From 82% to 61%: Where a Landfill Diversion Rate Goes Under Audit

A landfill diversion rate is a fraction, and a fraction gives you two independent places to lie. You can inflate the numerator by calling more of the stream "recovered," or you can shrink the denominator by never weighing what you quietly bury. Either move lifts the percentage without one extra ton leaving the site. Most of the landfill diversion rate audits I've run come down to proving which of the two happened, and in what proportion.

In 2022 we audited a client's 600 TPD materials recovery facility whose annual filing reported a landfill diversion rate of 82% (operator data). Their gate scale told a different story. Once we backed out the green material they were shipping off as alternative daily cover, reweighed the construction-debris line that was padding the recovered column, and reconciled baler output against what the sorters had actually tagged recoverable, the defensible number collapsed to 61%. Twenty-one points, and almost none of it fraud. Every reclassification was arguably legal under the accounting they'd chosen.

That gap is the subject here: not whether operators cheat, but how an honest-looking number gets assembled, and which pieces of the assembly an audit can check. And the pressure to assemble a flattering one keeps climbing. Landfill tipping fees averaged about $62 a ton in 2024, up roughly ten percent in a year, per EREF, so every ton you can book as diverted instead of disposed is money saved and a cleaner sustainability report in the same stroke.

The denominator you never see

Waste diversion rate calculation looks trivial: diverted tonnage over total generation, times 100. The trap is that only one of those two numbers gets weighed. Recovered material crosses a scale on its way to a baler or a buyer, and residue crosses the gate scale on its way to the tip face. Generation, the denominator, usually isn't measured at all. For a jurisdiction it's modeled from population and a per-capita generation factor; for a single facility it's often just recovered plus disposed, which quietly assumes you captured everything that came through the door.

Anyone who has scored a classifier knows the trap. Precision is easy; recall is where the model lies to you, and a diversion rate is pure recall: recovered over everything you decided was recoverable. It looks wonderful right until you check what you quietly dropped from the recoverable set. Shave the denominator, or model it low, and the rate climbs while the trucks to the landfill run exactly as full as they did yesterday.

The jurisdiction version is subtler. Plug in a per-capita generation factor from a boom year, keep it while consumption softens, and generation looks larger than it is, which flatters diversion without a single program change. That's precisely the estimation layer California decided it could no longer defend, and I'll come back to what it swapped in.

Calculating waste diversion so the number survives scrutiny comes down to weighing both flows. Put recovered product and residue-to-landfill across calibrated scales, and treat any modeled generation figure as a hypothesis rather than a measurement. The national aggregates are at least anchored in disposal reporting: of the 292 million tons of municipal solid waste generated in 2018, roughly half was landfilled and about a third recycled or composted, per EPA's national Facts and Figures data. A single site's self-reported rate carries far more slack than that, because nobody outside the fence is weighing its denominator.

Alternative daily cover, and other legal reclassifications

Every landfill has to cover its working face at day's end to control odor, disease vectors, and blowing litter. Regulators let operators use processed material instead of clean soil for that job: shredded green waste, ground construction debris, auto-shredder residue, treated sludge, contaminated soil. It's called alternative daily cover, and for years the green-waste version got counted as diverted even though every ton of it ends up spread inside the landfill.

Both framings are honestly true, which is what makes ADC the cleanest reclassification in the business. The material does a job you'd otherwise pay for with virgin soil. It's also, physically, in the landfill. The accounting has to pick one story, and California's picked the flattering one for years. That changed under AB 1594: as of January 1, 2020, green material used as alternative daily cover no longer counts as diversion and instead registers as disposal against a jurisdiction's per-capita target, according to CalRecycle. Other cover materials still count, and green-waste erosion control at the site still reads as beneficial use. One narrow loophole, closed.

Weight loading is the quieter cousin. Report diversion by mass and the heavy inert streams dominate the fraction: concrete, aggregate, glass, soil. Recover the concrete off a demolition job and a weight-based rate leaps, while the light, high-value, genuinely hard streams (mixed films, laminated packaging, small-format plastics) can still ride the belt straight to residue. A facility can post a 90% number by mass and landfill most of its packaging by count. That isn't a lie; it's a unit of measure doing exactly what you picked it to do. A claim of 100% zero waste to landfill earns the same question every time: by weight of what, and with which streams admitted to the denominator?

Construction and demolition debris deserves its own footnote, because a lot of it never gets weighed stream by stream. Mixed C&D sent to a processing facility is often credited back to the generating jurisdiction using a facility-wide recovery rate, an average rather than a measurement of that specific load. Average in, average out: your project's reported recovery is really the plant's blended number wearing your name. That's usually close enough, right up until the plant's residual creeps and nobody re-runs the average.

The numerator lies too: recovered versus residual

Shrinking the denominator gets the attention, but I've watched just as many points hide in the numerator, right at the boundary between recovered and residual. When we built a waste-classification dataset in 2023, 38,000 labeled frames pulled off a live sort line, the clean cases were easy. Clean PET, clean fiber, obvious residue. The fight was the contaminated-but-arguably-recoverable middle: a bottle wearing a full sleeve label, fiber with a grease shadow, a tray that's the right polymer in the wrong color. Where you draw that line is where the recovered column is really set, long before anyone computes a percentage.

