart, the line can represent a relationship be-

tween categories while the thickness of the line rep-

resents quantitative data. The objective of this visu-

alization is to get an overview of all available data to

guide users on the data exploration. Our overview vi-

sualization (see Figure 2-I) can represent: the number

of labels that were found in each quadrant (see Fig-

ure 2-I-A)); the number of labels that exist of each la-

bel type (see Figure 2-I-B)); and the number of labels

of each type is present in each quadrant (see exam-

ple in Figure 2-I-C)). The visualization interaction is

Mouse hover in left axle to show the amount of all

existing labels in that quadrant; Mouse hover in right

axle to show the amount of all occurrences of that la-

bel; Mouse click in right axle to open the zoom and

ﬁlter visualization; Mouse hover on lines to show the

amount of a given label in a given quadrant.

3.3.2 Visualization to Zoom and Filter the Data

The overview visualization (see subsubsection 3.3.1)

provides the view of the representativeness of each

type of label assigned to all the lean personas avail-

able in the dataset. Taking into account the label se-

lected (i.e., why, what, and how) previously in the

overview visualization, designers and developers can

explore the quantitative data by choosing which lean

personas have the selected label.

Rationale: Considering our objective of eliciting

requirements from UX data present in lean persona,

zoom and ﬁlter visualization aids to reduce and fo-

Supporting User-Centered Requirements Elicitation from Lean Personas: A UX Data Visualization-Based Approach

465

See the number of labeling of “Why” type per quadrant.

Click on the persona picture to explore the details

II)

III)

Figure 2: I) Sankey visualization for overview, II) Heat Map showing the labeling distribution for zoom and ﬁlter, and III)

Card visualization to see the detail of qualitative data of the persona — High resolution PDF on Google Drive.

cus on a more detailed level of information (i.e. only

personas with the select label) achieving the goal of

presenting the data in a funnel perspective.

Visual representation characteristics: We chose

Heat Map as zoom and ﬁlter visualization because

it can show the quantitative relationship between two

categorical variables by highlighting the variation of

this relationship through colors (i.e., the color scale is

associated with the value scale). The objective of this

visualization is to show distribution of a pre-selected

label in quadrants-persona relationship. In our pro-

posal (see Figure 2-II) the title of the quadrants (see

Figure 2-II-A) that organize the lean personas infor-

mation was inserted on the Y-axis, and the identiﬁ-

cation of the personas on the X-axis (see Figure 2-

II-B). The quantitative information presented in the

Heat Map is the number of labels that exist for each

set of quadrants and lean personas (see Figure 2-II-

C). The visualization interaction is Mouse click in x-

axis to show all information about the selected per-

sona; Mouse hover in cells to show a textual explana-

tion following the pattern: “[PERSONA NAME] has

[LABEL AMOUNT] in [QUANDRANT NAME]”.

3.3.3 Visualization to See the Detail

After selecting the cell in zoom and ﬁlter, the visual-

ization of details delivers the UX qualitative data in

the lean persona essence by showing all the quadrants

descriptions and the excerpts highlighted with labels.

Rationale: The visualization of the lean personas

detailed provides the UX data more contextualized

due to being available all the qualitative data. Card vi-

sualization presents small clusters of information, i.e.,

lean personas quadrants, which allows the delivery of

content appropriate to the domain, i.e., description of

needs and preferences from lean personas.

Visual representation characteristics: In the detail

visualization, by default, only labels that match the

selected label type on the overview visualization will

be loaded with highlights, however exists the possibil-

ity of seeing all the labels to give software practition-

ers an overview of the labeling data. The objective of

this visualization is to access all information for a pre-

selected persona. In our proposal (see Figure 2-III)

the card header (see Figure 2-III-A)) shows the label

chosen in the zoom and ﬁlter visualization. Labels are

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

466

identiﬁed by colors plus a symbol. By looking at the

personas, software practitioners can clearly recognize

where the labels are placed and where they are not

(see Figure 2-III-B)). Finally, the visualization pro-

vides a function that enables the viewing of all the la-

bels that were assigned to that persona (see Figure 2-

III-C)). The visualization interaction is Mouse click

to enable/disable the display of all persona labels (i.e.,

by default only shows labels of the pre-selected type

in the Sankey chart).

