October 2019: The PO I Shouldn't Have Signed
I still have the purchase order in my filing cabinet. October 14, 2019. A used magnetic separation line, $11,500, delivered to our yard in three weeks.
I approved it without testing a single kilogram of our actual material. That's the whole mistake, really. Everything else is just the bill coming due.
I've been handling recycling equipment purchases for about six years. I've personally made — and documented — four significant buying mistakes, totaling roughly $31,000 in wasted budget. This one was the most expensive, and the most avoidable.
I thought I knew our material flow. Iron was iron, right? A magnet was a magnet. The seller had years of experience, a clean shop floor, and a spec sheet that looked professional. I checked the price against three other quotes, confirmed the warranty terms, and signed.
In my first year (2014), I made the classic beginner's error of trusting a supplier's sample instead of running our own material. This time, I made a more expensive version of the same mistake: I trusted a spec sheet.
What Went Wrong (and How Much It Cost)
Here's the thing about magnetic separation: it works beautifully on ferrous metals. Our material — mixed electronics from small appliance recycling — had plenty of ferrous content, sure. But it also had aluminum, copper, and a bunch of non-magnetic junk that wrapped around everything.
Within two weeks, we had three problems:
- Iron recovery was closer to 75% than the promised 95% (the seller's “95%” was based on their sample, not ours — I should've caught that).
- Aluminum losses were brutal. The magnet was pulling ferrous fines into the aluminum stream, and our downstream buyer rejected two loads.
- We were paying overtime to manually re-sort what the machine should've handled.
Total damage: about $14,000 between rejected loads, overtime, and the eventual resale of the separator at a loss. Plus a very awkward quarterly review where I explained to ownership why our “efficiency upgrade” had made things worse.
The worst part? A colleague asked why we hadn't just tested the material first. She'd worked with sensor-based sorting before and said she always sent samples to the manufacturer's test facility. I didn't know that was an option.
The Trip to Ravensburg
That's how I ended up on a plane to Germany in February 2020, carrying 40 kilograms of material in two checked suitcases (which, honestly, got me some looks at security).
The destination was Steinert's Optik test center in Ravensburg — the site where they run optical and sensor-based sorting trials on real material. I'd reached out after my colleague's suggestion, expecting to be brushed off. We were a small operation. Our annual purchases were nowhere near what a major mining company would spend. I figured we'd get a brochure and a polite “minimum order” email.
Instead, I got a calendar invite and a request for material samples.
In Ravensburg, I met Trevor, an application engineer who'd spent the morning setting up a test line for our material, and Alexander, who ran us through the process. They didn't start with a sales pitch. They started with questions: What's your target purity? What's your monthly volume? What does your downstream buyer actually reject?
Then they ran the tests. It took about three days. We learned that pure magnetic separation was never going to work for our material mix. We needed sensor-based sorting — specifically, near-infrared and induction sorting — to distinguish between different non-ferrous metals.
Trevor mentioned at one point that the testing culture there traced back to Dr. Steinert's original focus on solving specific material problems rather than selling generic machines. I don't know if that's exactly the history, but it matched what I saw: nobody tried to sell me anything until they'd shown me data from my own material.
I'd spent nearly two years assuming “magnetic separator” was the answer. The conventional wisdom in our local recycling network was that magnets were the workhorse and everything else was a luxury. My experience in Ravensburg suggested otherwise: for complex mixed material, a combined approach was the only thing that actually worked.
What the Test Revealed
The numbers from those three days were uncomfortable to read next to my 2019 purchase order. On our material, a standalone magnetic separator achieved about 78% ferrous recovery and lost a significant fraction of aluminum to the ferrous stream. The sensor-based setup they tested achieved 93% ferrous recovery and reduced aluminum losses to under 3%.
I'm not going to pretend that meant we immediately bought a new system. We didn't. The price was more than double what we'd paid for the used separator, and I had to rebuild trust internally after the first disaster. It took another eight months of budgeting and a phased implementation before we ordered.
But the visit changed how I think about equipment purchases. Specifically:
- Test the actual material, not the spec sheet. The $11,500 separator wasn't “bad” — it was just wrong for our material. That distinction matters, and it only shows up in testing.
- Small volume doesn't mean small attention. Our 40 kg test batch was probably the smallest thing Steinert's Ravensburg team worked on that month. They treated it like it mattered. When I was starting out, the vendors who took my $200 orders seriously are the ones I still call for $20,000 orders. That pattern held here.
- Ask for the test, even if you're embarrassed. I thought asking for a trial would look naive. It looked like due diligence. The only naive thing was buying without one.
The Checklist I Now Use
I keep a one-page pre-purchase checklist taped inside my desk drawer. It's been there since 2021. Here's the short version:
- Have we run our actual material through the proposed process (not a sample, not a video, our material)?
- What's the target purity, and can the vendor show test data from material similar to ours?
- What happens to the non-target material? (The first separator “worked” on ferrous and wrecked everything else.)
- Is the quoted throughput based on our material density and moisture, or a generic figure?
- What's the smallest test batch the vendor will accept? (If the answer is “minimum 10 tons,” ask why.)
We've caught at least four questionable equipment choices with this list over the past three years — including a near-miss on a conveyor system that would've been undersized for our summer volume.
One more thing, and this is the part I'd underline twice: the trip to Ravensburg didn't just teach me about sensor sorting. It taught me that testing isn't a privilege reserved for big operations. When I emailed Steinert, I mentioned our volume in the first paragraph. Nobody flinched. The follow-up questions were all about the material, not the purchase size.
That matters more than a spec sheet.
One Caveat
This was my experience with one material stream in one facility. We're a mid-size electronics recycler with fairly consistent input. If you're processing construction debris or mining tailings, the material behavior will be different — test accordingly. Also, my cost figures are from 2019–2021; equipment pricing and sensor technology have changed since then, so verify current numbers before budgeting.
And if you're wondering whether it's worth flying your material to a test center: it cost me about $800 in travel and three days of time. The mistake it would've prevented cost $14,000. You can do that math yourself.
One last thing — we did eventually set up a camera above the sorting line to catch jams. The footage is strictly for training (not to make a John Wick-style action reel, though some of the material crashes do look dramatic). If you're going to invest in monitoring, keep it boring and useful.