Banksalad Data Organization Achieves 96% ML Classification Accuracy and Spreads OKR Culture through Planning Shop

Banksalad's Data Foundation organization held a 'Planning Shop' to increase transparency of company-wide target metrics and establish a data-driven decision-making culture. This event was run intensively for a full day at an external venue for team-by-team retrospectives and next-quarter goal setting, and with full leadership support, an attempt was made to define goals based on the ownership of working-level staff.

Financial Data R&D Results

The Financial Data R&D team set a goal of improving the category classification accuracy of household ledger card spending records. As a result of switching from the existing Rule-based model to a Machine learning-based model, classification accuracy rose sharply from the previous level of around 70% to 96%. This achievement became an opportunity to gain technical confidence that ML models could also be applied to bank transfer records.

Experiment Platform and Organizational Culture Improvement

The Experiment Platform team aimed to increase activation of the in-house A/B testing tool, and average query views increased by 81% and tool users increased by 80% compared to the start of the quarter. This was evaluated as an indicator showing that data-driven culture is spreading within the organization. In addition, the Data Engineering team and Planning team established guidelines for shifting from qualitative-goal-focused to quantitative-goal-focused approaches.

Company-wide Sharing and Guide Distribution

The action plans, OKR setting guide, and retrospective templates derived after the Planning Shop were shared with all members through a company-wide notice. In particular, by providing guidance on weekly and mass-sprint-level metric monitoring methods for people unfamiliar with data, a foundation was laid for data-driven goal management to expand across the entire organization.