4 UX DATA VISUALIZATION

EVALUATION

We conducted an experimental study with 20 partic-

ipants following the guidelines proposed by Lazar et

al. (2017) and Wohlin et al. (2012). We aimed to

evaluate the facility of interpreting the three visual-

izations. The objective of the visualizations was to

assist in extracting requirements based on the data

present in the personas, for this reason the interpre-

tation of the visualizations was analyzed. Subjects

are sampled by convenience. Our study was approved

by the ethical committee of the Federal University

of S

ao Carlos (UFSCar) under the process number

68524023.0.0000.5504. The next sections present the

study details.

4.1 Apparatus

We prepared four artifacts to support our study that is

described below.

An online questionnaire was elaborated to gather

participants’ consent to use the data collected (i.e.,

Informed Consent Form and their proﬁle data (i.e.,

demographic data, years of experience, or whether

they had no experience, position in the company, and

knowledge of information visualization and personas

techniques).

As personas descriptions are tightly associated

with the domain of the product, we prepared a sce-

nario description to explain the nature of the data that

participants will see. We got the same scenario de-

scription that the Lean Persona+ tool authors used to

produce the dataset. The scenario presents users’ in-

teraction with a tourism app: “A tourist usually uses

a mobile application (mobile, tablet, etc.) to plan and

guide their trips. The app displays places in a city or

region according to the interest and searches made by

users. These places can be hotels, monuments, muse-

ums, parks, and restaurants, among others. For each

location, the apps show details such as the location

name, photos, addresses, and feedback from other in-

dividuals that often are ranked by a score from 0 to

5. Moreover, these apps can present estimates of ex-

penses to spend in the places and also comments to

assist the user to decide and plan visits to places that

have good ratings”.

We constructed a questionnaire using Google

Forms to evaluate the interpretations of the visualiza-

tions. The questionnaire was composed of nine ques-

tions (see Table 1) and guided the participants in tasks

that encouraged the exploration of the three visualiza-

tions (i.e., overview, zoom and ﬁlter, and detail. At

the beginning of the questionnaire, an image of the

overview visualization was presented with numerical

labels to support the participants in identifying each

part of the visualization the questions were referring

to (see Figure 3).

Figure 3: Overview visualization image with numerical la-

bels.

4.2 Data Collection

The study was carried out with 20 participants, i.e., 14

undergraduate and 6 graduate students. The subjects

were invited in the context of the “Human-Computer

Interaction - (HCI)” course that was held in Com-

puter Science at the UFSCar. The participants ac-

cepted to take part in the study voluntarily and no type

of reward was involved. Although the participants

had different levels of experience in relation to UX

and data visualization, they mostly were profession-

als with experience in software development in prac-

tice. The study was conducted by two researchers, a

senior researcher with 15+ years of experience and a

researcher with 5+ years of experience at a laboratory

at UFSCar, and lasted 3 hours.

We initially presented the concepts about data

visualization (based on (Munzner, 2014)), levels of

UX (based on (Hassenzahl, 2018)), and lean persona

(based on (Pinheiro et al., 2019)) to level off the par-

ticipants’ knowledge. After that, we asked the par-

ticipants to navigate into the visualizations to explore

their features and understand the purpose of the per-

sonas’ quadrants. During this step, the participants

could clarify doubts about the concepts and the tool.

Supporting User-Centered Requirements Elicitation from Lean Personas: A UX Data Visualization-Based Approach

467

Table 1: Questionnaire to conduct users during the visualizations evaluation - Question TQ2 is a cluster of three questions.

ID Visualization objective Question Answer type Evaluation purpose

TQ1 Overview Is it possible to clearly visualize the dis-

tribution of the quantities of each label

(i.e., what, why, and how) in relation to the

quadrants?

We used a four-point Likert scale

as answer options ranging from

strongly disagree to totally agree

Verify whether that visualization informs the

participants of the labels with the highest and

lowest amounts in relation to the quadrants

without the need for absolute numerical veriﬁ-

cation

TQ2 Overview If you want to know more information

about UX data consider: [(a) application

functionality and their relationship with

the persona, (b) motivations and needs of

the persona, (c) concrete details of the

user’s interaction with the application] you

should click on:

The answers present the numbers

that are available in the overview

image in the questionnaire (see Fig-

ure 3). The participant should pro-

vide three answers, one for each

item (i.e., items a), b), and c))

Aim to verify whether the participants identiﬁed

the visualization element that provides access

to information from each label (i.e., why, what,

and how). The labels have semantic meaning

that represents the functionalities, user informa-

tion, and forms of interaction

TQ3 Overview Which quadrant has the fewest number of

labels?

