It From Bit · Research

The research record.

Our work is grounded in published research. Selected peer-reviewed contributions from our practice areas.

Peer-reviewed publications

  • Peer-reviewed journal article · 2026

    Comparative Analysis of Modern Machine Learning Models for Retail Sales Forecasting

    Mario Brcic, Luka Hobor, Lidija Polutnik, Ante Kapetanovic — Croatian Operational Research Review, 2026

    A triage framework for model selection in retail demand forecasting under real-world conditions: intermittent demand, missing data, and production constraints. Across major retail datasets, tree-based ensemble methods (XGBoost) consistently outperform deep learning architectures — a finding with direct implications for how organizations choose forecasting infrastructure. Co-authored with Babson College and mStart (Fortenova Group).

    arXiv preprint →
  • Conference paper · 2026

    AI-Assisted Unit Test Writing and Test-Driven Code Refactoring: A Case Study

    Ema Smolic, Mario Brcic, Luka Hobor, Mihael Kovac — MIPRO 2026

    Documents a client engagement in which AI coding models generated approximately 16,000 lines of unit tests in hours rather than weeks, enabling safe large-scale refactoring with up to 78% branch coverage. A practical template for organizations managing legacy codebase risk with AI tooling.

    arXiv preprint →
  • Working paper · 2026

    Policy-Bound Triple-Entry Receipts for Autonomous Commerce

    D. Kapusta, Mario Brcic — with COTRUGLI Business School & HashNet

    Proposes an accounting control architecture for AI-mediated transactions where execution speed outpaces human governance. The Policy-Bound Triple-Entry (PBTE) method uses an Accounting State Machine to gate transaction recognition against pinned policies — enabling event-time governance across ERP systems. Foundational work for agentic commerce infrastructure.

    Read on ResearchGate →
  • Peer-reviewed journal article · 2022

    Delivery pattern planning in retailing with transport and warehouse workload balancing

    Mario Brcic et al. — Croatian Operational Research Review, 2022

    Documents the methodology and results of the supply chain optimization engagement with Konzum — Croatia's largest retail chain — during the COVID-19 pandemic. Presents a discrete optimization model for weekly delivery pattern planning that balances warehouse and transportation utilization while maintaining service levels for fresh food across hundreds of stores in the CEE region.

    Read on HRČAK →

See these methods applied in practice → Case Studies