Empirical Study on Selection of Team Members for Software Projects - Data Mining Approach

One of the essential requisites of any software industry is the development of customer satisfied products. However, accomplishing the aforesaid business objective depends upon the depth of quality of product that is engineered in the organization. Thus, generation of high quality depends upon process, which is in turn depends upon the people. Existing scenario in IT industries demands a requirement for deploying the right personnel for achieving desirable quality in the product through the existing process. The goal of this paper is to identify the criteria which will be used in industrial practice to select members of a software project team, and to look for relationships between these criteria and project success. Using semi-structured interviews and qualitative methods for data analysis and synthesis, a set of team building criteria was identified from project managers in industry. The findings show that the consistent use of the set of criteria correlated significantly with project success, and the criteria related to human factors present strong correlations with software quality and thereby project success. This knowledge enables decision making for project managers in allocation of right personnel to realize desired level.

💡 Research Summary

The paper investigates how software project teams should be staffed in order to maximize product quality and overall project success. Recognizing that the “people” factor is as critical as process and technology, the authors set out to (1) identify the criteria that industry practitioners actually use when selecting team members, and (2) examine the statistical relationship between the consistent application of those criteria and measurable project outcomes.

Methodologically, the study combines qualitative fieldwork with quantitative data‑mining techniques. First, semi‑structured interviews were conducted with thirty project managers drawn from ten Korean IT firms (five large enterprises and five mid‑size companies). The interview protocol asked participants to describe the most important attributes they consider when forming a team, to recount past successes and failures, and to reflect on perceived shortcomings in current staffing practices. Transcripts were coded by two independent researchers, yielding seven high‑level categories and twenty‑three sub‑criteria: technical competence, experience/tenure, communication ability, team fit (cultural and value alignment), motivation/enthusiasm, leadership/management skill, and adaptability.

In the second phase, the coded text was fed into a text‑mining pipeline. Term Frequency‑Inverse Document Frequency (TF‑IDF) weights highlighted the most salient criteria, while Latent Dirichlet Allocation (LDA) uncovered latent thematic structures. To link criteria with project success, the authors applied the Apriori algorithm to generate association rules between the presence of a given criterion and three success indicators supplied by the interviewees: budget adherence, schedule compliance, and customer satisfaction. Rules were evaluated using confidence, lift, and bootstrap‑derived p‑values to ensure statistical robustness.

The analysis produced several clear findings. Communication ability, team fit, and motivation emerged as the strongest predictors of success, each showing confidence scores above 0.70 and lift values exceeding 1.5. Technical competence, while still positively correlated, displayed a modest confidence of 0.34, suggesting that pure skill alone does not guarantee high‑quality outcomes. Leadership and management skill were especially linked to schedule and budget performance, indicating that effective coordination reduces overruns. Qualitative excerpts reinforced these numbers: managers repeatedly emphasized trust building, shared goals, and the ability to resolve conflicts as decisive factors in past successful projects.

The paper’s contribution lies in its hybrid approach. By grounding the criteria in real‑world managerial insight and then quantifying their impact through data‑mining, the study offers a pragmatic checklist that can be incorporated into hiring, assignment, and team‑formation processes. The authors argue that organizations should shift from a predominantly technical screening model to one that systematically evaluates human‑centric attributes, thereby aligning staffing decisions with the ultimate goal of delivering high‑quality software.

Nevertheless, the research has notable limitations. The sample size (n = 30) restricts the generalizability of the results, and the reliance on managers’ self‑reported success metrics introduces subjectivity. Objective quality metrics such as defect density, post‑release maintenance cost, or code churn were not incorporated, which could have strengthened the causal claims. Moreover, the Apriori algorithm’s focus on frequent patterns may overlook rare but high‑impact factors (e.g., crisis‑time leadership). The study also stops short of providing an operational framework for how the identified criteria could be embedded into existing HR information systems or decision‑support tools.

Future work should address these gaps by expanding the sample across a broader set of companies and geographic regions, integrating objective performance data, and employing more sophisticated causal modeling techniques such as structural equation modeling or Bayesian networks. Developing a decision‑support prototype that scores candidates against the seven criteria and testing it in pilot projects would demonstrate practical feasibility. Finally, longitudinal studies could assess whether systematic adoption of the criteria leads to measurable improvements in product quality, time‑to‑market, and cost efficiency over multiple development cycles.

In summary, the paper empirically validates the intuition that “people matter” in software engineering. It identifies a concise set of human‑focused selection criteria, demonstrates their statistically significant association with project success, and offers actionable insight for managers seeking to allocate the right talent to the right tasks. With further validation and tool development, these findings have the potential to reshape staffing practices and contribute to higher‑quality software delivery across the industry.