Titles that identify the quadrant Identify particular information about the distri-

bution of labels presented in the visualization

TQ4 Overview Which label appears four times in the “Q4

- Difﬁculties” quadrant?

Name that identiﬁes the label Evaluate the use of interactive functions to iden-

tify speciﬁc information about the number of la-

bels in a speciﬁc quadrant

TQ5 Zoom and ﬁlter Select the incorrect alternative(s): a) The

persona Helena presents information about

the needs and motivations for using the

tourism application in all quadrants; b) The

persona Jorge presents information from

the tourists’ feedback feature; c) Quadrant

“Q2 - Needs and Goals” quadrant is suit-

able to know the types of concrete actions

that can be performed during the use of

the tourism application; d) Quadrant “Q3

- Behaviors and Preferences” is suitable to

know the types of concrete actions that ex-

ist during the use of the tourism application

Multiple-choice alternatives with

the statements as options

Verify whether by interacting with the zoom and

ﬁlter and details visualizations the participants

could evaluate the information in the alterna-

tives; see the participants interpretation on the

visual elements.

TQ6 Zoom and ﬁlter and details Which persona(s) provided more labels

that can give insights about features to the

tourism app? Justify your answer

Open-question answer from which

the participants write up the per-

sonas’ name and the motivation for

choosing a persona

Verify whether by exploring the zoom and ﬁlter

and details visualizations the participants were

able to identify which personas are suitable for

getting insights of the application features

TQ7 Overview, zoom and ﬁlter,

and details

What is the persona name that displays the

information “When the post text ﬁeld has

limited characters? In which quadrant is

this excerpt found? What label was used to

highlight this excerpt?”

Open-question answer to write up

the persona name, quadrant where

the information is placed, and type

of label assigned to the excerpt

Verify whether by navigating in all visualiza-

tions (i.e., overview, zoom and ﬁlter, and de-

tails) the participants could ﬁnd speciﬁc infor-

mation of a persona

We then proceeded by presenting the study’s objec-

tive and after that, we asked the participants to read

and answer the Informed Consent Form that state the

agreement on the use of the data and images for aca-

demic ends. After that, the participants ﬁlled out the

proﬁle questionnaire.

In the next step, all participants received the links

to: (i) the description of the scenario about the

tourism app; (ii) the online questionnaire that guided

the participants’ interaction with the three visualiza-

tions and evaluates the interpretation of the visualiza-

tions (see details in subsection 4.1); and (iii) the three

visualizations.

4.3 Data Analysis and Study Validity

We adopted a descriptive approach to examine the

data collected. To analyze the interpretation of the vi-

sualizations, we ﬁrst split the participants’ responses

by the perspectives of overview, zoom and ﬁlter, and

details considering the arrangement of the question

presented in Table 1.

We considered the conclusion, construction, inter-

nal and external threats to discuss our study validity

according to Wohlin et al. (2012) recommendations.

To deal with the conclusion threats, three researchers

adopted the same UX levels deﬁnition to label the

data in the personas dataset which avoided the pos-

sible bias that could be introduced by different inter-

pretations of UX data meaning.

To mitigate construct problems, we provided a set

of artifacts (i.e., a scenario of the tourism apps and the

questionnaire) to support the study conduction. All

the participants interacted with the same visualiza-

tions developed with the same dataset, which prevents

different results from being due to the types of data

or visualizations. We also delivered a presentation of

the relevant concepts to the study (i.e., visualization,

levels of UX, and lean persona). A hands-on exercise

ensured the participants had some experience with the

visualizations. These two activities allowed us to mit-

igate the impact of little knowledge of the participants

about the study concepts. Moreover, these activities

beneﬁted participants as they could clarify any doubts

about the concepts.

The internal threat related to participants’ fatigue

during the study. We proposed a lightweight set of

tasks guided by questions that stimulated the partic-

ipants to explore the visualizations. Besides, the ex-

ecution of tasks on the visualization lasted 1 hour at

most. Participation in the study was voluntary without

any ﬁnancial compensation. The participants were in-

vited to freely and optionally participate in the study,

and a few chose not to participate. It is worth men-

tioning that there was no reward to the participants

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

468

and the students who did not take part were not af-

fected in any way. The participants’ motivation was

due to the fact that the topics covered in the study

were popular and of interest to them. In the case

of an external threat, it commonly refers to the fact

that the study includes software developers who are

still students. However, Salman et al. provide ev-

idence of few performance differences between stu-

dents and professionals when performing new activ-

ities (Salman et al., 2015). In the case of this study,

the activity was new for all participants as none of

them had experience exploring UX data using visual-

izations.

