Just a few years ago, the finance team for the procurement and supply chain function at our client was facing unexpected complexity in answering fundamental cost questions. Questions like:

• What is the cost per hectoliter (commonly used to measure the volume of beer in brewing operations and is often used in international reporting of beer shipments) by brand and country?
• How have production costs changed year-over-year?
• Which specific brands or cost categories are driving the variance?
• Forward looking, what is the impact of large market-shifting events? What if?
• Which suppliers are above my benchmarks, and where can I find potential savings within our procurement?

These seemingly simple questions exposed fragmented landscapes, disconnected processes, and limited visibility into true cost drivers.

The first step on this journey was aligning and standardizing regional models to enable connection into a consolidated global framework. This was a critical move toward greater transparency and consistency, as there was a pressing need for deeper comparative insights across markets and time periods. As of 2024, we have achieved global visibility. Our client now can answer complex production cost questions at a highly granular level across dimensions such as product, brand, cost category, brewery, country, and time – comparing actuals, budgets, and forecasts with ease.

With this visibility, the next important questions arose quickly. Procurement and finance leaders began asking not just what changed, but why. What specific factors are driving shifts in production costs, and by how much? How can those insights be used to better anticipate and influence future outcomes?

These questions marked the next stage in the evolution: incorporating AI into the solution with predictive analytics and machine learning. In collaboration with our partner Predikt, we launched a pilot model for a single brewery in Belgium. The Predikt solution was implemented alongside existing planning models, complementing and enhancing them – not replacing them – to provide deeper business insight and enable more informed, data-driven decisions. The objective was to move beyond traditional planning, and move toward AI-powered forecasting.

Following a stepwise approach, the result was a dynamic, responsive model delivering accurate production cost forecasts and deep insights into why costs were shifting. With the perfect balance between leveraging ML algorithms and statistical predictive analytics, the Predikt platform has been able to identify the main drivers for the most important cost categories (raw materials, packaging materials, labor).

A defining advantage of Predikt is that it uses an AI language learning model to include the right external data sets. Research shows that up to 85% of forecast accuracy is driven by external indicators (e.g., commodity markets, inflation, energy prices), making their inclusion critical to increasing predictive power and forecast explainability. For instance, in the case of packaging materials, the model detected a lagged causal relationship between aluminum and metal market spot prices, and the cost of cans and lids. By using fitted regressions that demonstrate a high correlation, causation, and accuracy, we can conclude that, if for example the price of aluminum increases by 1% today, the magnitude of the effect will be on the total production cost of a beer a number of months from now. With this intelligence, our client can now anticipate the future cost impacts and take proactive measures in pricing strategy and cost control.