A Waste Collection Percentage Is Only as Useful as Its Denominator

A collection dashboard can show a reassuring percentage while leaving the manager unsure which street still needs a vehicle. The arithmetic may be correct. The problem is that nobody has settled what the number represents.

Does it describe people with access to a service, visits completed against a schedule, waste reaching a facility, or material managed under controlled conditions? Each measure can be useful. Treating them as interchangeable makes it harder to see where the service is failing.

This essay proposes a practical reporting approach for a fictional municipal operation. It is an engineering discussion, with no claim of a field deployment or a measured public-service improvement.

Name the thing being counted

Start with the unit. A household, a person, a scheduled visit, a vehicle trip, and a tonne of waste are different units. One truck trip may serve many collection points. One household may need several visits during a reporting period. A large load can come from a small part of the service area.

The official SDG 11.6.1 metadata distinguishes waste quantities collected from the population's access to collection services. It also treats management in controlled facilities as a separate part of the waste system. That distinction prevents an operational completion count from silently becoming a claim about citywide environmental performance. [1]

A dashboard label should therefore say exactly what it counts: recorded completed visits divided by planned visits, for example. Show the reporting period, service boundary, plan version, numerator, and denominator alongside it. A percentage without those details invites readers to supply their own meaning.

Keep the denominator independent of the upload

Consider a deliberately fictional shift with 100 planned visits. Ninety have a completion report; ten have no report yet. Dividing the 90 completions by the 90 uploaded records produces 100 per cent, but it answers a much narrower question than whether the plan was delivered. The documented completion fraction against the original plan is 90 per cent, with ten visits unresolved.

The denominator should come from the agreed service plan. It should not shrink when a device stops uploading. Likewise, repeating an upload should not create another completed visit. The counting unit needs a stable identifier that connects the plan and the evidence.

Real plans change. A road may close or an extra visit may be authorised. Preserve the original plan and the approved revision, with the reason and effective time. Then report performance against an explicitly chosen version. Quietly removing a difficult visit makes a trend look better while changing the question.

Give missing evidence its own place

An absent report does not establish whether a visit happened. It might reflect a missed service, a failed device, an unfinished route, or delayed synchronisation. Those possibilities call for different responses.

Use an unreported state until evidence supports a more specific classification. Show its count prominently, and give someone responsibility for resolving it. A supervisor may then confirm a missed visit, obtain the delayed record, or document an exception. Keep that decision and its basis.

This does not mean waiting indefinitely. Define a service deadline and a review deadline. An overdue reporting exception deserves action even before physical non-collection is established. Keep the operational urgency visible without describing uncertainty as an observed outcome.

Separate the service event from knowledge of it

A crew might record a visit at 10:00 and upload it at 13:00. A noon report should be reconstructible from what the platform knew at noon. A later reconciliation can add the report without pretending it had been visible earlier.

Temporal distinctions also appear in established observation models. OGC SensorThings separates the time of an observation from the time its result was generated. Platform receipt time is an additional field in the ledger proposed here; it should not be assumed to mean either of those times. [2]

Record the reporting cutoff and whether the displayed figure is provisional or reconciled. A revised result should link to the change record. That gives a manager a way to explain why yesterday's figure moved after an offline crew returned.

Use complementary measures to find the next action

Put planned-visit completion beside unreported visits, confirmed exceptions, and the age of unresolved reports. Add distinct measures for population access, collected mass, facility receipts, and controlled management when their evidence is available. Each should retain its own definition and source.

A high completion fraction with many unresolved records may be a reporting problem. A low completion fraction concentrated on one route may need dispatch attention. A strong collection count with missing receiving records needs investigation further along the material chain. These are prompts for review, not automatic diagnoses.

Avoid letting a district average hide repeated problems in a smaller area. Break the operational view down by route and service window, while protecting personal information. Keep enough context to distinguish a recurring gap from a single late upload.

Make the number answerable to a record

A useful metric needs a short contract: its purpose, unit, boundary, denominator, accepted evidence, cutoff rule, correction policy, and owner. Give the definition a version. If it changes, identify the break in the time series rather than presenting old and new values as directly comparable.

Test that contract with small examples before building predictions on top of it. Replayed uploads should leave the numerator unchanged. Corrections should affect the appropriate later snapshot. Missing records should remain visible. Invalid identifiers should be rejected for review rather than quietly added to the plan.

The companion collection ledger design study works through those rules using fictional visits. Its reference material is linked from the waste digital twin protocol. The immediate gain is an inspectable definition and a repeatable software check. Field effectiveness still needs field evidence.