5 RESULTS AND DISCUSSION

We present our results in three sections: (i) the partic-

ipants’ proﬁle, (ii) the discussion about the interpre-

tation of the three visualizations, and (iii) the compar-

ison of our results with the literature.

Participants’ Proﬁles. The proﬁle questionnaire re-

sults (see Table 2) reveal that 70% (14) of the par-

ticipants had a position in technology companies.

Considering the knowledge background of the par-

ticipants, 65% (13) have practical and/or theoretical

knowledge about InfoVis. On the other hand, more

than half of the participants (i.e., 60% (12)) had their

ﬁrst contact with the concepts of personas and lean

personas in the HCI course. We also noticed that

the more experienced professionals (i.e., > 1 year) re-

ported having little knowledge of personas technique

before attending the HCI course.

The Interpretation of the Three Visualizations. Ta-

ble 3 shows the participants’ answers to the questions

that guided the participants’ tasks (see the questions

in Table 1). The last line of the table provides the

correct answers of the questions. In next sections, we

will present the results for each visualization individ-

ually and the use of the three visualizations together,

and an overall discussion of the results.

In the Overview visualization the participants did

not have difﬁculties in identifying and interpreting

quantitative data about labels to select their ﬁrst step

for navigation. We saw that only two participants (one

participant for each question) did not answer correctly

questions TQ3 and TQ4 (see Table 1) about the dis-

tribution of UX labels, i.e., why, what and how. We

see that almost all participants agreed with the state-

ment on visualizing the distribution of the quantities,

i.e., 45% (9) of the participants “partially agree” and

55% (11) “strongly agree” with the statement (see

TQ1 in Table 1). We observed more than 70% suc-

cess rate in answering correctly questions on ﬁnding

Table 2: Participants marked with * were graduate students

— Participants proﬁle - Background on InfoVis: (1) I have

no knowledge about the area; (2) I have theoretical knowl-

edge about good practices; (3) I have practical knowledge

with visualizations, but I don’t know good practices; and

(4) I have practical knowledge and theoretical knowledge

about good practices to develop visualizations. Background

on persona and lean persona: (a) My ﬁrst contact with per-

sonas techniques was in the HCI course; (b) I already knew

the personas techniques but never used them in practice,

and (c) I already knew the personas technique and have also

used it in practice.

Identiﬁcation Background

ID Experience Position InfoVis Persona Lean persona

1* 1 to 3 years Developer 2 c b

2* > 8 years Developer 3 b a

3 < 1 year No job 1 a a

4 < 1 year No job 1 a a

5* > 8 years Developer 2 a a

6 1 to 3 years Analyst 3 a a

7 < 1 year No job 1 b b

8* > 8 years Developer 2 b b

9 < 1 year Developer 1 a a

10 1 to 3 years Developer 3 a a

11* < 1 year Support 3 b a

12* 6 to 8 years Developer 3 a b

13 < 1 year Analyst 2 a a

14 < 1 year No job 1 a a

15 < 1 year No job 1 a a

16 < 1 year No job 3 b a

17 1 to 3 years Developer 2 a b

18 < 1 year Developer 1 c b

19 1 to 3 years Researcher 4 b b

20 1 to 3 years Analyst 2 a a

UX labels throughout the entire process of navigat-

ing into the visualizations, i.e., overview, zoom and

ﬁlter, and details (see TQ2 in Table 1 and Table 3):

(i) 85% (17) of respondents correctly identiﬁed where

what labels were placed in the visualization; (ii) 65%

(13) of respondents correctly pinpointed where why

labels could be found; and (iii) 70% (14) of respon-

dents correctly identiﬁed the how labels. These re-

sults revealed the effectiveness of the Sankey chart in

(i) clearly showing the distribution of labels between

the quadrants; and (ii) allowing navigation and data

considering a speciﬁc purpose.

