The influence of the participation in Trello on teamwork quality, team performance and team members’ success of bachelor students completing a project in agile teams

1.9 Purpose and hypotheses

The purpose of this research is to study the influence of participation in Trello during a project on teamwork quality, team performance, and team members' success of bachelor students completing a project in agile teams. In order to reach it, the impact of the students' activity in Trello on each of the concepts and the variables they are composed of has to be examined. Additionally, the mutual influence of the concepts is to be studied as well. Those particular concepts were selected for the study as their connections with the agile team setting, and one another could be validated.

Teamwork quality is a complex concept that helps to evaluate the interactions within a team and consists of six other variables. It has been tied especially closely with the team members' success and team performance of both traditional and agile projects. Multiple researchers have studied teamwork quality in traditional projects' settings (Hoegl & Gemuenden, 2001). A fewer number of studies have investigated teamwork quality in regards to the projects of agile teams. It was found that in the project environment where the agile principles are applied teamwork quality is rather high (Freire, Perkusich, Saraiva, Almeida, & Perkusich, 2018; Lindsjшrn et al., 2016). However, it has been noted that the levels of the teamwork quality vary both in agile and traditional teams if the spheres, types or sizes of the projects studied are different. In the case of some projects, teamwork quality is higher in agile teams, while for the other projects, the difference between agile and traditional teams is insignificant (Lindsjшrn et al., 2016). In this research paper, Trello, a software application which creates an agile environment during the project, is studied during the educational project in the sphere of human resources management. Therefore, the following hypothesis is formulated:

H1: Participation in Trello positively influences teamwork quality.

Another concept, team performance, is composed of two variables, effectiveness and efficiency, and reflects how well the team meets the required by the project initiator objectives (time, costs, quality, etc.). As stated above, team performance is positively associated with teamwork quality (Hoegl & Gemuenden, 2001). Team performance has already been researched in many kinds of spheres (psychology, software development, management science, etc.) and both traditional and agile teams (Dingsшyr & Lindsjшrn, 2013). According to the paper by Schmidt, Kude, Heinzl, and Mithas (2014), the agile environment has a positive effect on the team performance of the software development teams. However, it is unclear whether the effect would remain in case of a project in a different sphere. Thus, there is the hypothesis:

H2: Participation in Trello positively influences team performance.

Team members' success is about their personal (e.g. increase in motivation and willingness to be a part of a team) and professional (e.g. increase in skills and knowledge) growths due to the participation in a project (Hoegl & Gemuenden, 2001). It is built of satisfaction and learning variables. What is more, team members' success is positively influenced by teamwork quality when a project is about software development (Lindsjшrn et al., 2016; Schmidt et al., 2014). Nevertheless, if a project is an educational one and in the sphere other than software development, the levels of team members' success may vary. Hence, the hypothesis below:

H3: Participation in Trello positively influences team members' success.

As has already been stated above, teamwork quality has a significant positive effect on the team members' success and team performance. For various projects in both traditional and agile teams in the software development field, there is a notable connection between these three concepts (Hoegl & Gemuenden, 2001; Lindsjшrn et al., 2016). However, it is possible that in the case of an agile project with different specifications (duration, sphere, team size, etc.), the connections may be not as secure. Thus, there is a hypothesis:

H4: Teamwork quality positively influences (a) team members' success and (b) team performance.



2. Methodology

2.1 Research topic and research questions

In this paper, the influence of the participation in Trello on teamwork quality, team performance, and team members' success of bachelor students completing a project in agile teams would be examined.

There are four research questions:

1. How does participation in Trello influence teamwork quality?

2. How does participation in Trello influence team performance?

3. How does participation in Trello influence team members' success?

4. How does teamwork quality influence team members' success and team performance?

2.2 The research model

The following graph (Figure 1) depicts the dependency relationships between the defined variables. There is the research model:

Figure 1. The research model



2.3 Hypotheses

The result of the analysis enabled the researchers to test the following hypotheses that emerged from the literature review:

1. Participation in Trello positively influences teamwork quality.

2. Participation in Trello positively influences team performance.

3. Participation in Trello positively influences team members' success.

4. Teamwork quality positively influences (a) team members' success and (b) team performance.

2.4 Setting and project description

The study was conducted in one university in Russia among first-year bachelor students of Business and Management taking a mandatory Human Resource Management course. The students were tasked to complete a consulting project after which they were required to complete two online questionnaires. The project was titled `HR Solutions', and its goal was for students to gain experience of working in agile teams using Trello to explore one of the fundamental human resources (HR) process, recruitment and selection (R&S) or training and development (T&D). The project lasted 3,5 weeks, from the week of April 6 to the week of April 27, 2020.

