A draft specification and an open tool for describing what happens to every material in a product after its useful life. Not one recycling rate. One fate per material.
The distance between these two numbers is the function being lost while it is still being counted as recycling.
Simulated on a demonstration dataset, not measured. Each bar is the whole of that material, read left to right from what comes back at full function to what is dissipated.
A product can report ninety per cent recycling by mass while losing every gram of the metals that were hard to get in the first place. Three things hide inside one number.
Copper rides the smelter's carrier metal all the way to a refined cathode. Lithium in the same furnace oxidises into slag on the first pass and never comes back. Both sit inside the same headline figure.
The same product sent to a smelter, a dismantling line or a shredder produces materially different outcomes. A product does not have a recycling rate. It has one per route.
Aluminium recovered as a cast alloy and aluminium recovered as slag aggregate are both called recovered. Only one of them can go back into the part it came from.
Every joint gets a separation class, every recovered gram gets a quality class, every number gets an evidence level. All three are proposed by this draft; none is an existing international standard.
| D0 | Separates during normal processing |
| D1 | Tool-less, hand separable |
| D2 | Common hand tool |
| D3 | Specialist tool |
| D4 | Thermal or chemical process |
| D5 | Destructive |
| D6 | Practically inseparable |
The class describes the joint, not the part. A battery held by a clip and an adhesive strip takes the harder of the two.
| C0 | Closed loop, equivalent function | 1.0 |
| C1 | High-quality open loop | 0.8 |
| C2 | Downcycled | 0.45 |
| C3 | Feedstock recovery | 0.25 |
| C4 | Energy recovery | 0.05 |
| C5 | Dissipative loss | 0 |
Energy recovery is C4. It is reported on its own line and never summed into a material recycling rate.
| E0 | Self-declared, no dataset |
| E1 | Manufacturer documentation |
| E2 | Engineering calculation |
| E3 | Process simulation |
| E4 | Experimental recycling trial |
| E5 | Independently verified industrial data |
A record's headline level is the lowest among the values that materially affect it. Not the average, and certainly not the highest.
A product with trial data on everything except one self-declared connector is a self-declared product.
Evidence takes the minimum across a record and everything it imports. That single rule removes the incentive to source the weakest link quietly, because the weakest link becomes the headline. It is also why a record can be complete and still honest about being weak: completeness and recyclability are measured separately here and are never combined into one score.
One document underneath. Editing a component changes the flow, the flow changes the rating, and conformance is checked against the specification while you work.
Enter parts, masses and material composition in grams. Mass declared but not assigned to a material is shown as a hatched remainder rather than quietly renormalised away.
Drag in shredders, separators, smelters, leaching, polymer lines. Each node holds a split matrix deciding which port every material leaves by. Fractions summing below one are flagged, not absorbed.
Every material's full fate, its recovery, its index, and the weakest link named. The grade is there, but the breakdown is what you are meant to read.
The classes are not decoration. A component's separation class feeds the dismantling node directly: move a battery from D1 to D5 and the share of lithium reaching battery pre-treatment falls with it, the remainder goes to the smelter, and the lithium bar in the wheel turns red. Design decision to recovery outcome, in one dropdown.
A 2018 study modelled a modular smartphone through three processing routes. Shredding produced the highest mass-based rates. It was not judged the best route.
Published figures, Reuter, van Schaik & Ballester, 2018. Percentages of product mass.
The authors judged dismantling the better option once the range of recovered materials was weighed alongside the recovered mass. Shredding wins on the scale and loses on the periodic table.
Notice also that smelting leads on the third bar only because combustion is counted. That is why energy recovery is kept on its own line here and never folded into a material rate.
No affiliation with or endorsement by any manufacturer. The composition and element-level figures in the tool are a demonstration dataset built for the prototype and are drawn from no publication.
The battery comes from a cell supplier, the board from an EMS. A format that cannot absorb their records forces every integrator to guess, and a guess published next to trial data is worse than no record at all.
The same document with recordType: "component". Composition, internal joints, issuer, evidence level, integrity hash. Publishable on its own.
It arrives locked and attributed. The integrator owns the interface joint and the disassembly sequence. The supplier owns the composition. Editing their numbers transfers ownership, and their evidence level stops travelling with it.
| Field | Owner |
|---|---|
| Composition, mass, internal joints | Component issuer |
| Interface joint to the assembly | Integrator |
| Disassembly sequence position | Integrator |
| Route and recovery outcomes | Whoever runs the route |
Confidentiality is the real obstacle, not the file format. A cell supplier will not publish a cathode recipe. Draft 0.1 supports element-level disclosure only, which loses the formulation and keeps everything a recycler needs. Declared ranges and third-party lab attestation are sketched but not specified. Only the third gets a full supply chain to the top conformance level, and it needs an institution that does not yet exist.
It may publish composition. That one restriction is what stops a number appearing with nothing behind it.
Identity, architecture, components, composition. Mass balance closes.
Connections and disassembly described for every component.
A full route, its outputs, and a quality class on every output.
Every declared value carries an evidence level and a source.
Every recovery value at E4 or E5. Trials, not simulation.
A high level means the claim is checkable, not that it is good. A record can reach L4 and grade badly, or grade well and sit below L1. Whether L5 is reachable at industrial scale is an open question — it demands a trial per material per route.