Client challenge

An American digital communications company approached KPMG in India with several critical challenges in their financial forecasting process

Lack of granularity and precision

Their existing forecasting approach relied heavily on traditional top-down models, which did not provide the necessary level of detail or accuracy. This limited the ability to make informed, data-driven decisions.

Difficulty explaining forecast variances

The FP&A team faced challenges in understanding and explaining forecast variances, particularly when there were shifts in the underlying product mix or non-ideal behavior at the transaction and contract level. This made it difficult to isolate key business drivers.

Manual, time-intensive processes

The forecasting process was largely manual and required significant time and effort from the FP&A team, diverting focus from strategic analysis.

Data volume constraints

The organisation had access to large volumes of historical data, but the limitations of excel and the compute power of the local machines made it impractical to process and analyse this data effectively, hindering the potential for more accurate, data-rich forecasting.

Lack of granularity and precision Difficulty explaining forecast variances Manual, time-intensive processes Data volume constraints

Lack of granularity and precision

Their existing forecasting approach relied heavily on traditional top-down models, which did not provide the necessary level of detail or accuracy. This limited the ability to make informed, data-driven decisions.

Difficulty explaining forecast variances

The FP&A team faced challenges in understanding and explaining forecast variances, particularly when there were shifts in the underlying product mix or non-ideal behavior at the transaction and contract level. This made it difficult to isolate key business drivers.

Manual, time-intensive processes

The forecasting process was largely manual and required significant time and effort from the FP&A team, diverting focus from strategic analysis.

Data volume constraints

The organisation had access to large volumes of historical data, but the limitations of excel and the compute power of the local machines made it impractical to process and analyse this data effectively, hindering the potential for more accurate, data-rich forecasting.

Services provided

KPMG in India has developed the intelligent forecasting workbench- a centralised, AI powered platform designed to transform the client’s forecasting capabilities. It includes:

Multiple business metrics forecast

Enabled simultaneous forecasting of key business metrics such as Average Selling Price (ASP), service bookings, and revenue, all within a unified platform-driving consistency and deeper insights across business areas

Explainability engine

Introduced transparency through explainable AI models, providing clear and actionable insights into both internal drivers and external market signals affecting forecasts

Automation and scalability

Automated key steps including data ingestion and model selection. The service, deployed on Microsoft Azure, offered scalable architecture capable of handling large volumes of historical data-eliminating excel limitations and manual bottlenecks

Human-in-the-loop design

Integrated expert judgment into the forecasting loop, allowing FP&A teams to validate, override, or adjust forecasts based on strategic or domain knowledge-combining the best of machine intelligence and human expertise

AI/ML powered forecasting

Leveraged machine learning algorithms to deliver highly accurate, data-driven and explainable forecasts, improving both precision and reliability

Data Ingestion

Model Building

Automation

Model Evaluation

Human-in-the-loop

Visualisation

Collaboration

Cloud Deployment

Impact

The implementation of Our Intelligent Forecasting Workbench delivered measurable improvements across multiple dimensions of the client’s forecasting function:

Improved forecast accuracy

The use of AI/ML-powered models led to a 12-15 per cent improvement in forecast errors, significantly enhancing the reliability of financial projections and decision-making

Enhanced insights and explainability

The built-in explainability engine provided a deeper understanding of forecast drivers, enabling the FP&A team to clearly identify and articulate the root causes of forecast variances-improving transparency and stakeholder confidence

Greater scalability

The scalable Azure-based architecture enabled the system to handle large volumes of historical data and support more granular, business-specific forecasting, positioning the client for future growth

Efficiency and time savings

Automation of data ingestion, model selection, and forecast generation resulted in significant time and effort savings, eliminating manual processes and saving several manhours per forecast cycle-freeing up the FP&A team to focus on strategic activities