Students were randomly allocated into 50 teams of five to six people, within which the members were asked to select one team leader by collective voting. Students were assigned to teams by facilitators to simulate a professional work environment, where teams are formed based on professional roles and scope of tasks they are able to perform, not personal relationships of employees (Hoda et al., 2013). Team leaders were responsible for providing leadership for the team, facilitating the project flow and the usage of Trello. They also had to contact project assistant, teaching assistant or professor in case of any questions related to the project tasks. Additionally, they facilitated the completion of the questionnaires and performed any additional responsibilities assigned by the team. Team members, in turn, were obliged to contribute to the project by sharing knowledge, expertise, ideas, and information. They had to perform the assigned tasks while exhibiting professionalism and reasonable diligence and supporting and respecting fellow team members. Besides, they were expected to complete the questionnaires at the end of the project.

At the next stage of the project, the teams had explored the project assignment (see Appendix 1). They assessed the initial content of their Trello boards, which was created by the facilitators of the project as a visual example for the students who had not had previous experience of using the application. Each board also provided a handout with an explanation of the use of Trello in agile teams (see Appendix 2) and a presentation with the justification of the usage of the agile approach and Trello (see Appendix 3). After that, the students proceeded to select a company with an HR department or HR generalist in any industry and geographical location. Next, the teams chose one of the HR processes of the company to study: recruitment and selection or training and development.

Students were expected to conduct a thorough study of materials that were made publically available about the chosen company to learn more about the company, its employees and HR processes. Using the information collected, students prepared the interview questions about the company's HR department, its activities, and the potential problems in R&S or T&D to address to an employee working in the HR department of the chosen company.

After that, students were encouraged to practice their networking skills by identifying potential company insiders in the sphere of HR and find a way to contact them to conduct an interview based on the questionnaire designed during the previous stage of the project. The number of employees in the chosen company that students could question and the number of interactions with them was not limited to allow participants to gather a sufficient amount of valuable information to be used later in the project. The ultimate goal of the interactions was to collect information about the company, its HR processes and problems the employees or the company faced that could have had a connection to these processes.

Next, after identifying problems in the company's R&S or T&D processes, students were expected to eliminate several of them and, finally, choose only one for further analysis. Consequently, they were to explore the academic literature (e.g. journal articles, books, reports), business cases, official legal documentation of other companies, as well as other sources of reliable information to research the chosen problem. The students were expected to describe the issue in terms used in Human Resource Management field, identify its typical causes and effects, present the existing solutions to the problem, and identify, describe and justify the best solution or a set of solutions for this particular company.

The project was designed to have two outcomes: a written report to the company and a presentation to the class. The former was to be sent to the interviewee or interviewees from the chosen company. The report could have been prepared in any language convenient for them. This report was not checked or graded by the professor and served as a tool to help students summarize their work on the project. It also provided them with a chance to practice writing a business report for a functioning company. The report should have been two-page long and needed to include an introduction (description of the chosen problem based on literature), methodology (summary of data sources and interviews), problem analysis (the problem's common causes, effects and solutions), solutions (description of the best possible solutions and their justifications), and references to literature used in the process of solving the case. The final presentation that students were required to show to the class and the professor, responsible for the group, was planned to be 7-10 minutes long. It needed to include a description of the company, HR department or generalist description, overview of the company's problems in the sphere of training and development or selection and recruitment, methodology, the chosen problem description, its causes and possible effects for the company, existing solutions description based on literature, the most fitting solutions descriptions and justifications selected by the team, a screenshot of the email with the attached final report sent to the company, and references.

The evaluation criteria for the presentation could be seen in Appendix 4. The maximum sum of preliminary points was equal to 20. However, two additional points were given to the students, whose team members filled in the questionnaires. After the presentation, the points received by the team were summed up and divided by two to calculate the final mark. The final mark was to be rounded. The curve was arithmetical (3.49 equals a 3; 3,501 equals a 4). Team leader received 1 extra point to the final mark for the project if Trello was used for all steps in the project during the 3,5 weeks.

The completion of the full pilot of the project before the initial start of the Human Resource Management course was considered impossible. Firstly, there was a limited amount of time before the beginning of the course. Secondly, the project itself was rather complicated. Finally, a considerable number of steps needed to be taken in order to complete the 3,5-week-long project. Consequently, it was decided that instead of launching a pilot, the final project description would be evaluated by several professors from Management faculty and bachelor students in their last year of study to reveal any inconsistencies the project plan could contain. In total, the project description was assessed by three professors and five fourth-year bachelor students. After their assessment was finished, all the necessary changes were made to the project. Thus, the project's quality and integrity of both its practical and theoretical value for the students was improved.

2.5 Sample

The overall number of students who were supposed to be observed in this research was 272. However, due to personal reasons, some of the students chose not to participate in the project. As a result, the end number of participants in the survey used for the study totaled 251.

The age range of the respondents was between 17 and 27 and could be seen below in Figure 2. The vast majority of the respondents were 18 (163 people) and 19 (55 people) years old. Considerably lesser number of people were 20 (16 students) and 17 (8 people) years of age. Only 5, 3 and 1 respondents were 21, 22 and 27 years old, respectively. The average age of the respondents was 18,5 years old.

