Ufone, one of Pakistan's leading telecom networks, needed clearer visibility into performance across its product portfolio and a reliable way to anticipate daily recharge behavior. With operations spanning 132 districts and 30 active national offers, decisions were often made with incomplete or delayed data, making proactive planning difficult.

The challenge was twofold: build dashboards that gave stakeholders real-time visibility, and develop a forecasting model accurate enough to actually inform business decisions.

Rather than treating this as a pure data science exercise, the approach centered on usability and business relevance from the start.

The strategy included:

Mapping which metrics mattered most across districts and offers before building any dashboard

Choosing a forecasting method (SARIMAX with Fourier seasonality terms) suited to recharge data's weekly and monthly cyclical patterns

Coordinating directly with product, brand, and technical teams to ensure forecasts and dashboards aligned with how the business actually made decisions

Validating the model against real historical data through a train-test split before rolling it out

The technical build included:

Performance dashboards covering all 132 districts and 30 offers for cross-functional visibility

A Python-based forecasting model using SARIMAX and Fourier seasonality, tested against a 90-day holdout period

Coordinated communication updates with product, brand, and technical teams as offers and portfolio changes rolled out

This project shows how data and marketing strategy work together in practice. The forecasting model wasn't built as a standalone technical exercise. It directly shaped how the business planned campaigns and timed communications, turning a prediction problem into a decision-making tool.

The forecasting model achieved 95% accuracy in predicting daily recharge, giving Ufone a dependable planning tool across one of Pakistan's largest telecom networks.

Tools: Python, Google Colab, SARIMAX, Fourier seasonality modeling