Waste Feedstock Quality Fails on Logistics Before It Fails on Sorting

The loss arrives as a line on somebody else's invoice. Your fuel reaches the buyer, their lab meters it, and a few months into the offtake the loads start coming back repriced as waste rather than fuel. Sometimes they come back on a truck. You've already paid to collect that material, shred it, screen it and bale it, so now you're paying a second gate fee to get rid of it, and the yield line in your model never recovers. That bill is what waste feedstock quality means on a live project, and almost none of it gets decided in the sorting hall.

Walk the bill backwards and you usually land on two things that rode in on the collection truck: water and chlorine. I've commissioned three conversion facilities and advised on market entry in two regions where the technology was never the thing that went wrong. Feedstock quality fails the way regional market entry fails, on logistics and offtake credit rather than on equipment choice. So here's the sequence I now walk, in the order the money moves.

Step one: the fuel spec is three numbers, and the standard behind it sets no limits

Solid recovered fuel is classified under ISO 21640:2021 on three parameters: net calorific value, chlorine and mercury. Five classes each, combined into a three-digit code, so a given fuel might be a 3-1-1. That code is what most RDF feedstock specification sheets lead with, and it tells a buyer less than developers assume.

The classification sets no limit values. It isn't a certification, it doesn't rank fuels as good or poor, and it has nothing to say about which fuel a given plant may burn. That sits with the permitting authority and with whatever the offtake contract says, as ZKG's walkthrough of the scheme spells out for the cement sector. A permit in hand beats a signed MOU every time, and a class code is not a limit, and binds nobody. Two of the parameters that decide your yield aren't in the classification at all.

Class limits are from ISO 21640:2021 Table 2 and the buyer requirements from Measurlabs' guide to SRF testing. Read the right-hand column again. Moisture and ash are the two things that move week to week, they're the two things your collection system controls, and neither one appears in the code everybody quotes.

Step two: water is yield you already paid to haul

Water does two jobs on a load, both of them against you. It displaces dry mass you could have burned, and then it charges you again on the way out, because boiling it off takes heat that would otherwise have made steam. Calorific value drops twice over, and the second drop is the one feasibility models leave out.

According to a 2026 study in the International Journal of Heat and Technology, RDF taken from roughly 24 percent moisture to 43 lost more than a quarter of its gross calorific value in a bubbling fluidized bed co-firing with coal. Combustion efficiency fell from about 93 percent to 73 across those same two tests, and per the authors the net plant heat rate rose 27.5 percent. That was one run per condition at lab scale, so treat the slope as a direction and not a coefficient. The direction is steep enough to reorder a capital plan.

Most of that water is organic fraction plus rain on the tipping floor. According to EPA's national figures for 2018, food was 21.6 percent of US municipal solid waste generation at 63.1 million tons, and food was also the single largest category inside the 34.6 million tons combusted with energy recovery. Food scraps run wet, and anything stored outdoors through a wet month runs wetter.

So the fix is mostly civil works and scheduling, which is exactly why it never makes the technology shortlist. Covered reception. A tipping floor you can sweep dry. Route timing that keeps wet commercial organics out of the fuel line on the days you're baling for export. On the Caribbean island RDF plant I commissioned in 2022 we came in 9 percent over budget, and most of that overrun was a roofed reception bay and a second screen that weren't in the feasibility estimate (we had priced the plant off a data room and a dry-season site visit). The site visit tells you what the data room hides, but only if you make it in the rain.

Step three: chlorine is a refusal, not a discount

Chlorine behaves nothing like moisture at the buyer's end. Wet fuel gets repriced; chlorine-heavy fuel gets turned away, because the damage lands on the buyer's own asset rather than on their fuel bill. In a cement kiln, chloride cycles in the preheater and pushes bypass rates up, which costs the operator clinker. On a grate boiler it goes after superheater tubes and takes the outage with it. No buyer carries that for a discount.

The sources are ordinary: PVC, some packaging films, a few textile and cable fractions. A near-infrared sorter, a Tomra unit or equivalent, will pull PVC out of a shredded flow, but it reads whatever surface passes under the lamp, and the film and composite fractions that carry chlorine into your bales are often the ones it was never configured to eject. Chlorine is also lumpy in the stream, which makes it a sampling problem before it's a sorting problem. One bad bale in twenty can put a composite sample over the line.

