Collo's deeptech fix for the food & beverage industry’s water waste problem

Despite efforts to reduce water consumption, the beverage industry is known for its resource-intensive nature: for example, producing each litre of Coca-Cola requires up to 1.8 litres of water, and in dairy, this ratio is twice as high. 

Both industries mainly rely on legacy sensor technology and measurement principles developed in the 1800s to distinguish between different liquids in the production process.

Finnish deeptech startup Collo wants to change this.

Developed after years of scientific research at Tampere University, the company has developed IoT analysers for optimising industrial liquid processes (and a corresponding platform), helping cut liquid losses in beverage and dairy production. Its tech is trusted by industry giants like Danone, Fonterra and Valio.

I spoke to CEO Jani Puroranta to learn more.

The three pain points that sparked Collo's RF innovation

According to Puroranta, the company spun out an original science project where some big companies had identified the need for new types of analysers in their processes. The major things they were struggling with were, first of all, fouling — the accumulation of unwanted material on solid surfaces. 

"In many processes, you get fouling, and on optical devices especially, it's a challenge because for one or two weeks it works nicely, but then it starts to drift as fouling accumulates.  Eventually, you need to either recalibrate or clean it, and you can't rely on those instruments all the time.

The second challenge was that instruments are often very specific to the location where they are used. In one place it works, but at the next station in the process, it may not work.

Then at the third stage, you need another analyser. It becomes a game of mixing and matching."

Then there's the need for predictive maintenance, rather than having a maintenance person go through the plant, check instrumentation, and follow a maintenance schedule.

Puroranta contends, "Wouldn't it be better if the instrument could tell you when it needs service? That was a challenge." When asked why such a critical solution hadn't already been developed by the giants of the food and beverage world, Puroranta contends that genuine breakthroughs like this rarely originate inside large corporations. 

"They typically come from universities, where researchers test unconventional ideas, like a completely new method for liquid measurement that falls outside the industry's standard playbook," he explained.

"Once a technology shows real promise and reaches industrial validation — around Technology Readiness Level 6 to 8 — that's when the big players start paying attention. That's where we are now: the technology is proven in live environments, we're shipping to customers, and continuously improving with each deployment."

The RF breakthrough lets factories see inside their pipes in real time

Collo uses radio-frequency (RF) sensing to analyse the behaviour and composition of liquids in real time. Instead of relying on traditional measurements like temperature, flow, or pH, Collo sends low-power RF signals through the liquid and measures how the signal changes as it interacts with the fluid. 

Different liquids – and changes within the same liquid – affect the RF signal in unique ways (a kind of "liquid fingerprint").  By interpreting these RF responses, Collo can detect transitions between products, levels of dilution, residue during cleaning, and other subtle changes that standard sensors can't capture.

This allows manufacturers to track what is happening inside pipes and tanks with high precision, enabling better optimisation of product pushouts, cleaning processes (CIP), and overall process efficiency.

A three-layer platform bringing "liquid intelligence" to the beverage and dairy industries

After the original project, the tech turned into a startup, and according to Puroranta, "Over the past years, it's become evident that this technology works particularly well in the food and beverage environment, tackling the challenges customers have there today. Dairy is a big vertical for us, beverage bottling plants as well."

Collo offers three core products that together deliver end-to-end "liquid intelligence" for industrial processors.

Collo Insights is the analytics and visualisation layer, turning complex liquid behaviour and sensor data into clear, actionable insights for operators to optimise processes, reduce losses, and monitor batch consistency in real time. 

Collo Connect integrates this intelligence directly into plant automation systems such as PLCs and SCADA, enabling precise, data-driven control of flow rates, transitions, and CIP sequences so processes can adjust automatically based on live liquid composition. 

Complementing these, the Collo Lab Analyser is a portable device for real-time liquid analysis in labs, pilot sites, or on the production floor, supporting R&D, quality control, and new product development through rapid testing and benchmarking.

