Mitsui increases forecast accuracy with AI and Predictive Analytics

As part of Mitsui Chemicals Inc., Mitsui Chemicals Europe has a wide range of product groups in its portfolio. With the existing planning solution based on Microsoft SQL and Access and the typical Excel scenarios, this was too complex. In 2017, the Performance Controlling department, together with Jedox partner HighCoordination, implemented a new planning solution with Jedox that significantly simplified and accelerated sales planning for Mitsui Chemicals Europe. Since the Jedox platform integrates various sub-plans, Mitsui can quickly adapt to changing market conditions at the push of a button enabling much more agility in planning.

Predictive Analytics in sales forecasting

With a goal of increasing forecast accuracy, the project team now implemented the Jedox AI module “Predictive Forecast” in a pilot project. The module, which is based on the integrated Jedox AIssisted™ Planning solution and includes predefined business logic, supports the 12-month rolling sales forecast at Mitsui with AI-generated sales predictions. After just a few months, the accuracy of the forecast in relation to the actual values has seen increases to up to 95 percent for several product groups – and further optimizations are expected by continuously growing the database.

In this pilot project, the AI standard model provided by Jedox was implemented, which already delivers initial reports and sales evaluations. At Mitsui, the primary focus was on a higher level of detail, e.g. to obtain forecasts the different product groups and product SKUs, the individual customers, materials and regions.

The major challenge was the large amount of data required for AI-analyses. Jedox recommended a data set that covers at least the three previous years as the foundation. However, initially only 1.5 years of data was available. Mitsui used the Jedox Integrator, the data hub in the Jedox solution, to map additional legacy data in SAP and migrate it to Jedox. Their existing data cubes were also optimized for data quality and performance.

Mitsui uses the sales prediction from the AI engine in parallel to the manual forecast values to support the decisions of their sales planners with the automated comparison. The aim is to strengthen confidence in the AI-generated values and to further automate forecasting on the basis of broad system acceptance.

Precision and transparency in detail

The high level of detail achieved in planning and forecasting was a big win for Mitsui. The powerful Jedox In-memory engine supports agile planning such as splashing of values on the granular level of individual customers and materials or data aggregation across all hierarchy levels at the push of a button.