We motivated the participants to interact with the

three different visual representations (i.e., overview,

zoom and ﬁlter, and details) based on statements

about the personas’ data to use ﬁltering feature to

achieve the target information (see question TQ5 in

Table 1). We understood that TQ5 suffers the inﬂu-

ence of previous knowledge about the UX labels. We

observed that 70% (14) of the participants could se-

lect at least one right answer (i.e., options c and d in

Table 3). Our ﬁndings suggest that participants’ per-

formance on question TQ5 was inﬂuenced by their

prior familiarity with UX labels, with 70% (14) suc-

cessfully selecting at least one correct answer from

the available options.

In question TQ6, the participants should exam-

ine the personas to ﬁnd the ones that provide insights

about the features of the tourism app (see Table 1). To

Supporting User-Centered Requirements Elicitation from Lean Personas: A UX Data Visualization-Based Approach

469

Table 3: Participants’ answers to the tasks on the three visualizations - TQ2-A to C: answers were guided by Figure 3; TQ5:

see the alternatives in Table 1; TQ7:“N/A” was assigned to participants that did not report an answer.

ID TQ1 TQ2-

TQ2-

TQ3 TQ4 TQ5 TQ6 TQ7

1 Strongly agree (6) (4) (5) Q2: Needs and Goals Why b, d Claudia, Leandro (Jana

ına), (Q4), (How)

2 Strongly agree (6) (4) (5) Q2: Needs and Goals Why d Leandro (Jana

ına), (N/A), (How)

3 Strongly agree (6) (4) (5) Q2: Needs and Goals Why b, c Leandro, Clara, Cl

audia (Jana

ına), (Q4), (How)

4 Strongly agree (6) (4) (5) Q2: Needs and Goals Why b, c Clara, Leandro, Claudia (Jana

ına), (Q4), (How)

5 Strongly agree (6) (2) (1) Q2: Needs and Goals Why a, c Claudia, Lucas, Luiz, Jana

ına,

Lyn, Elena

(Jana

ına), (Q4), (N/A)

6 Strongly agree (6) (4) (5) Q2: Needs and Goals Why b, c Leandro, Claudia, Clara (Jana

ına), (Q4), (How)

7 Partially agree (4) (6) (5) Q2: Needs and Goals Why b, c, d Clara, Leandro, Claudia (Jana

ına), (N/A), (How)

8 Partially agree (5) (4) (6) Q2: Needs and Goals Why b, c F

abio (Leandro), (Q3), (N/A)

9 Partially agree (5) (4) (6) Q2: Needs and Goals Why c F

abio, Francisco (Jana

ına), (Q4), (N/A)

10 Partially agree (6) (1) (2) Q2: Needs and Goals Why a, b, c Clara, Leandro, Claudia (Jana

ına), (Q4), (N/A)

11 Strongly agree (6) (4) (5) Q2: Needs and Goals Why b, c, d Clara, Leandro, Claudia (Jana

ına), (N/A), (How)

12 Strongly agree (6) (4) (5) Q2: Needs and Goals How b, c, d Carlos, Clara, Leandro, Claudia,

Luiz, Francisco, Jorge, Adriana,

Helena, Rog

erio, Patr

ıcia

(Jana

ına), (Q4), (N/A)

13 Strongly agree (6) (4) (5) Q2: Needs and Goals Why c, d Clara, Leandro, Claudia (Jana

ına), (Q4), (How)

14 Strongly agree (6) (4) (5) Q4 - Difﬁculties Why d Jorge, Adriana (Jana

ına), (Q4), (How)

15 Partially agree (6) (2) (5) Q2: Needs and Goals Why a, c Leandro (N/A), (Q4), (Why)

16 Partially agree (6) (4) (5) Q2: Needs and Goals Why c, d Carla, Lauren, Francisco, Jorge (Jana

ına), (Q4), (How)

17 Strongly agree (6) (4) (5) Q2: Needs and Goals Why a, d Ana (Jana

ına), (Q4), (How)

18 Partially agree (6) (1) (2) Q2: Needs and Goals Why a, b Patr

ıcia (Jana

ına), (Q4), (N/A)

19 Partially agree (6) (5) (4) Q2: Needs and Goals Why a, c F

abio (Jana

ına), (Q4), (N/A)

20 Partially agree (6) (1) (5) Q2: Needs and Goals Why c, d Leandro (N/A), (Q4), (How)

Key answers Agree (6) (4) (5) Q2: Needs and Goals Why c, d Clara, Leandro, Claudia (Jana

