Automating Forecast Management Operations in the FP&A Department Using Predictive Models: Advancing Management Control Through AI Utilization Tailored to Account Characteristics

LIXIL’s FP&A function is responsible for forecasting performance and providing analysis and recommendations to support management decision-making. However, forecasting processes relied heavily on the experience of individual personnel in each business division, requiring significant effort for data collection and processing while leaving insufficient time for analytical work.
To address these challenges, LIXIL had previously attempted to automate forecast generation using machine learning models based on statistical analysis. However, due to competing priorities and limited internal resources, the desired level of accuracy could not be achieved despite developing models across various segments, products, and accounts. As a result, many manual tasks remained and independent model development proved difficult.
Against this backdrop, LIXIL engaged ABeam Consulting for its specialized expertise, project execution capabilities, and overall capabilities. ABeam began supporting the initiative from the planning phase in April 2024, and the Performance Forecasting Model Development Project officially launched in January 2025.

In this project, achieving the required forecast accuracy for key profit-related accounts within a one-year pilot period was critical. The team initially adopted a broad modeling approach across account and product dimensions, then refined areas where accuracy fell short. As the initial results did not meet expectations, the modeling granularity was increased.
For revenue and standard cost, products were segmented into eight categories by combining product and market classifications, enabling the required level of accuracy.
For cost variances, other costs, direct labor costs, and other allocated expenses, machine learning models were first evaluated through proof-of-concept testing. Where the required accuracy proved difficult to achieve, formula-based calculations (shown in blue in Figure) were adopted to ensure an acceptable level of accuracy and quality. By tailoring the methodology to the characteristics of each account, the project successfully met business requirements within a limited timeframe.
A key success factor was the flexibility to adjust requirements and implementation approaches based on actual results rather than adhering to the original plan.

For more than a decade since 2014, ABeam Consulting has supported LIXIL across a broad range of areas, including logistics and finance. Drawing on this deep understanding of the client, ABeam identified key management challenges and defined a roadmap centered on three transformation themes. The first initiative was the Performance Forecasting Model Development Project, aimed at improving efficiency and analytical capabilities through the use of forecasting models.
Rather than simply delivering a system, ABeam combined its knowledge of LIXIL’s business and operations with AI expertise, bridging business and IT while supporting future internalization of the solution.

During the one-year pilot, both teams maximized limited resources through short, iterative cycles, expanding coverage as required accuracy was achieved. By continuously refining plans based on project outcomes, the first phase was successfully completed in March 2026.
Building on the pilot foundation, LIXIL plans to accelerate deployment across additional business divisions. The company also aims to enhance analytical capabilities by leveraging forecast, plan, and actual data, with a long-term vision of introducing simulation functions to support profit improvement through FP&A.