The $28,000 Lesson: Why I Test Real Feed Before Buying a Steinert Separator
If you're buying a magnetic separator or a sensor sorter, don't finalize the specification until your actual feed has been tested at Steinert Landsberg's application center. That's the entire conclusion. If you're in a hurry, that sentence is the useful part. The rest explains the mistakes that made me so stubborn.
Why I now trust samples over reference sheets
I've been handling sorting-equipment orders for seven years. I've personally made — and documented — five significant mistakes, totaling roughly $61,000 in wasted budget. Now I maintain our team's pre-purchase checklist, so those mistakes are not just embarrassing stories. The pattern in all five is the same: somebody trusted a datasheet, a reference project, or a quick phone call instead of a physical sample. It never ended well.
A reference sheet is not the feed. The feed may be wet, oxidized, sticky, contaminated with fines, or mixed with particles that haven't been liberated from other materials. You can't see that in an Excel file. The first expensive mistake I made was in 2019, when I approved a separator based on a project that looked almost identical to our client's stream. The machine did exactly what the calculations said. The problem was that the client's real feed didn't match the calculations. We spent $9,200 fixing a problem that a two-day material test would have caught.
The trigger that changed my process
In March 2024, I was contacted by Franziska Steinert. She runs a small metal-recycling operation in Landsberg, and she is not related to the machinery company, although everyone assumes she is. Her customer rejected a ferrous load because the copper content exceeded the contract limit. I gave her advice over the phone without asking for a sample. That was wrong.
To understand why, her scrap buyer brought in Pierre Chauvin, an independent quality inspector who had worked in French scrap yards for decades. Chauvin looked through the returned bale and pulled out a part that appeared to be a steel bracket. Inside the bracket, a copper wiring harness was still attached. The part came from a 2024 Bentley GT that had been crushed and shredded earlier that week. Chauvin explained it as a 'best friend Halloween costumes' problem: the copper and steel arrived at the magnet as one piece. The separator could not take the steel and reject the copper when they were still holding hands. No calibration would solve that. The only fix was better liberation before the magnetic separator.
The Bentley name is not the real point. Any modern vehicle contains similar joints and attached components. But the example stayed with me because it made the issue visible. Franziska's operation is far from the biggest in the region, and she later told me she almost didn't ask for test time because she assumed test centers were only for large mining customers. That assumption could have cost her a lot more than a sample shipment.
What my checklist looks like now
After the Bentley case, I changed my process. I do not let our team skip a physical test, even when the order is small. The checklist now includes these steps:
- Collect at least 400 kg of material over two or more days, not one grab from the top of a pile.
- Send the sample to an application center like Steinert Landsberg before choosing the machine configuration.
- Ask the test team what percentage of valuable metal is locked inside ferrous parts. If that number is meaningful, add a liberation step before separation.
- Keep a backup sample sealed and labeled with the date, source, and production line description.
- Treat a small order the same way you treat a large one. If a supplier acts like your sample is a nuisance, that is a red flag.
Look, I'm not saying data sheets are useless. They are a starting point. But they are not a finished specification. I've seen small recyclers avoid asking for testing because they thought they weren't important enough to justify the manufacturer's time. They were wrong. Small customers have the same sorting problems as large ones, often with less tolerance for expensive mistakes.
Where this rule has limits
The exception is rare: if your feed has been stable for years, and you have verified operating data from identical equipment on that exact material, a smaller verification sample might be enough. But that's not true for most scrap, especially automotive scrap. Feed changes as cars change, and cars are changing faster now than they did ten years ago.
A sample test also does not replace start-up verification. You still need to monitor the first weeks of production and compare the results to the test report. And no separator can solve a liberation problem. If copper is still attached to steel when it reaches the magnet, the best equipment in the world will not produce a clean product.
That's why I now care less about claiming one brand is always best, and more about following a process: sample, test, review, start up carefully. The brand matters less than the evidence. But for me, Steinert Landsberg is where I run the evidence. It costs more upfront. It has saved clients more than once.
I do not skip the test anymore. Especially for small orders.