ına), (Q4), (How)

accomplish the detail visualization to explore the lean

personas information, we motivated the participants

to interact with zoom and ﬁlter and detail visualiza-

tions more than once and brought an interpretative

approach to the task. To correctly answer this ques-

tion, the participants should pinpoint three personas

(key answers: Clara, Leandro, and Claudia). Our re-

sults showed that more than half of the participants

(i.e., 11 participants - 55%) answered the question

by mentioning the three personas or at least one of

them (Table 3). Our ﬁndings indicate that the funnel-

based approach helped participants (i.e., 11 partici-

pants - 55%) to successfully address the question by

mentioning at least one of the key answers.

To get a perspective of the interpretation with the

three visualizations connected one each other, we pro-

posed question TQ7 (see Table 1). In this question,

the participants should explore the different paths in

the visualizations and browse the overview and zoom

and ﬁlter visualizations to achieve the answers avail-

able from the detail visualization to thus select the

lean persona. The results showed (see Table 3) that

(i) 85% (17) of the participants selected the right lean

persona; (ii) 80% (16) selected the right quadrant;

and (iii) 60% (12) selected the right label. From this

result, we could conclude that the elements and the

interactions provided by the three visualizations sup-

ported the search for particular UX data.

The results showed that the participants did not

have difﬁculties in ﬁnding speciﬁc information about

UX data (see questions TQ1, TQ3, TQ4, and TQ7

in Table 1 and their results in Table 3). Our results

revealed that all participants got at least 50% of the

questions right. We understand that the knowledge

in the UX labels could inﬂuence the use of the visu-

alizations in particular to determine which label has

the desired UX data (e.g., questions TQ2-A, TQ2-B,

TQ2-C, TQ5, and TQ6 in Table 1).

We noticed the results of the questions TQ2-A and

TQ6 were connected. In the TQ6 question the partic-

ipants should inform which persona had the informa-

tion about features. In a lean way, in TQ2-A question

they should inform where in the visualization they

had to click to access information about features (i.e.

which label). Two of the three participants who chose

the wrong answer in TQ2-A also selected the wrong

option in TQ6. This result suggests that some par-

ticipants could struggle to understand the relationship

between labels and the embedded UX data of each la-

bel. To mitigate this issue, we conducted a warm-up

to leverage the participants’ knowledge about UX la-

bels; however, we concluded that we needed to ﬁnd

alternatives to present the UX label meanings in the

visualizations.

Comparison with the Related Work. We can see

three main differences in relation to the use of visual-

ization to explore personas data, the adoption of com-

plementary approaches for the conceptions of the vi-

sualizations, and the use of different chart formats to

view the data.

Firstly, we proposed three visualizations that to-

gether and connected aid developers and designers in

streamlining the search for useful UX data into lean

personas textual information. As far as we know there

is no other similar proposal that works to make eas-

ier the exploration of personal data although personas

are recognized as a proper technique for UX-related

requirement elicitation as we mentioned in (Pinheiro

et al., 2019).

We have the consciousness that there are many

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

470

works that adopt the Nested Model of Munzner

(2014) to support the elaboration of visualizations.

However, we used a different approach by putting har-

monically together the good practices proposed by the

Nested Model and the systematic process of visual-

ization construction, i.e., Design Activity Framework

(DAF) of McKenna et al. (2014) (see section 3). The

complementary perspective of the approaches sup-

ported us in following a sequence of steps (DAF), ob-

serving the details of the visualizations, and clearly

specifying the elements of each one (Nested Model).

While the approaches above guided our process

of visualization elaboration, Shneiderman’s Mantra

sustained the conception of the visualizations by in-

troducing ways to gradually conduct developers and

designers to the lean personas data. Although Bern-

haupt et al. (2020) and Bakiu and Guzman (2017)

have applied Shneiderman’s Mantra, they do not use

all three levels of visualization. The visualizations

proposed by Bakiu and Guzman (2017) do not apply

zoom and ﬁlter levels. Bernhaupt et al. (2020) work

presents a close perspective of ours by applying the

mantra; however, the authors focus only on the levels

of overview and detail.