For conversion routes that make a liquid product, the tolerance is tighter again. Chlorine that would merely corrode a boiler leaves a reactor as organochlorides in the oil, and the refinery buying that oil writes its chloride spec in parts per million. Feedstock contamination in pyrolysis hits product quality well before it hits yield, which is why serious pyrolysis systems carry front-end separation that mass burn never needed. Same molecule, different failure mode, much shorter fuse.

Step four is where most projects leak money: the sampling plan

By the time a buyer's lab result lands, the fuel is theirs and the only thing left to you is an argument. Most RDF contracts I've read specify composite sampling at the receiving end, at a frequency that suits the buyer's lab budget. Which means your yield is being characterized by someone else's sample, taken after the load has moved, against a stream that varies by day of week. At that point waste feedstock quality has stopped being an engineering question and become a measurement one. ISO 21645 covers how to take that sample properly. It says nothing about who takes it, or how often, and those two blanks are where the margin goes.

Better to meter upstream, on your own belt, continuously, so you argue from your own record. Microwave or near-infrared moisture on the fuel line, a chlorine proxy from the sorting telemetry (imperfect, but it trends), and every reading keyed to the inbound route and the date, so you can see the wet Tuesday before your buyer can. Continuous feedstock telemetry of that kind is what a platform such as Optimal Waste Intelligence is built around, and it earns its keep in contract arguments long before it earns anything in yield. I've argued elsewhere that a conversion plant is only as good as the feedstock nobody measured, and this is the cheapest place on the site to stop being that plant.

I used to put the drying question after the sorting question in feasibility work. Actually it belongs first, because every throughput rating on a sorting line assumes a moisture band that nobody wrote into the specification, and a wet month quietly derates the whole front end.

Where the logistics argument stops holding

Below a certain scale, none of this pays. A plant taking a couple of hundred tonnes a week can't amortize a roofed reception bay and a continuous moisture meter against the yield it would recover, and the better move there is to buy a tighter fuel rather than manufacture one. In a wet tropical market with no real dry season, covered storage and biodrying only take you part of the way; you size the boiler for wet fuel from day one and stop pretending the annual average means anything. Where the stream is heavily organic and unsorted at source, no amount of back-end separation substitutes for a separate collection, and the landfill diversion number you report to the municipality stays flat even as your waste feedstock quality improves.

There's also a ceiling on the whole exercise. EIA's explainer on waste-to-energy puts the burnable share of US municipal waste at about 85 pounds in every 100, with 2,000 pounds of garbage leaving 300 to 600 pounds of ash. You are optimizing inside that band, not escaping it. And if the offtake has no creditworthy buyer at the price your model needs, better fuel changes nothing; I've killed a project at that stage myself and would do it again.

Does any of this hold on a mass-burn grate with a large furnace volume and a reciprocating feed? Partly. A big grate tolerates more water than a fluidized bed does, and the operators I talk to usually describe moisture as a steam-output problem rather than a burnout problem. But the economics land in the same place, because the tonnage you paid to haul is the tonnage you have to pay to dry.

So before anybody prices a drier or a second optical sorter, pull twelve months of weighbridge tickets and every receipt analysis your buyer has sent, then plot delivered moisture by week against rainfall and against which collection routes fed the pile that week. On the sites where I've done this, the variance tracks the route and the weather more than it tracks anything inside the process, and that single chart tells you whether the next dollar belongs in civil works or in separation. It costs a week of somebody's time, and it usually reorders the capital plan before the first vendor quote arrives.

Sources & Notes

• The class limits in the table come from ISO 21640:2021, Table 2, as reproduced in Measurlabs' SRF testing guide, which also lists the application-specific moisture, ash and calorific value ranges that lime kilns and coal plants ask for. The standard itself sits behind a paywall.
• On what the classification does and doesn't do, see ZKG's article on SRF specifications and classes. It is explicit that the scheme sets no limit values and does not decide which plant may burn what; permits do.
• Moisture and combustion numbers: Putrawan et al., "Moisture Content Effects on Coal-RDF Co-Firing Performance in a Bubbling Fluidized Bed Incinerator," International Journal of Heat and Technology, Vol. 44 No. 4 (2026). Lab scale, one combustion test per moisture condition, blend prepared on a wet basis.
• Composition figures are EPA's, from the national overview of materials, wastes and recycling, data year 2018. EPA changed its food measurement method that year, so don't compare the food share straight back to older editions.
• Burnable share and ash residue are from EIA's waste-to-energy explainer. Those are US fleet averages and will move with local composition.