Figure 2. The age distribution of the respondents

Figure 3 below shows the distribution of the respondents' gender. The majority of the participants (n=178; 71%) were female, while 73 (29%) remaining were male. Each team was designed to choose a team leader among its members. Therefore, there is also a statistic of the gender distribution of team leaders. Females accounted for 70% of team leaders, while 30% of student teams had a male leader.

Figure 3. The gender distribution of the respondents



Figure 4 depicts the respondents' countries of origin. The majority of the students were from the Russian Federation (n=206; 82%). The rest of the respondents were from 16 different countries all over the world. These countries included Kazakhstan (16), Uzbekistan (8), Moldova (3), Latvia (3), Sri Lanka (2), Estonia (2), Armenia (2), Azerbaijan (1), Belarus (1), China (1), Ecuador (1), France (1), Italy (1), Taiwan (1), Tajikistan (1), and Vietnam (1).

Figure 4. The countries of origin of the respondents

Figure 5 shows how many students had any previous experience with Trello before the project. The vast majority of the respondents (n=207; 82,5%) had not used or known Trello before the project was introduced.

Figure 5. The respondent's previous experience with Trello



2.6 Data collection

As for the research strategy of the study, the quantitative methods were used to obtain the necessary data. Data came from two sources: (1) participants' self-reports via surveys and (2) a record of participants actions in Trello (contribution markers). Participants' self-reports were collected via two surveys. The first one (Appendix 5) consisted of 4 questions and collected demographic data (age, gender, and country of origin), as well as familiarity and previous experience with Trello.

The second questionnaire (Appendix 6) was used solely to gather data necessary to answer the research questions. It was initially developed and proposed by Hoegl and Gemuenden (2001) and served as a starting point for its further adaptation to the context of the research and its exploitation as a data collecting tool. This questionnaire consisted of several parts that assessed (a) the teamwork quality, (b) the subjective perception of team members' success, and (c) team performance. Each of these sections was comprised of additional subscales, which characterized the particular aspects of the parent construct. As discussed by Hoegl and Gemuenden (2001) and Lindsjшrn et al. (2016) the teamwork quality is measured using the scales of communication, coordination, mutual support, effort, cohesion, and balance of member contribution. The team members' success was comprised of two sub-constructs, one of which defined the level of team members' satisfaction with the current teamwork, and the other identified the team members' learning outcomes. The last section of team performance regarded the success of the project in terms of effectiveness and efficiency of the work done. In the studies, Hoegl and Gemuenden (2001) and Lindsjшrn et al. (2016) measured the internal consistency of the questionnaires calculating the Cronbach's alpha. In the first case of Hoegl and Gemuenden (2001), the Cronbach's alpha coefficient for each measured scale was higher 0.72, while the Cronbach's alpha for TWQ construct as a whole was 0.91. In case of the research by Lindsjшrn et al. (2016), the only subscale with the same coefficient below 0.7 was “balance of member contribution” (0.58), while the Cronbach's alpha for the majority of the subscales exceeded 0.81. These results indicate a high reliability of the implemented data collection tool, according to Nunnally (1994).

The original questionnaire consisted of 61 statements measuring the specific characteristics of the mentioned constructs. For each statement, it was required to show the agreement on the 5-point Likert scale from 1, which meant “strongly disagree”, to 5, which meant “strongly agree”. In this research, the original questionnaire was adapted to the context of the study by making two changes. First, the part investigating team performance in terms of effectiveness and efficiency was excluded from the original version of the questionnaire and replaced with the mark provided by the professor according to the specified project evaluation criteria. Thus, the structure and definitions of the remaining two constructs can be observed in Appendix 7. Second, the part with general questions about the student's team number and whether a participant of the survey was a team leader or not was included. Altogether, the second survey included 46 statements, among which 38 statements were dedicated to the teamwork quality concept while 8 statements measured team members' success, and two student-related questions. The Cronbach's alpha was estimated for every subscale to measure the internal consistency of the resulting questionnaire.

Overall, the data collection process via questionnaires took place via Google Forms. The corresponding links to the surveys were allocated on every Trello board by the facilitators so that every student could easily access the questionnaires. The students had a certain period when they could participate in the questionnaires and receive additional points if all of the team members completed the surveys. This period took place after the teams presented the final presentation of their project, but before the professor provided the marks.

The second source of data came from participants' actions recorded in Trello. To conduct this analysis, we adopted Gonзalves et al. (2017) approach. Gonзalves et al. (2017) studied three different agile-based applications during three different timeframes, and, therefore, considered only five types of actions and calculated the average number of actions. In contrast to the study of Gonзalves et al., in this research, all recorded activities in the application were considered as contribution markers. In total, 19 kinds of activities were identified and included in the data analysis. These are creating, changing and deleting cards, leaving and deleting comments, uploading and deleting files, creating, marking and changing due dates, joining and leaving cards, moving cards, changing a board's background, marking, creating and removing checklists. As a result, the sum of the actions was attributed to a particular group. It served as the `participation in Trello' variable. To sum up, all the data collected in this research was quantitative and appropriate data analysis techniques were exploited.