Taking into account the chart formats, we see that

Buono et al. (2020) also used the Sankey chart and

that this chart was highly accepted in the evaluations

conducted by the authors. In the work, the authors

built the Sankey chart to show the number of visits

(i.e., quantitative data) for each page (i.e., categories)

that the users navigated on the website. Similar to

the authors, we adopted the Sankey chart to show the

relationship between quantitative data (i.e., numbers

of labels) and categorical data (i.e. why, what, and

how). At the design granularity of zoom and ﬁlter,

we adopted the Heat Map to show a three-leg rela-

tionship. We associated a particular UX data with the

lean personas and the quadrants. In this visualiza-

tion, developers and designers can identify, through

the color pattern, where UX data is concentrated (i.e.,

why, what, and how). M

oro et al. (2014) work men-

tions the Heat Map as a visualization that is desired

by users but they do not implement it.

6 FINAL CONSIDERATIONS

In this paper, we presented a new visual approach to

UX requirements gathering from lean personas. The

three visualizations were developed to assist develop-

ers and designers in searching and exploring UX data

available in the textual descriptions of lean personas.

The visualizations are intended to guide the elicitation

of UX-related requirements following Shneiderman’s

mantra (1996), i.e. provide the overview of data ﬁrst,

make available zoom and ﬁlter of the data, and ob-

serve the details. In this way, we proposed a funnel-

based approach for data exploration. By connecting

the three visualizations, we provided ways of explor-

ing the data gradually throughout the visualizations.

We labeled the lean personas descriptions by using

Hassenzahl’s UX levels. It aimed to give meaning to

the UX data available since qualitative data can intro-

duce subjectiveness to the requirement elicitation. We

evaluated the three visualizations with 20 participants

with regard to the interpretation of the visualizations.

Overall, our ﬁndings revealed that the three visu-

alizations provide easy ways of identifying and ﬁnd-

ing UX data available in the lean personas. Nonethe-

less, the interpretation evaluation revealed the need to

add some kind of help that aids the better understand-

ing of the UX level which was used to provide mean-

ing to the data. The participants’ feedback showed us

that improvements on the zoom and ﬁlter features are

requested to give more ﬂexibility to users’ search and

ﬁnd information. We also presented a method contri-

bution by using Nested Model (Munzner, 2014) and

Design Activity Framework (McKenna et al., 2014)

in a complementary perspective for building the visu-

alizations. This same combination of approaches can

be replicated in other designs of the visualizations.

In future work, we intend to improve our visual-

izations to add the features suggested by the evalua-

tion participants. We also intend to conduct new stud-

ies in real scenarios (i.e., software projects with de-

signers, requirement analysts, and developers) with

different samples of lean personas. Regarding the

evaluation, we intend to analyze the usability and ef-

ﬁciency of the visualizations to facilitate the organi-

zation and speed up the understanding of the UX in-

formation collected through lean personas.

ACKNOWLEDGEMENTS

This study was ﬁnanced in part by the Coordenac¸

de Aperfeic¸oamento de Pessoal de N

ıvel Superior

– Brasil (CAPES) – Finance Code 001 and grant

309497/2022-1 and 147915/2022-8, Consellho Na-

cional de Desenvolvimento Cient

ıﬁco e Tecnol

ogico

(CNPq - Brazil).

REFERENCES

Bakiu, E. and Guzman, E. (2017). Which feature is un-

usable? detecting usability and user experience is-

sues from user reviews. In 2017 IEEE 25th Inter-

Supporting User-Centered Requirements Elicitation from Lean Personas: A UX Data Visualization-Based Approach

471

national Requirements Engineering Conference Work-

shops (REW), pages 182–187.

Bernhaupt, R., Martinie, C., Palanque, P., and Wallner, G.

(2020). A generic visualization approach supporting

task-based evaluation of usability and user experience.

In Human-Centered Software Engineering, pages 24–

44. Springer.

Billestrup, J., Stage, J., Nielsen, L., and Hansen, K. S.

(2014). Persona usage in software development: Ad-

vantages and obstacles. In ACHI 2014.

Buono, P., Caivano, D., Costabile, M. F., Desolda, G., and

Lanzilotti, R. (2020). Towards the detection of ux

smells: The support of visualizations. IEEE Access,

8:6901–6914.

Choma, J., Guerra, E. M., Alvaro, A., Pereira, R., and

Zaina, L. A. M. (2022). Inﬂuences of UX factors in

the agile UX context of software startups. Inf. Softw.

Technol., 152:107041.

Desai, R. (2019). 4 ways to implement data for a better user

experience design.

