Liquats Vegetals makes its forecasts reliable thanks to AI with VOLTA Forecast

Liquats Vegetals banks on artificial intelligence with VOLTA Forecast to increase the reliability of its forecasts and streamline its S&OP process

Liquats Vegetals is a leading player in the field of plant-based drinks in Europe. Based in Catalonia, the company produces and markets the brands Yosoy, Monsoy, and Natrue, distributed in over 40 countries. Since 2017, Liquats has relied on the BEVOLTA platform to manage its production planning and the replenishment of its components and raw materials. This historic partnership has allowed for the consolidation of reliable and controlled management of production flows.

But as volumes increased and ranges expanded, the complexity of demand management became more noticeable. In 2021, Liquats decided to take a new step: to integrate artificial intelligence at the heart of its forecasts. On the occasion of the launch of VOLTA Forecast, the automatic forecasting module based on machine learning developed by BEVOLTA, the company agreed to become a pilot client.

Better anticipation without burdening the processes

By this stage, Liquats already had a solid planning foundation. Forecasts are built on historical data, manually enriched by planners and sales teams, following a proven business logic. But this process consumes time and valuable resources. And most importantly, it becomes difficult to evolve in an increasingly unstable market context: last-minute promotions, unusual weather conditions, emerging nutritional trends, trade-offs between channels...

The goal was not to call into question internal expertise but to rely on technology capable of processing a large number of variables, automatically generating reliable forecasts, and offering teams a more robust, dynamic, and exploitable starting point.

A collaborative and progressive project approach

The project was co-led by the supply chain teams, IT coordinators, and BEVOLTA consultants. It began with a functional framing phase aimed at defining target scopes (product families, geographic areas, update frequencies), the stakeholders involved, priority use cases, and data interfaces.

The first step involved connecting existing data from the JDE ERP: sales histories, stock levels, product sheets, life cycles. From this base, BEVOLTA teams supported Liquats in identifying and structuring explanatory variables useful for modeling, directly linked to their activity.

Explanatory variables adapted to Liquats' business

Besides the classic calendar effects and promotions, several elements proved particularly relevant in the context of Liquats:

• Local weather data, notably to anticipate the consumption of fresh or plant-based dairy drinks in hot regions.

• Customer location: some export customers (particularly in Asia or Northern Europe) present consumption profiles strongly linked to cultural or logistical contexts.

• Specific commercial campaigns, with volume engagement mechanisms that strongly impact the demand curve.

• Cannibalization between families, for example between almond, soy, or oat-based ranges, which sometimes respond to similar uses but evolve differently depending on the markets.

• Packaging changes, which can lead to an anticipation or delay effect in orders.

• Logistical variations, notably constraints on production or transport capacities, which indirectly influence expressed demand.

These variables are integrated into the forecasting engine via VOLTA Forecast. Thanks to the AutoML approach, the platform automatically selects the most efficient models, based on available data and the specifics of each product/channel/country combination.

A reliable foundation for management and dialogue

In less than three months, the first models were trained, tested, validated, and then deployed in production. Users now have access to three levels of probabilistic forecasts (P10, P50, P90), directly integrated into the BEVOLTA platform. These data feed the planning, simulation, and decision modules already used daily by supply teams.

Beyond gain in precision, it's the overall functioning of the S&OP process that has evolved. Exchanges between planners, sales, and management are no longer focused on manually producing figures, but on interpreting them, adjusting them, and their strategic impact. Meetings are shorter, better prepared, and allow for quicker engagement in trade-offs.

• Shorter, better-prepared meetings

• Faster and more informed decisions

• Enhanced alignment between supply, commerce, and management

A logic of alignment, time gains, and stronger decisions

The implementation of VOLTA Forecast has enabled Liquats to gain rigour, fluidity, and consistency in its decision-making process. The time freed up by forecast automation is reinvested in analysis, coordination, and building more relevant scenarios. The tool does not eliminate uncertainties, but it allows them to be addressed more methodically, and to align key players more quickly around realistic action plans.

Planners remain at the heart of the system, with an enhanced role: they are no longer number producers but facilitators of dialogue between data, industrial constraints, and commercial dynamics.

Towards an extension of forecast intelligence

Liquats is already studying the next axes of VOLTA Forecast extension: integration of e-commerce trends, forecasting raw material needs based on harvests, dynamic adjustment of supplies according to market signals. The collaboration with BEVOLTA continues in a long-term partnership logic, serving a more agile, resilient supply chain more connected to ground realities.