Data quality and governance

Data inventory

From source to screen

1. SourceFive open datasets downloaded from open.toronto.ca, kept unmodified.
2. PipelineOne documented script cleans codes, applies exclusions, joins by station and by location, and prints reconciliation totals.
3. AggregatesSmall files by run area, ward, month and category. No addresses, no individual incidents.
4. SiteStatic pages read only the aggregates. Every chart has a data table behind it.

Field completeness

Exclusions and privacy

Incidents in and out

Emergency incident records, 2023 to 2024

Location is withheld by design

Share of incidents published without map coordinates, by call type

Medical calls are published to the first three postal characters only (MFIPPA), so this site maps by run area, never by point.

Data quality register

FindingEvidenceTreatment hereWhat would fix it at source

Methods

Response time

  • Interval from TFS alarm time (when Toronto Fire was notified) to TFS arrival time (first unit on scene).
  • Reported as the 90th percentile using the nearest-rank method, with the median alongside.
  • Emergency call types only. Excluded: no arrival time, zero or negative intervals, and intervals over 30 minutes.
  • Reference of 6:24 from NFPA 1710 (1:04 + 1:20 + 4:00). Shown as a yardstick; Council-approved service levels are for Toronto Fire to state.
  • Run-area figures are suppressed on the map below 20 responses.

Fires, inspections and the risk index

  • Fire trends count incident types 01 (fire) and 02 (explosion). Type 03, no-loss outdoor fires, is shown separately because it enters the data in 2018.
  • Category charts cover 2020 to 2024 and are keyed on the Ontario Fire Marshal code, so wording changes between years do not split a category.
  • Inspection rows are grouped by inspection thread to give one record per inspection; a row with no Fire Code reference is an inspection with no violation.
  • The risk index converts six measures to percentile ranks across the 84 run areas and combines them with the weights shown on the risk page. Areas with fewer than 10 inspections take a neutral rank on that measure.

Privacy and disclosure

  • Uses only data the City has already released, with personal information removed and locations generalised by the publisher.
  • This site adds a further step: only aggregates are published. Inspected addresses, though public in the source, are not repeated here.
  • Fire incident records exclude some incidents under section 8 of MFIPPA (law enforcement), so counts may be slightly below operational totals.
  • Before any internal version combined these records with patient-care or personnel data, a privacy impact assessment and PHIPA review would come first.

Accessibility and reuse

  • Designed to WCAG 2.1 AA principles: keyboard-operable controls, a data table behind every chart, colour never the only signal, light and dark themes chosen separately.
  • Charts never rely on hover: every hovered value also appears in a table.
  • No tracking, cookies or third-party scripts.
  • The pipeline is one standard-library Python script, so a refresh is a single command when the City publishes new data.