2.7 Data analysis

Regarding the data analysis process, the subject of the study was the team as a whole and its internal processes. Thus, further analysis was conducted on the team level. For data collected via the surveys, the arithmetic mean of the student responses was calculated in accordance with their team number. This data later represented the value of the whole team. The number of teams investigated in the present study was 50, so was the number of observations. The descriptive statistics for all investigated variables were measured, as seen in Table 2. In the study, each variable was represented as the arithmetic mean of the individual elements' comprising it. It was implemented in order to reduce the number of free parameters in the further analysis using Structural Equation Modeling given the number of teams explored. All variables were tested on whether they are normally distributed using the Shapiro-Wilk test as implemented in R. Table 2 reports the Cronbach's alpha, internal-consistency/reliability indicator, as well indicating that all of the subscales used to measure the identified constructs are satisfactory with the values being higher than 0.7 (Nunnally, 1994). Also, Spearman's correlation was calculated to indicate statistically significant relationships between investigated variables.

Table 2 Descriptive statistics of the investigated variables

Construct	Variable	No of items	Mean	St. dev.	Alpha
Teamwork quality (TWQ)	Communication	10	3,80	0,28	0,87
	Coordination	4	4,17	0,38	0,87
	Mutual support	7	4,37	0,38	0,88
	Effort	4	3,77	0,54	0,94
	Cohesion	10	3,94	0,42	0,94
	Balance of member contribution	3	4,08	0,41	0,92
Team members' success	Work satisfaction	3	4,09	0,43	0,92
	Learning	5	3,83	0,37	0,89

For there is an already existing theoretical model behind the research, confirmatory factor analysis was conducted using Structural Equation Modeling implemented in the lavaan package in R (Rosseel, 2012; R Core Team, 2015). The parameters were estimated using maximum likelihood. The advantage of Structural Equation Modeling is that it enables the estimation of both measurement and structural models at the same time. As for the measurement model, it defines how certain concepts are represented using the observed and the latent variables. The latent variables in the study were the following: teamwork quality, which was represented by the six observed variables (communication, coordination, mutual support, effort, cohesion, the balance of member contribution), and team members' success, represented by two observed variables (work satisfaction and learning). As it was mentioned above, teamwork quality is comprised of six observed variables, each of which are related to the teamwork quality concept and describe the quality of collaborative working within the teams. Therefore, a team which exhibited behaviours related to all of the six observed variables can be considered a team with high collaboration activity present (Hoegl & Gemuenden, 2001). Following is a more detailed description of TWQ variables and how they each constitute a part of this concept.

According to Hoegl and Gemuenden (2001), communication within a team needs to have a high degree of openness and a smooth direct method of communication. So, all the project-related information was available to not only the team leader but also the members. Also, there were no additional mediators between the members, when transferring information as it had been found to slow the processes with a team and hinder its overall productivity (Pinto & Pinto, 1990). What is more, the communication between the team members should also be frequent, to ensure that each member possesses up-to-date information, and most of it should have an informal nature. There were some operational activities that, undoubtedly, required some level of formalization (official written reports, communication with a supervisor, etc.). However, it has been found, that informal communication (such as short phone calls, messaging colleagues online, etc.) is necessary to accelerate the transfer of information and create a productive environment within a team (Hoegl & Gemuenden, 2001).

A common understanding and unity concerning the current progress on the project and its overall goals and objectives, in this study, referred to as coordination, is considered an essential facet of the teamwork quality (Hoegl & Gemuenden, 2001). As the project was a complicated practical endeavour, each member usually received their own set of tasks and responsibilities. In order to function efficiently, a team, therefore, had to develop a synchronized unified approach to the goal-oriented scheduling and the structure of the project completion plan, so that each member had a clear understanding of its contribution and members' tasks did not overlap in the process.

Competitiveness within one team tends to be destructive to a project's efficiency when members are working towards the same goals, so a sense of mutual support should be present in a team. When there is intensive collaboration present, the responsibilities of team members are usually interdependent, and members are expected to be respectful, collaborate on ideas of their peers, and offer assistance when there is a need (Hoegl & Gemuenden, 2001).

Additionally, as team members share the same objectives and time constraints, there is a need to have a standard for the amount of effort spent on achieving the project goal by each team member. The workload needs to be shared as equally as possible. Otherwise, the lack of sufficient effort from everyone can lead to a conflict within a team (Pinto & Pinto, 1990).

As to the cohesion facet of the TWQ, it displays the willingness of members to remain with the current team. Three parameters allow determining if the cohesion is present within the team. The members have to be eager to continue working on the tasks, a mutual sense partnership between the members, and a team spirit of appreciation and pride related to the project (Hoegl & Gemuenden, 2001).

