AI piles on new capabilities

Nestlé plant in Iowa uses Aveva’s Connect, historian and advanced analytics services to make fluffier, powdered-chocolate products

Jim Montague

Oct. 2, 2025

3 min read

To make fluffier Nestlé Quik and Ovaltine, the company’s plant in Waverly, Iowa, recently partnered with Aveva, and transferred production data to its Connect, Historian and Advanced Analytics services to predict quality parameters, recommend optimal setpoints, and subsequently save 10% on powder. This graph shows the accuracy of AI calculations trained to predict product density compared to actual density found in the company’s lab.

Key highlights:

The article showcases how Nestlé applied AI and machine learning to improve real-time process control, moving beyond manual operations to data-driven decision-making.
By using real-time predictions and setpoint recommendations, the project illustrates how AI tools can support or even replace traditional manual adjustments, reducing variability and improving consistency in production.
Despite advanced analytics and AI, human understanding is still crucial for developing, deploying and interpreting models.

One emerging force promising to enable analytics is artificial intelligence (AI) and its interactive accessories. However, as usual, it seems like some human understanding and logic is necessary to develop and apply them effectively.

For example, Nestlé recently partnered with Aveva on a lighthouse project to optimize production of its Nesquik and Ovaltine chocolate-powder products at its plant in Waverly, Iowa. These products undergo an agglomeration process during production that involves wetting and drying the powder until it achieves the right moisture and density, or optimal fluffiness. Operators previously ran the agglomeration process manually, but this sometimes caused unwanted variability in the powder’s moisture and density.

In addition, operators on Nestlé’s production line used to rely on manual samples to test powder moisture and density every 30 to 60 minutes. This procedure left operators with lagging indicators throughout each production run, which resulted in inconsistent power, as well as wasted product because the company had to put more product in each jar to reach a visually satisfying fill level.

“We wanted real-time predictions of the density and moisture of our product,” says Greg O’Brien, project owner and engineer at Nestlé. “We also wanted to give our operators manipulated variable suggestions, so they could make more consistent powder on their own.”

Consequently, Nestlé designed a solution that uses real-time production data and advanced analytics to generate setpoint recommendations that keep moisture and density levels within desired limits, and minimize variability to consistently produce lower-density, fluffier chocolate powder. The lighthouse team began by transferring production data from one of Waverly’s agglomeration towers to a cloud-computing service and Aveva’s Connect data services and historian software, where they could perform advanced analytics with AI and machine learning (ML).

Get your subscription to Control's tri-weekly newsletter.

Next, Nestlé developed a dashboard in Connect that visualizes predicted moisture and density, along with the setpoint recommendations, which are enabled by Aveva’s Advanced Analytics software. Operators can also view customized dashboards next to SCADA screens that show predicted and actual density and moisture levels (Figure 1). Also, the recommendations screen takes the guesswork out of process control because operators can simply follow Advanced Analytics’ guidance to ensure product consistency. In the future, the team expects to close the loop by feeding Advanced Analytics’ setpoint recommendations directly into Nestlé’s SCADA system to automate adjustments.

Figure 1: To make fluffier Nestlé Quik and Ovaltine, the company’s plant in Waverly, Iowa, recently partnered with Aveva, and transferred production data to its Connect, Historian and Advanced Analytics services to predict quality parameters, recommend optimal setpoints, and subsequently save 10% on powder. This graph shows the accuracy of AI calculations trained to predict product density compared to actual density found in the company’s lab.

More chocolatey yield, less waste

In addition, the lighthouse team trained no-code analytics models using historical process and lab data, establishing a correlation between the product’s quality parameters and process variables. These models automatically generate setpoint advice for operators to use in the plant when making equipment adjustments, and subsequently achieved several positive results. During the lighthouse project’s one-shift, eight-hour trial, the team saw more consistent powder produced with optimal moisture and density levels. At the end of the trial, the socks used to catch waste were mostly empty due to reduced fine-powder circulation and waste. Finally, Nestlé uses inline check weighers underneath conveyors to weigh every product as it passes along the line. The team found that powder from the trial had consistently higher moisture and lower density, eliminating the need to overfill jars to reach desired fill levels. In fact, the trial revealed that for every 10 1-kg jars produced, Nestlé saved up to one 1 kg of product—up to a 10% savings.

“Working with Aveva’s lighthouse team helped us understand how we can realize even more value in the future,” concludes O’Brian. “For instance, running in the cloud means we can easily roll out the analytical model to other plants across different regions without investing in added on-premises infrastructures.”