Five Claims About Production Waste That Don't Survive a Customer Audit

A manufacturer pays for its production waste four times and books it once. There's the resin or the coil stock or the corrugate you bought and didn't ship. There's the energy and labor you spent turning it into something unsellable. There's the handling to move it from the press to the compactor. And then there's the tip fee, which is the only one of the four that appears on a waste line in the ledger. So when a plant quotes me a cost per ton for production waste disposal, I ask which of the four it covers, and the answer is the tip fee.

That gap is why the claims below keep getting made. I hear them from plant managers, from corporate sustainability teams, and from the vendors selling into both. Each holds up fine internally and then breaks the first time a customer's supplier-audit team, or a state inspector, asks for the document underneath it.

"We divert 92 percent of our production waste"

Ask where the 92 came from and the trail usually ends at a monthly report from the hauler. That report is two things glued together: weights off a truck scale at your fence line, and a disposition code the hauler assigned to each load. The weights are good. The codes are a routing intention recorded at pickup, and your material changes hands two or three more times after that.

A baled stream leaves the compactor, lands at a transfer station, gets consolidated with similar bales from other generators, and is sold to whoever bids that month. The buyer might be a domestic reprocessor, an exporter, or a cement kiln. Any of them can re-route a load on arrival for moisture, contamination, or price, and nobody tells the generator, because the generator isn't a party to that transaction. Your broker agreement almost certainly obliges them to take title, and says nothing about reporting end use. So what does a disposition code describe once the material is two handoffs past your gate?

EPA's Toxics Release Inventory analysis has the same structure at national scale: facilities reported managing 34.6 billion pounds of listed chemicals as waste in 2023, with about 90 percent recycled, burned for energy, or treated, and 10 percent disposed or released (per EPA's TRI waste-management summary, published in August 2025 on 2023 reporting). That's a reported-management figure rather than a verified end-use figure. (And TRI covers only listed chemicals above reporting thresholds, so most plants' scrap by weight never enters it at all.)

An audit wants the end-use record: downstream facility name, permit number, and what happened to the material once it got there. Produce that for a single quarter and your number will hold. If you can't, you're publishing your hauler's estimate under your own logo.

"Our scrap numbers beat the national recycling rate"

They might. But the national rate describes a different stream. Per EPA's facts-and-figures tables, the United States generated 292.4 million tons of municipal solid waste in the 2018 data year and recycled or composted 32.1 percent of it, with roughly an eighth going to combustion with energy recovery and about half to landfill.

Municipal solid waste, as EPA defines it in those tables, excludes non-hazardous industrial waste, construction and demolition debris, and wastewater sludge. Your production waste sits in the excluded set. Benchmarking a stamping plant against a figure assembled from household trash tells you nothing about the stamping plant, and it's the single most common error I see in industrial scrap management reporting.

There's a harder problem underneath. No current federal inventory of non-hazardous industrial waste exists. EPA's own pages on industrial secondary materials put the volume at "hundreds of millions of tons" sent for disposal each year and leave it there, which is a range, not a measurement. The 6.9 billion metric ton total that still circulates at conferences traces back to a 1985 survey, and MIT's 2016 re-estimate found it wrong by close to an order of magnitude, largely because the old number had counted dilute process water sitting in surface impoundments rather than solid scrap.

Even the best-instrumented state can't close that gap. California collects quarterly submissions from recyclers, haulers, brokers and landfills, and CalRecycle's 2024 summary puts statewide generation at 77.2 million tons against 44.7 million tons landfilled. Those are facility-side figures. They tell you what crossed a scale in California, never what the injection molding sector sent there.

So there's no denominator, which leaves you comparing your own plant against itself last quarter, line by line.

"It's a byproduct, so it isn't a waste"

This one has teeth, because sometimes it's true. The legitimacy test for recycling sits in 40 CFR 260.43, and it asks four questions: does the material make a useful contribution, does the process yield a valuable product or intermediate, do the generator and the recycler manage the material as a valuable commodity, and are hazardous constituents present at levels comparable to the analogous virgin product.

Factor three is where I watch plants lose the argument. If an analogous raw material exists, the secondary material has to be managed at least as protectively as that raw material. Totes of off-spec polymer stacked uncovered behind building four, with rain standing in them, are not being managed the way anyone manages the virgin resin they paid for. An inspector doesn't need a lab to make that finding; they need a photograph.

Then there's the clock. Storage without a working recycling route becomes speculative accumulation, and the test for it is quantitative:

The amount of material that is recycled, or transferred to a different site for recycling, equals at least 75 percent by weight or volume of the amount of that material accumulated at the beginning of the period. Materials must be placed in a storage unit with a label indicating the first date that the material began to be accumulated.

That text is from 40 CFR 261.1, and two separate obligations hide inside it. You need a date on the storage unit (or an inventory log, where labeling isn't practicable), and you need three quarters of your opening inventory to move within the calendar year. A pile that stops moving because the offtake price went sideways fails the second obligation no matter how good the intent behind it was.

A scope limit, because this is the point where consultants overreach: those four factors and the speculative accumulation test live in the hazardous waste rules, and most production scrap isn't a hazardous secondary material. States run their own programs, several stricter than the federal floor, and a state beneficial-use determination is usually what converts a scrap stream into a legitimate product on paper. Get that determination before you build a process waste minimization business case on it, not after.