What is more, as the professional environment is becoming more sophisticated and the changes made to the project during its completion are fast-paced, it is crucial that each member can contribute his ideas and innovative thoughts during the team discussions. The process of idea generation is not overpowered by a small percentage of individuals to avoid bias and allow for more innovative practices in knowledge-sharing experience within a team. The facet of teamwork quality that is testing this dynamic in a team is called a balance of member contribution (Hoegl & Gemuenden, 2001).

The second latent variable used in the study was team members' success. Its observed variables work satisfaction and learning both provide a sense of personal achievement to an individual working in a team. High work satisfaction can lead to a prolonged period of working in the same team, which is vital for establishing a comfortable social environment in a team that is aimed at better collaborative work in future projects. As to the variable connected with learning, it has been found that an opportunity for both personal and professional growth can act as a motivating factor. It also contributes to the success of the project (Lindsjшrn et al., 2016).

One of the observed variables was the participation in Trello, a sum of actions on the board made by group members, excluding teacher's or facilitator's comments, that were made during the project. Lastly, another observed variable was a team performance, a mark from the professor following the criteria for the evaluation of the project outcome described previously, which were used in the structural model.

In the data analysis, the structural model allows investigating the relations between the constructs. In the current study, the structural model had five paths according to the research model, from teamwork quality to team members' success and team performance, and also from the participation in Trello variable to TWQ, team members' success and team performance. Further, this model fit was tested and reported by the root mean square error of approximation and 90% confidence interval and several other commonly used fit indices, such as SRMR (standardized root mean square residual), CFI (comparative fit index) and TLI (Tucker-Lewis index). agile team education modeling

2.8 Limitations

The study was conducted in one university in Russia and covered only students of Business and Management in their first year of study. It is fair to consider sample size, its homogeneity, and restriction to one university to be limitations of the research, as the majority of the students were the same age and had a similar academic background. Time constraint was also a limitation because the project, in which the students used the Trello application, was only 3,5 weeks long and that might not have been enough to learn how to use it in the most efficient and useful way for their study group. What is more, there were different patterns of application usage, as each group followed a unique project flow. For instance, some groups distributed time and effort spent on each task equally between every project step. In contrast, others chose to perform all the actions necessary to complete the project in the last days before the final deadline. In the latter case, the usage of Trello application might not have influenced the project completion in the same way, as when the execution of the project was performed step-by-step for the duration of the whole project.

Additionally, the data collection method may have been flawed because some students, due to their limited academic background, were not experienced enough in survey taking. They might have misinterpreted some statements dedicated to assessing their teamwork quality and other variables used in the research, which could have had an impact on the conclusions of the study.

What is more, the data for this research was cross-sectional and not longitudinal. A longitudinal study would have provided a thorough look at the development of collaborative work, project management skills, and success of teams over time. Unfortunately, the academic environment was not fit for a study into prolonged collaboration in the same teams. The study courses were designed to last a reasonably short amount of time, and students could choose different subjects and change their social environment often throughout their education.



3. Results

Exploring the data of Trello usage gathered during the project, and data collected via two questionnaires are aimed at exploring a possible relationship between usage of an application designed for project management, teamwork quality, the personal success of members, and team performance in university settings.

The results are able to provide valuable insights into the implementation of the agile approach to project management in the academic environment.

3.1 Shapiro-Wilk test

The first step towards the exploration of the data was to check the missing or inappropriate values in the dataset, which revealed no problematic issues to account for. Further, the Shapiro-Wilk test of normality was conducted on the investigated variables as implemented in R 4.0.0 to see whether they are normally distributed or not. The Table 3 shows that the variables can be considered as normally distributed according to the p-value > 0.05 except for the mutual support (p = 0.018) and balance of member contribution (p = 0.021).

Nonetheless, regarding the relatively small size of the sample, it was decided to retain all the variables causing non-normality in order to maintain statistical power and significance.

Table 3 Shapiro-Wilk test for normality

Variable	N	W	P-value
Communication	50	0.97685	0.4981
Coordination	50	0.97534	0.4447
Mutual support	50	0.93761	0.01759
Effort	50	0.97354	0.3863
Cohesion	50	0.95333	0.06798
Balance of member contribution	50	0.93981	0.02116
Work satisfaction	50	0.94999	0.05073
Learning	50	0.96696	0.2239

3.2 Frequency of Trello usage

The exploration of how frequent and in what exact way the students interacted with Trello was conducted through collection and analysis of the actions stated in the history section of every Trello board. The 20 types of activities identified in the methodology were grouped by the subject to simplify the data collection process and make the representation of the results more compact. Thus, the number of distinct actions reported halved and can be observed in Table 5. The “Added to the board” action reflecting the number of team members added to the personal Trello board is excluded from further consideration as it served as a technical indicator and helped the facilitators to track if all students of HRM course participated in Trello.

