Case Study · Finance
Kuveyt Türk: master–apprentice mentoring for the Digital Marketing and Retail Customer Analytics teams
It began in 2023 with SQL, Power BI, Python and machine learning training for Digital Marketing's analytics team. The team then built its own projects under Lumtify's mentoring: digital score, target management, churn cause analysis, UI usage analysis, product recommendation and churn prediction models. Since 2025 the work has continued with the Retail Customer Analytics team on spending analytics, a luxury spend score and next-best-action prioritisation. The team writes the code; Lumtify sets direction, brings the method and checks the output.
- Master – apprentice
- Model
- 8+
- Digital projects
- 2025 –
- Retail analytics
- 4 modules
- Training
On-the-job training; the team writes the code
From descriptive to predictive
Spending, luxury score, NBA
SQL · Power BI · Python · ML
Context
In 2023 Kuveyt Türk’s Digital Marketing Group was building its own analytics team. It had the people and the data; what it lacked was technical skill and a way of working that would take a project from start to finish. From the start, Lumtify’s role was defined as getting the team to do the work itself: the master–apprentice model, on-the-job training.
Training first, then projects (2023)
Three trainings were delivered between April and June: data awareness and an introduction to AI, SQL with reinforcement projects, data visualisation with Power BI. Project consulting continued from June to December and the team carried five projects to MVP with its own hands:
- Digital score: a score computed from each customer’s use of digital and physical channels; a dashboard that can be analysed at both customer and bank level.
- Target management: a sustainable data preparation that tracks customers’ digital activity over the last 3–6–9–12 months and allows audiences to be built from demographics and product use.
- Digital assistant insight: an analysis comparing customers who discussed specific products with the assistant against their activation on those products. It showed which products get activated after a conversation.
- Churn cause analysis: the errors customers who left the digital channels hit at login; the profile of those who got an error on a given login method and never came back.
- UI usage analysis: time per screen, clicks, errors and daily trends; a dashboard measuring the effect of design changes.
On every project the team member wrote the plan, the SQL and the dashboard, with Lumtify alongside at every step. The goal was a usable MVP. Each project ended in a Power BI dashboard that was sustainable, documented and ready to be integrated into the core system with IT.
From descriptive to predictive (2024–2025)
In 2023 the questions were mostly about “why”; in 2024 they moved to “what will happen”. Python and machine learning training was delivered and the projects were taken to the next step: segmentation on the digital score, a product recommendation model from digital assistant data, a churn prediction model, at-risk customer detection producing alerts, a retention dashboard and DataAlert. Work on a digital score for corporate customers began. A data model combining all projects into a single customer life story, with natural-language audience building on top of it, was added to the roadmap.
The team changed over time; shortened repeats of the trainings for new joiners became part of the engagement.
The second team: Retail Customer Analytics (2025–)
A new period began in 2025 with the Retail Customer Analytics team. This team was already technically strong. What it needed was to back its projects with data and get small models into use quickly. So each team member started with a small model; projects were kept short, and models were grown once they were in use.
Spending analytics roadmap: RFM-based segmentation of card spending with segment naming; usage frequency, day- and hour-based spending behaviour and its relation to the portfolio; spending distribution by merchant category group (MCG). Luxury spend score: a customer-level score derived from merchant categories.
Next-best-action (NBA) prioritisation: scores produced by dozens of the bank’s models were being combined in the same expected-value formula. The churn score was a real probability but the cross-sell scores were not, so the same action always came out on top. Lumtify’s methodology note defined the two missing pieces. The first was a calibration layer on top of the outputs, leaving the models untouched (isotonic regression). The second was bringing the action’s own incremental effect into the formula. The note also proposed deriving the threshold from cost and tying capacity to an assignment problem. The one non-negotiable was the measurement loop: logging and a control group. Giving the method the team had already tried its name in the literature was one of the most valuable parts of the work.
Customer Experience and Academy
In 2025 a training-and-mentoring programme was designed for the Customer Experience team, running from data literacy to GenAI use, with an applied project at the end of every module. In 2026 an evaluation of the 2023–2024 period and an expanded programme proposal were prepared for Kuveyt Türk Academy: pre-assessment, small groups, practice on the institution’s real data, post-training project mentoring, before-and-after measurement.
Outcome
Over three years two teams moved from descriptive analytics to machine-learning models, and from MVP to dashboard to productisation, writing the code themselves throughout. Lumtify set direction, brought the method and checked the output.