Methodology
How Cirqular compares destinations
Cirqular first checks which destinations can take a material — requirements such as facility approvals, capacity and material compatibility. Only destinations that meet every requirement are scored. Destinations that need confirmation, or that are unavailable, are shown separately and never scored.
01Material profile
Every analysis starts from a structured material profile: the site, the residue and how it is produced, its physical form, the annual volume and the current arrangement (route, destination, collector, gate fee, haulage and revenue per tonne). Values come from what you type, documents you upload, or inference from the description — and each keeps its source and, where it was inferred, a qualitative confidence.
Volumes are converted to tonnes a year (weekly × 52, monthly × 12, kilograms ÷ 1,000). Net cost per tonne is gate fee plus haulage minus revenue; if a component is missing the cost is marked as partial, not completed with zero. A few material facts matter for destination requirements — catering-waste contact, animal-derived ingredients, packaging and waste or by-product status. Cirqular infers them where the description supports it and otherwise asks one targeted question. “Not sure” is always accepted and recorded as unknown.
02Eligibility and destination matching
Each destination is checked against the profile by fixed, inspectable rules — never by AI. Every check passes, fails or cannot be decided yet because information is missing. The outcome is one of three:
- Suitable — every mandatory requirement is met. Only these destinations are scored and ranked.
- Needs confirmation — nothing fails, but something is unknown (for example animal-derived ingredients answered “Not sure”). Not scored until resolved; answering the question re-runs the checks straight away.
- Unavailable — a mandatory requirement fails. Never scored or recommended.
The checks in this prototype:
| Check | Area |
|---|---|
| Accepts this type of material | Material requirement |
| Accepts this physical form | Material requirement |
| Suitable for farmed-animal feed | Animal-feed restriction |
| Approved for animal-derived material | Animal by-products |
| Authorised to receive this material | Destination requirement |
| By-product status | By-product status |
| Permitted collection available | Waste collection |
| Capacity available | Capacity |
| Within volume limits | Capacity |
| Meets contamination limits | Quality and contamination |
| Collection frequency fits (advisory — flags, never excludes) | Operational fit |
These rules are a prototype encoding of the requirements described in the Cirqular proposal and need professional regulatory review before operational use. A by-product claim alone never passes the by-product check. Material classification and regulatory obligations remain the responsibility of the producer.
03The Cirqular Score
Cirqular compares suitable destinations across five factors:
- Cost
- Net cost per tonne: gate fee plus haulage, minus any revenue for the residue. Lower is better; net revenue scores best.
- 30%
- Resource value
- Where the route sits in the Cirqular resource-value hierarchy — redistribution / reuse, animal feed, anaerobic digestion, composting, disposal.
- 25%
- Emissions
- Estimated transport plus treatment CO₂e per tonne. Lower is better.
- 20%
- Logistics
- Estimated distance, collection-frequency fit and how many handling requirements the receiver sets.
- 15%
- Reliability
- How strong and current the data behind the comparison are. Quotes and public registers count for more than estimates.
- 10%
These are pilot starting weights and can be adjusted by the user. They are not a regulatory or scientifically validated standard.
Score = 100 × (0.30 × Cost + 0.25 × Resource value + 0.20 × Emissions + 0.15 × Logistics + 0.10 × Reliability)
04Normalisation, missing data and ranking
Each factor is turned into a 0-to-1 score, where 1 is best. For cost, emissions and distance, the best option for that material gets 1, the worst gets 0, and the rest fall in between in proportion. If every option has the same value, each gets 0.5.
That means a score is relative to the options available for that material. A score of 85 for one residue is not necessarily “better” than 80 for an unrelated residue — they were compared against different sets of destinations. Scores a few points apart should be treated as comparable.
If a factor can’t be calculated for a destination — for example, a gate fee is unknown — it is left out, and the remaining weights are scaled up so they still total 100%. The result says so (“calculated using 4 of 5 factors”), and the missing data also lowers that destination’s reliability. Unknown values are never treated as zero.
Destinations are ranked by score, highest first. Ranks are given on the score as displayed (rounded to a whole number), so two destinations shown with the same score share a rank (1, 1, 3); between them, the unrounded score and then the name set the order. Scores less than 3 points apart are described as comparable. Weights that don’t total 100% pause the ranking rather than produce a score.