Faily, S. and Lyle, J. (2013). Guidelines for integrating

personas into software engineering tools. In Proceed-

ings of the 5th ACM SIGCHI Symposium on Engineer-

ing Interactive Computing Systems, EICS ’13, page

69–74, New York, NY, USA. ACM.

Gothelf, J. (2012). Using proto-personas for executive

alignment. UX Magazine, 1.

Hassenzahl, M. (2018). The Thing and I (Summer of ’17

Remix), pages 17–31. Springer International Publish-

ing, Cham.

Isherwood, M. (2020). Guide to audience data for ux de-

sign.

Kashﬁ, P., Feldt, R., and Nilsson, A. (2019). Integrating

ux principles and practices into software development

organizations: A case study of inﬂuencing events. J.

Syst. Softw., 154(C):37–58.

Lazar, J., Feng, J. H., and Hochheiser, H. (2017). Research

Methods in Human Computer Interaction. Morgan

Kaufmann, Boston, NY, USA.

Luther, L., Tiberius, V., and Brem, A. (2020). User ex-

perience (ux) in business, management, and psychol-

ogy: A bibliometric mapping of the current state of

research. Multimodal Technologies and Interaction,

4(2):18.

McKenna, S., Mazur, D., Agutter, J., and Meyer, M. (2014).

Design activity framework for visualization design.

IEEE Transactions on Visualization and Computer

Graphics, 20(12):2191–2200.

oro, R., Dar

az, J., and Bielikov

a, M. (2014). Visual-

ization of gaze tracking data for ux testing on the

web. In HT (Doctoral Consortium/Late-breaking Re-

sults/Workshops).

Munzner, T. (2014). Visualization analysis and design.

A.K. Peters visualization series. A K Peters.

Orme

no, Y. I. and Panach, J. I. (2013). Mapping study about

usability requirements elicitation. In Salinesi, C., Nor-

rie, M. C., and Pastor,

O., editors, Advanced Infor-

mation Systems Engineering, pages 672–687, Berlin,

Heidelberg. Springer Berlin Heidelberg.

Pinheiro, E. G., Lopes, L. A., Conte, T. U., and Zaina, L.

A. M. (2019). On the contributions of non-technical

stakeholders to describing ux requirements by apply-

ing proto-persona. Journal of Software Engineering

Research and Development, 7:8:1 – 8:19.

Salman, I., Misirli, A. T., and Juristo, N. (2015). Are

students representatives of professionals in software

engineering experiments? In 2015 IEEE/ACM 37th

IEEE International Conference on Software Engineer-

ing, volume 1, pages 666–676.

Salminen, J., Jung, S.-G., Chowdhury, S., Seng

un, S., and

Jansen, B. J. (2020). Personas and analytics: a com-

parative user study of efﬁciency and effectiveness for

a user identiﬁcation task. In Proceedings of the 2020

CHI Conference on Human Factors in Computing Sys-

tems, pages 1–13.

Shneiderman, B. (1996). The eyes have it: a task by data

type taxonomy for information visualizations. In Pro-

ceedings 1996 IEEE Symposium on Visual Languages,

pages 336–343.

Slone, D. J. (2009). Visualizing qualitative information. The

Qualitative Report, 14:489–497.

Teixeira., G. V. and Zaina., L. A. M. (2022). Using lean per-

sonas to the description of ux-related requirements: A

study with software startup professionals. In Proceed-

ings of the 24th International Conference on Enter-

prise Information Systems, pages 211–222. INSTICC,

SciTePress.

Ware, C. (2012). Information Visualization: perception for

design. Morgan Kaufmann Publishers Inc., San Fran-

cisco, CA, USA, 3 edition.

Wohlin, C., Runeson, P., H

ost, M., Ohlsson, M. C., Reg-

nell, B., and Wessl

en, A. (2012). Experimentation in

software engineering. Springer Science & Business

Media.

Ye Lim Rhie, J. H. L. and Yun, M. H. (2017). Eval-

uating representativeness of qualitative text data in

identifying ux issues. International Journal of Hu-

man–Computer Interaction, 33(11):868–881.

Zaina, L. A., Sharp, H., and Barroca, L. (2021). Ux in-

formation in the daily work of an agile team: A dis-

tributed cognition analysis. International Journal of

Human-Computer Studies, 147:102574.

ICEIS 2024 - 26th International Conference on Enterprise Information Systems

472