"A better sorter will fix our scrap problem"

Production scrap is the easiest classification problem in this industry. One material most of the time, known geometry, known color, arriving from a known tool at a known rate. Build a confusion matrix on a belt like that and the off-diagonal cells come back nearly empty, which tells you the model was never the bottleneck. Most 'AI waste sorting' sold into a scrap line is a rule-based pipeline with a CNN bolted on, and here that rule engine is the right architecture under the wrong name.

Compare that to mixed municipal input. On a Tomra autosort retrofit I worked at a 600 TPD materials recovery facility in 2022, the PET model held 94 percent recall through a dry week and gave up more than twenty points of it once input moisture climbed past 18 percent [commissioning data]. Nothing in a production scrap stream behaves that way. Which is why sorter capacity bought for a scrap line is usually capacity bought against somebody else's problem.

The hard problem is attribution. Scrap reaches the compactor already mixed across sources, so the question "which of our forty presses makes sixty percent of this" has no answer anywhere in the weight data. What you need is a join key: every scrap event tagged to line, shift, tool, operator and part number, at the moment it's generated. That's a counter and a barcode reader at the drop point. It costs less than any sorter and is far harder to get approved, because the first week of data makes somebody's line look bad.

Actually, that holds down to a limit. Run short lots with frequent material changeovers and the scrap stream starts to resemble mixed waste, at which point a vision model does begin to pay for itself. The same ordering applies upstream, where waste feedstock quality fails on logistics well before it fails on sorting. Measurement placement beats model architecture in both cases, and manufacturing waste reduction programs that start with instrumentation rather than equipment tend to find their savings in scheduling and tooling instead.

"Zero waste to landfill means nothing of ours reaches a landfill"

The phrase is a certification claim, and the certification has tiers. UL's landfill waste diversion validations band Silver at 90 to 94 percent diversion, Gold at 95 to 99 percent, and Platinum at 100 percent, validated site by site rather than across a company [UL 2799 service description]. Two plants can therefore both publish "zero waste to landfill" while one of them sends nearly a tenth of its output somewhere else entirely.

Where it goes is usually combustion, and UL's own published material is inconsistent about how that counts. The 2799A validation is described as requiring at least 90 percent diversion by methods other than waste to energy, while the tier descriptions elsewhere treat energy recovery as diversion. So read the validation letter rather than the press release, and check which revision of the standard it cites.

Energy recovery is a legitimate destination for production waste that has no viable recycling route, and for some streams it's the right one. Mixed-polymer film and contaminated absorbents don't have a reprocessing market at any price I've seen quoted. But it doesn't close the accounting. Combustion leaves bottom ash and fly ash behind, and ash turns out to be an offtake problem rather than a disposal line for whichever facility receives your material. If their ash goes to a monofill, part of your tonnage reached a landfill with an extra step inserted in front of it. Residue fractions differ sharply by route; pyrolysis systems and mass-burn lines leave very different quantities behind to place.

Much of the search traffic on this phrase is somebody checking a named operator's claim: whether that company runs zero waste to landfill, or what its landfill diversion figure includes. The answer lives in the validation letter, in which standard and revision were applied, which tier was awarded, which sites were in scope, and whether energy recovery counted toward diversion.

None of this is within reach for every plant. Below roughly fifty tons a month of total scrap you have no leverage with a hauler and no second bidder to call, so the disposition code is the only record that will ever exist. Single-compactor sites can't segregate finely enough for end-use documentation to mean much. And in states with no beneficial-use determination pathway, a legitimate secondary material can sit classified as waste for years while the paperwork catches up. In that position the better move is to stop publishing a precise diversion percentage at all, rather than defending one you can't support.

So the first concrete step, before any number goes into a customer deck or a sustainability annex, is a letter to your hauler asking for the name, permit number and state of every downstream facility that received your material last quarter, along with the final disposition of each load by weight. Give them thirty days to answer. Most haulers can produce it for the recyclable fraction and will stall on the residual, and that gap is the shape of your zero-waste-to-landfill program as it exists today, which is a far better thing to learn from your own hauler than from a customer's auditor.

Sources & Notes

• The 34.6 billion pound figure and the 90/10 split between managed and released waste come from EPA's TRI National Analysis waste-management section, published August 2025 on 2023 reporting year data.
• Generation, recycling and disposal tonnages, plus the definition that excludes industrial and C&D waste, are in EPA's materials and waste figures. Note the data year is 2018; EPA also changed its food-waste methodology that cycle, so the series isn't comparable with earlier years.
• For the legitimacy factors I used the text of 40 CFR 260.43 as published by Cornell's Legal Information Institute. The speculative accumulation language quoted above sits in 40 CFR 261.1, and I'm summarizing regulation here rather than giving legal advice.
• Tier bands and the waste-to-energy wording discrepancy: UL's landfill waste diversion claim validation pages. The inconsistency between the 2799A description and the tier bands is UL's own, which is why I'd read the letter.
• The 1985 provenance of the 6.9 billion metric ton industrial total, and the re-estimate putting it in the hundreds of millions, come from an MIT thesis on accounting for non-hazardous industrial waste. California statewide tonnages are from CalRecycle's state of disposal and recycling summary for 2024.

Disclosure: I build computer-vision and sensor systems for waste and recycling operations at Renewable Waste Energy. The 2022 sorter retrofit referenced above is anonymized at the operator's request, which is why the facility isn't named.