Replication package - Training as an Ecosystem: How Software Engineering Professionals Perceive the Quality of Corporate Training

This package supports a cross-sectional mixed-methods survey with 282 Software Engineering professionals in the Brazilian software industry (data collected from September 9 to November 2, 2025). The survey combined 27 Likert items, built on the training framework of Salas et al., with open-ended questions. Files survey_instrument_pt.pdf: questionnaire as applied (Brazilian Portuguese). survey_instrument_en.pdf: English translation of the questionnaire. survey_responses.xlsx: the 282 valid responses, one row per participant. Sheet codebook gives, for each column, a short code, the dimension of the framework of Salas et al. (TNA, AC, IMS, PTC), an English description, and the original question. It also documents the Likert coding: 1 = Discordo totalmente (strongly disagree) to 5 = Concordo totalmente (strongly agree). Q11 is negatively worded and kept raw; it was reverse-coded only in the analyses. survey_responses.csv: the same responses in UTF-8 CSV, with the short codes as header. thematic_analysis.xlsx: working sheets of the reflexive thematic analysis. QQ2 v2 (final): coding of the open question analyzed in the paper (suggested improvements), which produced the five reported themes. QQ2 v1: first coding round. QQ1: coding of the other open question (most useful learning experience), analyzed in prior work and included for completeness. item_construction_matrix.xlsx: construction of the questionnaire items, from candidate questions collected from the research team and the literature to the final 27 items. factor_output_efa.txt: full output of the exploratory factor analysis in FACTOR 12.06.07. anonymization_log.csv: list of all anonymizations applied. Participant codes. participant_id (P1–P282) matches the participant codes quoted in the paper. Anonymization. Research-team members appear as PESQUISADOR A–E. Names of employers or institutions mentioned by participants were replaced by [EMPRESA] (company) or [INSTITUIÇÃO] (institution). Answers, codes, and themes are otherwise reproduced exactly as recorded, in Portuguese. Not included. The analysis scripts are not shared. All analyses can be reproduced from the shared data with the software and settings reported in the paper: FACTOR 12.06.07 for the exploratory factor analysis; lavaan with the WLSMV estimator for the confirmatory and bifactor models; non-parametric tests with Benjamini–Hochberg correction, and a linear regression with robust standard errors, for the group comparisons. Ethics. The study was approved by an institutional Research Ethics Committee (approval number withheld for double-anonymous review). Participants gave informed consent, and the committee authorized sharing the anonymized responses.