Table 4 Statistics of the frequency of Trello usage across teams

Activity	Sum	Mean	Median	Max
Make/delete a comment	392	8	7	42
Attach/remove a file	336	7	8	32
Create/mark/delete/change due date	563	11	9	77
Join/leave card	323	6	6	50
Change/create/delete a card	205	4	3	24
Move a card	628	13	18	36
Change background	25	1	1	10
Mark/unmark a checklist	1667	33	39	67
Create/remove checklist	45	1	2	7

Table 4 reports some of the statistics aggregated across the teams. It can be observed that the most frequent action made by the students is marking/unmarking items in the checklists (1667 in total as well as 33 actions on average in every team). The second most common action is moving cards across the lists which frequently denoted changing the status of the task, for instance, from “to do” to “in progress” status/list. The total amount of that particular action made by the students is 628 whereas the mean value is 13 and the maximum among the teams was 36. Not far behind are the actions connected to the compliance with the deadlines (their creation, removal, and meeting), the sum, mean and maximum are 563, 11 and 77, respectively. On the contrary, the least frequent actions of students are changing the background, creating/deleting checklists, and changing/creating/removing the cards with 25, 45, and 205 total actions across all teams accordingly. Figure 6 depicts the sums of activities sorted from the largest to the smallest.

Figure 6. Sums of actions across teams

Additionally, the four most frequent actions among the students were further analyzed by presenting their distribution via boxplots. Figure 7 shows that the middle 50% of the teams marked/unmarked checklist items around 30-43 times, whereas for comment and due date actions the figures are 5-12 times, and for moving cards the values are approximately 11-25. The highest number of actions in one group is attributed to the due date activities (77), although this number is an outlier as depicted on the plot. Thus, the maximum score of these activities, excluding the other two outliers as well, is around 25 across all teams. Notably, the activity accounting for moving the cards is normally distributed. It does not have any outliers while the other three kinds of action have some skewness (positive or negative) and several outliers. Nonetheless, all of the observations are retained, as later the actions are summed up across each team to form a Trello usage variable, and there is no distinction between the activities in the end.

Figure 7. Distribution of the top four actions in Trello across teams

3.3 Correlation matrix

Table 5 presents the correlation matrix of the investigated variables. The values representing the correlations between the variables of the TWQ and the team member's success constructs are significant at p < 0.01 with the correlations between effort and learning as well as the balance of member contribution and learning being significant at p < 0.05. Besides, most of those values indicate moderate and high correlations between the variables except for the values significant at p < 0.05, which present weak correlations. Concerning the values of the frequency of Trello usage and the team's grade for the project, these variables are not significantly correlated with the six variables of the TWQ. However, there are significant correlations between the Trello usage and the learning variables, and between the grade and the Trello usage where both are significant at p < 0.05, although still represent a weak correlation.



Table 5 Correlations between investigated variables

	1	2	3	4	5	6	7	8	9	10
(1) Communication	1.00
(2) Coordination	0.68	1.00
(3) Mutual support	0.76	0.77	1.00
(4) Effort	0.83	0.76	0.71	1.00
(5) Cohesion	0.78	0.73	0.80	0.87	1.00
(6) Balance of member contribution	0.72	0.77	0.78	0.84	0.87	1.00
(7) Work satisfaction	0.67	0.60	0.70	0.68	0.82	0.74	1.00
(8) Learning	0.47	0.44	0.51	0.37	0.54	0.37	0.64	1.00
(9) Trello usage	-0.02	0.02	0.04	0.18	-0.06	0.10	0.10	0.34	1.00
(10) Grade	0.08	0.04	-0.08	-0.08	0.00	0.02	0.10	0.16	0.33	1.00

Note: The correlations above 0.32 are significant at p < 0.05, the correlations above 0.37 are significant at p < 0.01. N = 50 for all variables.

3.4 Teamwork Quality model

The initial step before analyzing the dependent paths between the constructs is conducting the confirmatory factor analysis of the teamwork quality concept. This analysis allows supporting the existing theoretical background concerning the internal structure of the concept. The measurement model of the construct was specified with the TWQ defined as the latent variable, and communication, coordination, mutual support, effort, cohesion, and balance of member contribution considered as the observed variables to do so. As it was stated previously in the methodology section of the paper, these observed variables were calculated as the arithmetic mean of the individual items within their scales to decrease the number of free parameters and allow the model to be identified and computed. At first, the unstandardized factor loadings of the variables were constrained to be equal (tau-equivalent model), so that every variable loaded on the factor the same.

The estimation of this measurement model, however, produced unsatisfactory results: ч2[14] = 62.066, p =0.000, and RMSEA = 0.276 with 90% confidence interval [0.208-0.348], CFI = 0.832 and TLI = 0.820. The RMSEA exceeds the threshold of <0.10 and thus indicates a poorly fitting model, which is also supported by unacceptable comparative fit index and Tucker-Lewis index as both measures are below 0.90.