With Collo's tech, the primary goal is automation. Its instrument sends an automation signal to the plant's SCADA system, which then determines when to turn valves to eliminate losses that can be identified in the process. The automation connection is key.

However, on the back of this, the company can also conduct cloud data analysis.

"Or if the customer doesn't want any cloud connection, we can collect the data on site and do the final analysis separately," explained Puroranta. 

"Then we walk the customer through their process: for example, if there's a sudden spike in the loss of raw milk somewhere, we can show that now it's going into the drain, and when it went into the drain, we can go back in the process and see which valve turned at the wrong time."

The hidden drain on profits: product changeovers and cleaning waste

The main applications for Collo's tech are so-called push-outs. 

"When you make product changes, you push the previous product out with water and then bring in the next product. The water needs to go to the drain, but sometimes you lose some of the product, or you may be too aggressive in the push-out, trying to save product, and end up diluting it instead. We help get that timing right," he explained.

Further, when workers clean the pipes with acids and caustic chemicals, they need to flush them and determine when it's the right time to start production again. 

Collo's tech can save huge amounts of water by dynamically determining when the pipe is truly clean and when you can move to the next stage. And the impact is profound.

Customers like Coca-Cola, which have publicly announced for years their targets for reducing water use, currently  use 1.8 litres of water for every litre of Coca-Cola they bottle. 

Additionally, when considering product losses in dairies, according to Puroranta, there are approximately 12,000 dairies in the EU, which process 160 million tonnes of raw milk annually:

"On average, 4 per cent of that gets lost. That ends up practically in the drain. That's almost €1 billion annually lost in the drain.

Additionally, there's an extra half billion in wastewater management costs, as it needs to be treated — it cannot simply be poured into the sewer. 

You need to add polymers, flocculants, and so on. This means big investments for water treatment plants. Just getting the extra milk out of the water is a big cost, and of course, there's the environmental impact as well."

Reducing the average from 4 per cent to 3 per cent — a 25 per cent decrease — results in more than a billion euros saved annually across the industry.

In other words, even incremental waste saving has a massive (excuse the pun) flow-on effect. 

Augmenting with smarter sensors and self-learning models

Crucially, Collo tech augments existing supply chain set-ups, reducing the cost of replacement.

Puroranta asserts, "We're not expecting customers to replace anything. Usually, they've gone through the process of identifying that they can't live with a no-parameter setup, which is usually time- and flow-based: just looking at the clock and turning the valve. That's wasteful. Then they add a conductivity instrument or some optical devices to monitor what's going on. With that, they get down from, say, 7 per cent loss to 5 per cent, maybe close to 4 per cent. But how do you get better than average? Then you need new technology. That's where we come in."

Collo's instrument stands out because it measures nine variables, including temperature for temperature compensation.  "How do you turn nine variables into a signal that automation can use? Because automation can only use a monotonic one-dimensional signal that goes up or down. It can't use nine-dimensional signals," shared Puroranta.

"For that, we use machine learning models. We have developed about a dozen different algorithms depending on the liquid type and the problem you're trying to address. Are you doing push-outs? Cleaning in place? Product quality?

That's also an aspect — we do raw milk quality fingerprinting.

We convert those nine variables with machine learning models tailored, if needed, to the customer process. They are trained on the actual data on the customer site, and we turn that into an automation signal."

"The algorithms are also self-learning," explained Puroranta.

"Sometimes it needs more work if it's a very special product.

For example, we have customers using it for other purposes like measuring the viscosity of resin. Then we need to do some lab analysis to correlate it with our data and refine the machine learning models."

In August last year, the startup raised €5 million.

  "There are 12,000 dairies in Europe. That's thousands of addressable customers," shared Puroranta.

Further, Collo's technology has numerous applications in liquid processes beyond the beverage industry, including oils, resins, and ceramics, as well as mining processes and mineral processing.