How a simple test boosted KSI Recycling’s recovery by 10%

Process Engineer Tjerk Wiersma is one of the people responsible for ensuring that KSI’s recovery rates continue to grow. With a small team, improving recovery means testing innovative ways to maximise the performance of existing machinery and human resources.

Before the arrival of AI, experimentation was a time-consuming process based on historical data:

Before we deployed Analyzer units, we would run an experiment maybe once a week — then wait one to three weeks before the quality team had the results for us. We couldn’t run multiple tests at a time, because it would be too much material for the team to sample."

Reliance on manual sampling delayed testing and iterative improvement, often forcing staff to rely on intuition when making process adjustments. Automated analysis filled that vital data gap, unlocking rapid testing and improvement.

Mandatory sampling is still conducted by KSI’s experienced staff, but AI waste analysis has enabled Tjerk to automate the non-mandatory tests that maximise facility performance:

Now, we can do several tests in a single shift. Instead of waiting for results or overloading our team, we get live data."

Tests quickly led to results that changed the way KSI’s team operate. Faster process improvements based on live material data have already translated to business-critical improvements for KSI.

How a single test unlocked an insight that boosted compliance — and yield

The data gap

Tjerk and his team knew that cleaning certain sorting machinery would improve material separation. That is especially important on heavily-regulated residue lines: whenever residue material contains over 25% recyclable materials (excluding fibre), KSI faces penalties.

Tjerk didn’t know how much cleaning would improve would improve performance, though — or how often they could clean machinery without impacting facility productivity:

We knew that performance would drop as material became more dirty — but not by how much, nor what time it was going to be dirty enough to have an effect.

We also didn’t know what kind of materials were making up the residue, so it was hard to tell whether we were over the 75% threshold for residue and fibre material before it was too late."

AI waste analytics had the answer.

The test

To find out whether it was worth briefly pausing operations for cleaning, Tjerk conducted a test. During the morning shift, he instructed his team not to clean sorting machinery:

What you are seeing here is the composition of our residue line during the morning shift, in the same day. The pink and the grey colours at the bottom are what we want to see — they need to be above 75% for us to avoid penalties.

In the morning shift, I asked my team not to clean our machinery, and you can see the amount of plastics — like the blue PET and PP layers – increasing."

On the same day, he instructed his team to clean the machinery at the point quality had dropped earlier on:

This chart shows the afternoon shift, after the team cleaned our machinery.

You can then see the effect of that cleaning, almost immediately. The difference was clear. We're now reaching the above 75% target"

The results

For Tjerk, the results were significant. Even if sorting was briefly interrupted, it made better business sense to clean machinery and avoid penalties while recovering more material:

Cleaning helped us maintain higher performance, and ensured less valuable material ended up in residue.

Before this, we were losing around 30% of our valuable commodities. Now that we’re optimising based on data, we’re losing just 20% — and our residue composition is back within regulatory thresholds."

That shift — the result of a single process change — means that KSI now avoid penalties, and recover more valuable plastics in their mission to return materials back into the circular economy.

How AI led to a data-first, human-centric mindset

The cleaning test was just one of an ongoing process of adjustment, measurement, and improvement. Data-driven changes are now part of regular operations at KSI:

Residue made a major 10% improvement, but Analyzer’s impact is really about the fact that we can control our whole process better.

We adjust NIR parameters or blast separator angles and measure the immediate effect on performance. They’re very simple changes, but we didn’t know their specific impact before adopting AI."

KSI is known for its tight-knit team, and Tjerk has empowered his colleagues by giving them the same access to transformational performance data. They shared his natural curiosity:

Before we had this technology, shift leaders had to make decisions on intuition and past experience. Now, they have the power to make those decisions with facts.

We provided the data dashboards, and then they came up with their own questions about it. Now, they use that data to adjust machinery, and allocate staff where they are needed most."

Like KSI’s ambitious vision of a future without waste, Tjerk has high hopes for the potential of waste intelligence technology. He envisions a system guided by experienced staff, with AI waste analytics making suggestions about process adjustments — or even making those adjustments automatically:

This technology will soon be able to make some of the decisions that an operator does. But we use AI to help make the people we already work with even better, not just automate the picking process. AI-powered robots can’t clean equipment."