Regarding the poor fit of the tau-equivalent model, the option of a congeneric model was considered. This model is unconstrained and allows the factor loadings and the variances of the variables to vary. Proceeding to the fit of the model, the goodness-of-fit measures, CFI and TLI, were equal 0.954 and 0.923 respectively. The values above 0.90 for CFI and above 0.95 for TLI indicate a good fitting model and the values close to 1 suggest an excellent fit (Hu & Bentler, 1999). The other fit measures did not perform well, although were improved to a certain extent: ч2[9] = 22.185, p = 0.008, and RMSEA = 0.18 with 90% confidence interval [0.086-0.277]. The significant chi-square statistic, as well as the RMSEA >0.10, serve as the indicators of a poorly fitting model (Hu & Bentler, 1999).

Thus, the model was explored for the modification indices that could improve the fit of the model and comply with the existing theory. As a result, the model was respecified in order to covary error terms of the mutual support and effort variables. It could be expected that these two variables share a connection as mutual support describes to what extent team members support and help each other when completing tasks. In contrast, effort describes the extent of the involvement of each member in working toward project success. The final model provided the following results:

Table 6 The measurement model of TWQ

TWQ subscale	Std. factor loading	P-value	R-Square
Communication	0.843	0.000	0,743
Coordination	0.821	0.000	0,686
Mutual support	0.846	0.000	0,782
Effort	0.923	0.000	0,896
Cohesion	0.938	0.000	0,840
Balance of member contribution	0.915	0.000	0,794

According to Table 6, all observed variables are loading high on the TWQ construct (the standardized factor loadings are significant as the p-value < 0.001), with the coordination loading the lowest (0.821) and cohesion loading the highest (0.938). The variance of the variables explained by the TWQ concept is considered good as the R-Square values range from 0.686 to 0.896. The fit of the model considerably improved: ч2[8] = 10.732, p =0.217, and RMSEA = 0.087 with 90% confidence interval [0.000-0.207], p-value RMSEA <= 0.05 = 0.291. The RMSEA met the threshold of <0.10, and the upper confidence interval exceeds the mentioned threshold, which does not allow to say with confidence that the model fits the data well. However, the p-value RMSEA <= 0.05 is insignificant, which allows retaining the hypothesis of a close fit of the model. The comparative fit index and the Tucker-Lewis index of the model are 0.99 and 0.982, respectively, suggesting an excellent model fit.

3.5 Structural model

The examination of the measurement model of TWQ concept allows to proceed with the specification of the structural model mentioned in the methodology. In this study, the goal of the structural model was to investigate the possible relations between the TWQ, team member's success and the team's grade for the project as well as the effect of those relations. According to the literature review and the research questions, the suggested relations of the concepts were as follows:

The eight investigated variables are depicted as the rectangles and represent the observed variables, and so are the observed frequency of Trello usage variable and the team's grade. The latent variables, or the constructs, are shown as ellipses. The double-headed arrows represent the error variances. The arrows from the TWQ concept to the six rectangles represent the relations in the measurement model and indicate the factor loadings of the scales. Similarly, the arrows from the team member's success to the work satisfaction and learning in rectangles represent the second measurement model. The other arrows from the TWQ to the team member's success and team's grade variables, as well as the arrows from the frequency of Trello usage to the TWQ, grade, and team member's success represent the structural paths in the model and are to provide the coefficients of the relations.

Figure 8. Investigated measurement and structural model

The model based on the specification above provided the following fit: ч2[30] = 47.114, p =0.024, and RMSEA = 0.119 with 90% confidence interval [0.044-0.182], p-value RMSEA <= 0.05 = 0.060, SRMR = 0.053; CFI = 0.948, TLI = 0.922. The fit of the model can be characterized as contradictory due to the RMSEA and its upper confidence interval exceeding the threshold of 0.1. However, the lower confidence interval (0.044) is less than 0.05, and the p-value RMSEA <= 0.05 is insignificant, which indicates that the close-fit of the model hypothesis is not rejected. Kline (2005) argues that this type of RMSEA behavior is caused by the bias of the metric towards smaller sample studies. Concerning the other fit indices, the SRMR is below 0.08, which, according to Hu and Bentler (1999) indicates a good model fit. Also, the CFI is above the threshold of 0.9, which supports an acceptable fit of the model.

Figure 9 depicts the investigated measurement and structural model, note that the reported values are standardized. Starting with the factor loadings, all of them are significant at p < 0.001. Concerning the TWQ construct, the six manifest variables are loading high with effort and cohesion loading the highest (0.933 and 0.934 respectively) and coordination having the lowest loading of 0.824 and the highest error variance of 0.321. The team member's success is represented by work satisfaction and learning which standardized factor loadings are 0.967 (high) and 0.661 (moderate) respectively, and their variances are 0.066 and 0.564.