On the sort line itself the same slack shows up physically. That same plant was our 2022 optical-sorter retrofit at a 600 TPD MRF, and for the better part of a wet-season shift the Tomra units lost more than twenty points of PET recall after input moisture spiked past 18% (wet fiber smears the near-infrared signature, so the ejectors miss). The belt was recovering far less than the nameplate implied, and the reported diversion never reflected the gap. If your number leans on rated equipment performance instead of weighed output, you're booking recall the machine didn't deliver. Label quality sets the ceiling on what a classifier can see; feedstock condition sets the ceiling on what the hardware can act on.

There's a second numerator leak downstream of your gate. You book a bale as recovered when it crosses your scale, but the mill that buys it pays on accepted tonnage, after it strips out the outthrows and prohibitives. A contaminated fiber bale can lose ten to fifteen points of yield at the receiving end (worse if it caught rain in transit), and that loss never shows up in your diversion rate, because your number closed the moment the bale left the building. If you want the honest figure, reconcile against buyer settlement weights, not baler counts. I've had that one reconciliation move a reported rate by four or five points by itself.

What TRUE zero waste certification actually audits

A third-party audit closes most of this, which a self-reported figure can't. TRUE zero waste certification, run by GBCI, sets the bar at 90% or greater diversion from landfill, incineration, and the environment, measured across the most recent twelve months of real records. Two design choices in that sentence carry the weight. It's judged against a full year of data, not a good quarter. And incineration counts as disposal, not diversion, so you can't reach the threshold by routing residue to a combustor.

That second choice is stricter than many operators expect. Sending a stream to energy recovery is a legitimate destination, and our own waste-to-energy services exist precisely because some materials have no better home. But it isn't landfill diversion in the TRUE sense; it's thermal disposal with heat recovery. Conflating the two is one of the most common ways a diversion rate gets inflated on paper, and a certification that shuts the loophole is worth more than one that leaves it open.

Measure disposal, not diversion

The fix I keep coming back to isn't a better diversion formula. It's measuring the thing you can actually weigh. California reached the same conclusion at the policy level almost twenty years ago. Under SB 1016, starting with the 2007 reporting year, the state stopped calculating jurisdiction diversion rates and switched to per-capita disposal, pounds landfilled per person per day, drawn straight from facility disposal reports, per CalRecycle. Disposal crosses a scale. You can't inflate it by redefining what counts as recovered, because you aren't counting recovery at all.

Actually, that's not the whole story, and per-capita disposal carries known limits of its own. One is a failure mode in the other direction: in a recession, generation falls, disposal falls with it, and a jurisdiction looks like it diverted more when it simply consumed less. The other is scale. This approach breaks down below the point where a weighbridge pays for itself, so a small rural transfer station handling a few thousand tons a year ends up estimating disposal with a fixed volume-to-weight factor, its tonnage a modeled number wearing a measurement's clothes. The metric is weakest exactly where data quality is poorest. It's still harder to game on purpose than a diversion fraction, though, because the number you'd need to move sits on a scale somebody else controls.

When I audit a landfill diversion rate now, I don't argue with the percentage on the cover page. I pull the inputs and rebuild it:

1. Gate-scale tickets for every outbound stream, recovered and residue both, across a full twelve months rather than a sampled month.
2. The ADC and beneficial-use log, so any green waste or shredded debris booked as cover can be moved back to the disposal side when that's where it belongs.
3. A recent composition study of the recovered streams, to catch contamination riding along in the recovered column.
4. The denominator's provenance: whether generation was weighed, or modeled from a per-capita factor that happens to flatter the result.

None of that needs exotic instrumentation. It needs calibrated scales, an honest cover log, a composition sample, and the discipline to treat a reported rate as a claim to be tested rather than a result to be filed. The gate-scale and composition telemetry we wire into waste intelligence software does the same job continuously, but the clipboard version works fine. Show me a diversion rate and I'll ask for the disposal tonnage underneath it. The number nobody can redefine is the one worth trusting.

Disclosure: I build data and process-optimization systems at Renewable Waste Energy, including the Optimal Waste Intelligence platform referenced above.

Sources & Notes

• National generation, recovery, and landfill shares are drawn from the U.S. EPA's national overview of materials, waste, and recycling, reporting year 2018, the most recent full national dataset.
• For the certification bar, GBCI's TRUE certification for zero waste program sets the 90% threshold, the twelve-month basis, and the exclusion of incineration.
• CalRecycle spells out the January 2020 reclassification under AB 1594 in its guidance on green material used as alternative daily cover.
• The move from diversion-rate accounting to a disposal-based metric is documented in CalRecycle's explainer on per-capita disposal and goal measurement.
• Tipping-fee context comes from EREF's analysis of MSW landfill tipping fees. The 82%-to-61% audit and the sort-line recall numbers are from RWE project work in 2022 and a classification dataset we built in-house in 2023.