Case Study · Finance
Türkiye Finans: training and project mentoring for the HR analytics team
In 2024 we spent one day a week with the participation bank's HR Analytics team: first Python and machine learning training, then mentoring on the team's own projects. The relationship between the recruitment assessment inventory and performance and retention was analysed by job family, and a literature review and model design were done for employee attrition prediction. The team applied what it learned to the institution's own data straight away.
- 1 day a week
- Format
- Python · ML
- Training
- 2
- Projects
- Master – apprentice
- Approach
Training + mentoring, six months
Assignment-driven, with exams
Recruitment inventory · attrition
The team does the analysis
Context
Türkiye Finans’s Human Resources Analytics team wanted to make better use of its technical skills. Lumtify Academy’s “data analytics journey” catalogue for banks has introductory, foundation, reinforcement and advanced trainings. From it we drew a path specific to HR analytics. The engagement put training and mentoring on the same calendar and ran one day a week for six months.
The roadmap followed familiar steps: meetings with managers and goals, one-to-ones with the team, assessing competence levels, an inventory of tools and data, identifying the target project for mentoring, the trainings needed, and starting the project.
Training
Weekly Python training started in February: language fundamentals, assignment-driven progress, examples from the team’s current work. Machine learning followed in May: preparing data for a model, feature engineering, model validation, regression and tree-based methods. The training was measured with an exam and closed with feedback.
Project 1: Recruitment inventory analysis
The bank uses a comprehensive assessment tool in recruitment, covering candidates’ general knowledge, aptitude and personality inventories. This data, accumulated over years, had never been related to the later performance of those hired or to whether they stayed at the bank.
The analysis was done by job family: retail banking, call centre, information systems, branch support, digital banking, loans and relationship management roles. For each job family the team looked at which inventory dimensions relate positively or negatively to performance, compared dimension means between leavers and stayers with a t-test, and examined the effect of the university attended. The result was a set of job-family-specific findings on which dimension to look at for which role in the hiring decision. In some roles the relationships ran against expectation.
The team did the analysis and prepared the presentation; we supported them on method and interpretation.
Project 2: Employee attrition
The second project was a model predicting the probability of an employee leaving. The team did a literature review and screening, identified the data needed and designed the model. The findings of the inventory analysis became feature candidates for this model.
Outcome
At the end of six months the team could apply statistical tests and machine learning to its own data and prepared its own presentations. We ran training and project with the same team in the same period because we wanted each method to be used on the institution’s own question as soon as it was learned.