Figure 9. The measurement and structural model with the standardized factor loadings and structural path coefficients

Proceeding to the explored relations between the constructs, the significant structural path coefficients are observed from the TWQ to the team member's success, as well as from the frequency of Trello usage to the team member's success and the team's grade for the project. The reported path coefficients are standardized and thus imply that the fluctuations in the standard deviations of the independent variable produce the change in the standard deviation of the dependent variable as given by the estimated coefficient. For instance, an increase of 1 standard deviation in the Trello usage variable will produce with a 95% confidence interval an increase of 0,31-0.34 in the standard deviation of the team's grade given the estimated coefficient of 0.32 between the constructs. The lowest structural path coefficient is observed from the TWQ to the team's grade, which is negative as well (-0.03). However, this path is not as significant as p = 0.864. On the contrary, the highest coefficient was from the teamwork quality to the team member's success (0.83) with the p < 0.001.

Regarding the hypotheses of the study, the results can be interpreted as contradictory.

The study did not support H1: Participation in Trello positively influences teamwork quality.

The conducted analysis using Structural Equation Modeling revealed that the standardized structural path coefficient between the concepts is below zero and equals -0.08, and the R2 = 0.006, although p = 0.593 which makes these estimations insignificant, resulting in the rejection of the first hypothesis.

The analysis supported H2: Participation in Trello positively influences team performance.

The standardized coefficient of this relation is 0.322, with the p = 0.03 indicating the significance of the relationship. The cumulative effect imposed on team performance can be characterized as medium, and R2 = 0.114.

The study supported H3: Participation in Trello positively influences team members' success.

The standardized estimation of the relation is 0.185, and the p-value = 0.038, which provides a significant result and allows to retain the hypothesis suggesting a positive effect of Trello usage on team member's success.

The analysis supported H4a: Teamwork quality positively influences team members' success.

The standardized coefficient of the path is 0.825 with the p < 0.001, providing a significant estimation. Thus, the cumulative effect of the Trello usage and the TWQ can be considered as significant and large. It is represented by R2 = 0.690.

The study did not support H4b: Teamwork quality positively influences team performance.

The estimation of the structural path coefficient (-0.03) occurred insignificant with the p-value = 0.864. Since the relation from the TWQ to the grade did not prove to be significant, the medium effect on the grade can be entirely attributed to the Trello usage variable (R2 = 0.114).

To sum up, the graphical representation of the results of hypotheses testing is presented in Figure 10. The analysis supports the green-coloured hypotheses. The red-coloured ones are rejected. R-squared denotes the proportion of the variance in the dependent variable explained by the independent variables in the structural model.

Figure 10. Results of hypotheses testing process.



Conclusion

Data collection and its further analysis using Structural Equation Modeling were used to investigate the influence of using an application designed for project management in the university environment. This study was mainly dedicated to examining whether the usage of Trello influenced teamwork quality of the students in one university in Russia. Additionally, team performance and team members' success were examined concerning Trello usage. What is more, the influence of teamwork quality on the individual success of team members' and overall team performance was explored.

It was found that participation in Trello positively influenced the team performance of the students. The team performance variable consisted of a grade provided by one of the professors responsible for the Human Resource Management course. Thus, it may be speculated that the application helped students in planning and developing the project, therefore, affecting the quality of it, which allowed students to receive a high grade.

Additionally, it has been found that Trello also had a positive influence on team members' success. The team members' success variable was measured by two observed variables, learning and work satisfaction. The students who used Trello more actively also showed higher satisfaction with the work performed for the project and higher scores connected to learning new information during the teamwork.

However, the data analysis showed no significant relationship between Trello usage and teamwork quality. As Trello is a tool designed for project planning and collaboration between team members, this result was somewhat unexpected. The reason for this outcome may be connected to the amount of time dedicated to this project during the HRM course. The project lasted four weeks, and, as students were allocated to the teams in random order, they had no previous experience in working in the same group of people before and the majority of them (about 83%) did not work with Trello before. It may be argued, therefore, that the time constraints and inexperience in project management of first-year students had been an obstacle to developing a significant connection of Trello and teamwork quality.

Lastly, the inspection of the relations between the TWQ and individual team member's success revealed a significant positive result. The teams in which the level of teamwork quality was higher provided higher scores on the questionnaire concerning the learning within the team as well as the satisfaction with the work done. This discovering was expected as the aspects of these constructs are thought to be connected and correlated. Indeed, the atmosphere in teams where productive communication, support of the fellow team members and cohesive goal-achieving take place is considered as a facilitator for more in-depth learning and a motivator to participate in the future work with the same team. Nevertheless, the mentioned merits of the high TWQ did not provide a significant relationship between this concept and the performance of the team. From the conducted analysis, it cannot be said with confidence whether there is a positive or negative effect of the TWQ on the grade. The rater bias might explain this fact as several professors assessed the teams' projects, and regardless of the grading criteria, these grades might tend to be subjective.

...