05From ranking to decision
The ranking answers which suitable destination scores highest under these priorities. The decision answers what the site should do. They are not the same thing, so every stream gets exactly one decision:
- Change — the top-ranked destination is not today’s, and it saves at least €1,000 and 5% of the current annual cost, cuts estimated emissions by at least 1 tCO₂e a year, moves the material up the Cirqular resource-value hierarchy, or today’s destination fails a mandatory requirement. A change can cost more when it improves emissions or resource value; that is always shown as a trade-off.
- Keep current — the top-ranked destination is already today’s destination, or no suitable alternative clears those thresholds. Nothing changes, so no saving is claimed. Never chosen when today’s destination is known to fail a mandatory requirement.
- Needs information — open questions, missing details, or a baseline (such as current costs) too incomplete to compare the top-ranked destination with today. The top-ranked destination is shown only as provisional evidence.
- No suitable alternative — no destination meets every requirement.
- Ranking paused — the priority weights don’t total 100%.
Today’s destination is recognised among the compared destinations only by an explicit link, or by an exact match of its name (ignoring case and punctuation) at the same location (within 1 km) and on the same route type. There is no fuzzy matching. When today’s destination is the top-ranked one, a modelled difference — for example haulage estimated by the model against the haulage you pay — is explained as a modelling difference, never as a saving.
The thresholds are pilot assumptions. They never change eligibility, scores or rankings.
06Factor methods and pilot assumptions
The figures below are prototype assumptions for the pilot. They are kept in one configuration file so they can be reviewed and replaced as pilot data arrive.
Resource value
The order follows the Cirqular proposal. The numeric values are a pilot scoring assumption — the gaps between them do not come from legislation.
| Redistribution / reuse | 1.0 |
|---|---|
| Animal feed | 0.8 |
| Anaerobic digestion | 0.6 |
| Composting | 0.4 |
| Disposal | 0.0 |
Cost
Net cost per tonne = gate fee + haulage − residue revenue. Annual cost = net cost × annual tonnes; saving = current annual cost − destination annual cost. Where no haulage quote exists, a prototype model estimates it at €6/t plus €0.15/t per km.
Emissions
Transport (0.12 kg CO₂e per tonne-km) plus treatment by route (redistribution / reuse 25, animal feed 40, anaerobic digestion 30, composting 45, disposal 450 kg CO₂e/t). These are prototype assumptions, not validated life-cycle assessment values.
Logistics
Logistics = 60% distance + 25% collection fit + 15% operational fit. Unknown parts never get full marks: unknown distance and unknown handling requirements count 0.5, unknown collection availability 0.5, and an unknown collection frequency 0.75. See Distance and transport.
Reliability
The average trust in the price, capacity, accepted-materials and distance data. Data older than 180 days count for 90%, older than a year 75%, undated 85%.
| Measured | 1.00 |
|---|---|
| Public source | 0.95 |
| Quoted | 0.85 |
| User provided | 0.70 |
| Estimated | 0.50 |
| Example assumption | 0.40 |
| Unknown | 0.00 |
07Distance and transport
Distance from your site to each destination feeds three things: the distance part of the logistics score, the prototype haulage estimate (where no haulage quote exists) and transport emissions (0.12 kg CO₂e per tonne-km, a pilot assumption).
Road routes are preferred. Where road routing is available, Cirqular uses the routed travel distance and, where returned, an estimated driving time. In this prototype routing comes from the public OSRM demonstration server — open-source, but without any availability guarantee and not intended for production traffic.
Straight-line estimates are the fallback. If routing is off, unavailable or times out, distance is the straight-line distance × 1.3 (minimum 5 km), labelled “Straight-line estimate”. It is not an actual haulage distance, and no route is drawn for it.
Every distance shows how it was obtained (road route or straight-line estimate). Road-routed distances count for more in the reliability factor (0.85) than straight-line estimates (0.5). Without a site location, distance — and therefore cost, emissions and the distance part of logistics — is shown as unavailable rather than zero.
Site locations come from coordinates stored with the analysis (the example site’s location, or a point you set on the map). If none is set, a town named in the location text gives an approximate town-centre location. Location text is not sent to any geocoding service. Example facility locations are approximate, fictional points.
