Orbico

Data Exploration & KPI Definition

Assessment of available transactional data, customer history, product information, and weekly purchasing patterns. Definition of measurable KPIs such as prediction accuracy, customer category stability, and SKU-level forecast performance.

Domain Discovery & Infrastructure Setup

Mapping data flows, analyzing seasonal variability, identifying challenges related to new products, and defining model integration points within Orbico’s existing systems.

ML Model Design (Proof of Concept)

Development of initial machine learning models for:

• Customer Categorization – grouping similar B2B buyers based on purchasing behavior

• Weekly Order Prediction – SKU-level demand forecasting for each customer

• Scoring Model – supporting sales teams in prioritizing clients with the highest predicted demand or risk of inactivity

This stage also included evaluation of different ML approaches, feature engineering, and baseline accuracy measurement.

Full Model Implementation & Automation

End-to-end development of automated prediction pipelines, including:

• daily ETL workflows

• weekly forecast generation

• automatic model retraining

• data quality validation

• new SKU detection

• alerting and notification systems for sales teams

This phase ensured full production readiness and scalability across product categories and markets.

Model Monitoring, Calibration & Rollout

Continuous monitoring of model performance, recalibration based on market dynamics, and adjustments for new product categories. Following successful validation, the solution was expanded to Bulgaria and Slovenia.