08Decision priorities
Shortcuts for common priorities. They are prototype weight sets, not industry standards, and every weight can be inspected and changed.
| Priority | Cost | Resource value | Emissions | Logistics | Reliability |
|---|---|---|---|---|---|
| Balanced | 30% | 25% | 20% | 15% | 10% |
| Cost-focused | 50% | 15% | 15% | 10% | 10% |
| Carbon-focused | 15% | 20% | 45% | 10% | 10% |
| Resource-focused | 15% | 50% | 15% | 10% | 10% |
09Provenance labels
Every important figure carries one of these labels. A figure built from several inputs takes the weakest one. Values from the fictional example destinations add “example”, for example “Quoted · example”, so example data is always identifiable as fictional.
- Measured
- Weighed or metered, e.g. weighbridge records.
- Quoted
- A price or figure quoted by the facility or contractor.
- Public source
- Taken from a public register or published source.
- Document extracted
- Read from a document you uploaded.
- User provided
- Entered or confirmed by you.
- Inferred
- Worked out by Cirqular from other information; shown with a confidence level.
- Road route
- Road distance from a routing service.
- Estimated
- Calculated from a model or approximation, e.g. a straight-line distance or the haulage model.
- Pilot assumption
- A Cirqular pilot scoring assumption, open to review.
- Example assumption
- Fictional value from the example data. Not real data.
- Unknown
- Not provided yet. Never treated as zero.
10Site overview totals
The site overview runs every residue stream through the same steps as a single analysis — eligibility, then the Cirqular Score, then the decision — under the current priorities, and adds up the changes Cirqular actually recommends. A stream kept as it is counts as no change: it adds nothing to the saving or the emissions change, stays on today’s route in the route mix and isn’t shown as a new destination. A stream still needing information adds no future arrangement. Changing the priorities re-ranks and re-decides every stream.
A total includes only the streams that have the figure, and always says how many: for example “Based on 4 of 5 streams”. A missing figure is never added as zero. Cost totals use streams whose current costs are all known and whose proposed destination has known costs; savings are today’s cost minus the cost after the decision over those same streams. Emissions totals use streams where the current destination’s location is known, so current transport can be estimated with the same prototype factors as the alternatives. Distance-dependent figures can shift slightly when road routing is available.
Route mix and resource-value movement are measured in annual tonnes, not numbers of streams. The resource-value positions are the Cirqular resource-value hierarchy, not a statutory ranking.
There is no site-level Cirqular Score. Each score is relative to the destinations suitable for that material, so scores for different streams are not on a common scale and an average would mislead. Each stream is also checked on its own: combined volumes sent to one destination are not yet checked against its capacity.
11How Cirqular uses AI
AI may help Cirqular:
- extract data from invoices, reports, specifications and other documents;
- interpret free-text descriptions of a material;
- suggest structured classifications, such as material category or possible animal-derived content;
- identify missing information and which question to ask;
- draft a plain-English explanation of a recommendation.
AI does not decide whether a destination is available. Mandatory requirements are checked by inspectable rules that read the structured residue profile, and only destinations that pass them are scored. A suggestion from AI becomes part of the profile only as a labelled value with its source and a qualitative confidence (high, medium or low), never as a percentage it can’t support.
Cirqular never overwrites a value you provided. When a document or suggestion disagrees with you, or two documents disagree with each other, you choose. A stated value from a document outranks an inference; low-confidence values that would change the result are confirmed with you first; unknown stays unknown.
Spreadsheets (CSV, XLSX) and text files are read directly in your browser without AI. PDFs and images need a document-reading service; when none is configured they are listed as unsupported and you continue manually. Simulated demo extraction is always labelled as such. A generated explanation is shown only if every facility, route and figure in it matches the calculated analysis; otherwise the standard explanation is used.
In this prototype, file contents are not stored: only file names and the values you keep, in your browser. When a document-reading service is configured, files sent to it are processed for that request and the provider’s own data terms apply. There is no server-side storage, retention policy or certification yet.
12Current prototype assumptions
- All company, facility and commercial data shown in this example are fictional.
- Emission factors, resource values, the haulage model, logistics and reliability weights, presets and overview status thresholds are pilot assumptions, not validated values.
- Road routing uses the public OSRM demonstration server; without it, distances are straight-line estimates × 1.3.
- Location text is matched against a small offline list of Irish towns; nothing is sent to a geocoding service.
- Eligibility rules are a prototype encoding and need professional regulatory review.
- Everything you enter stays in this browser (local storage). There are no accounts and no server-side database.
Cirqular provides decision support based on the information available. Estimated and example values should be confirmed before operational decisions are made.
Analyse a residue stream to see the score in use.