Methodology

Every record on this site comes from one of the sources listed below and carries a link back to the original report. There is no sample or placeholder data.

Data collection

Incidents are collected automatically. On a fixed schedule the pipeline fetches each monitored source, extracts the details of any reported incident into the schema below, checks it against existing records so the same event isn’t listed twice, and publishes it with its original source URL preserved.

Monitored sources

• NCCM Incidents Portal — individual Islamophobic and anti-Palestinian incidents documented by the National Council of Canadian Muslims.
• B’nai Brith Canada — published reporting on antisemitic incidents. Items that are policy announcements rather than incidents are filtered out.
• League for Human Rights — annual audit statistics on antisemitic incidents in Canada.
• Statistics Canada — police-reported hate crime counts, national, by region and for the Winnipeg census metropolitan area (tables 35-10-0066-01, 35-10-0191-01 and related), via the public Web Data Service.
• Toronto Police Service — per-incident hate crime open data, already coded by the police service, including the bias recorded.
• Winnipeg Police Service — monthly counts of violent offences by neighbourhood. The feed records no motive, so none is claimed for it.
• Manitoba news reporting— CBC Manitoba, the Winnipeg Free Press, the Winnipeg Sun, Global, CTV and others, found through a standing search and read from the publisher’s own page. Each record cites the publisher, never the aggregator.
• Anti-Sikh and anti-South Asian hate, Canada-wide — national and community news reporting of incidents against Sikh, Hindu and South Asian people and places of worship, found through a standing search and read from the publisher.

Reference sources

Some organisations document this field without publishing a feed a crawler can read. Their work informs the categories used here and is cited for context, not ingested as records:

• World Sikh Organization of Canada — Understanding and addressing anti-Sikh hate in Canada
• Institute for Strategic Dialogue — The rise of anti-South Asian hate in Canada (police-reported hate crimes against South Asians up 227% between 2019 and 2023)

Sources that publish individual events become incident records on the map. Sources that publish aggregate statistics are stored and shown as statistics, never converted into individual incident records, and the two are never added together in a headline figure.

One incident, however many reports

Several outlets report the same attack, each on its own day and in its own words. Before a record from a news or community source is published, it is compared with records held for the same province in the surrounding fortnight. A second independent report of an incident already held is attached to that record as a corroborating source and raises its evidence level to Corroborated Incident; it is not counted as another incident. Totals therefore count unique incidents, not the number of sources reporting them.

Evidence status

Records differ in how strongly they are evidenced, and the site says so rather than presenting everything as equally established. A record documented by a police service is not the same thing as a single report awaiting assessment, and neither is a finding by a court. Every record carries one of the following, shown on the search table and on the record itself, and usable as a filter:

• Reported / Pending Review — information received or identified, assessment not complete.
• Under Review — actively being assessed against its sources.
• Corroborated Incident — the core occurrence is supported by multiple independent, credible sources.
• Officially Recorded — documented by police, a court, a government body or another competent authority.
• Formal Determination — a court, tribunal or inquiry has made a finding.
• Disputed — material aspects are contested.
• Unable to Verify — the evidence does not meet the threshold. This does not mean the incident did not occur.
• Removed / Withdrawn — withdrawn over an error, a legal or privacy concern, or a retraction.

What happened, and why

The description of an incident is kept separate from any account of its motive. A motive is recorded as established, alleged, or undetermined, and is never inferred from the identity of those involved or from the nature of the act alone. Where the evidence does not support a motive, it stays undetermined.

Public reports

Anyone can report an incident through the Report an Incident form. Unlike the monitored sources, public reports are reviewed by a person before they are published.

Inclusion rules

• A record describes a specific event — violence, a credible threat, harassment, intimidation, property damage, incitement or comparable conduct — at a known time and in a known municipality, with at least one source.
• A hate or identity-based classification rests on a source documenting that the targeting was identity-based. The identity of those involved is not evidence of motive.
• Lawful protest, dissent and criticism of government are not incidents of any category, and are never classified as extremism.
• Private individuals are not named as offenders, criminals or extremists on the strength of an unverified accusation.
• CPO verification is a research assessment. It is not a police finding or a judicial ruling, and the record says which of the three it rests on.

Schema

Each incident record carries the following fields:

• Temporal/geographic: event date, latitude/longitude, province, city
• Research category: one primary category from twelve — hate and identity-based incidents, violent extremism, anti-government violence, political violence, threats and intimidation, community conflict, foreign-conflict spillover, foreign interference, public disorder, critical infrastructure, online harms, other social tensions — and any number of secondary categories
• Targeted identities: any number, each with a dimension (race or ethnicity, religion, sexual orientation, gender identity, sex, disability, age, language, immigration status, political hostility), a main group, optional detail, and whether the targeting rested on perceived identity. Religion is recorded apart from race or ethnicity. Unknown stays unknown.
• Classification: primary and secondary violence type
• Conduct: what was done — one of nine types, from physical violence to discrimination and exclusion — recorded apart from motive
• Targeted community: one of eight identity categories with a finer subgroup, recorded only where a source documents the targeting
• Actors: perpetrator group, target group
• Motivation: primary and secondary motive category, with its status
• Source: source type (police or official record, news reporting, community organisation, court or human-rights record, public submission)
• Themes: cross-cutting tags (e.g. election-related, opioid crisis)
• Impact: killed, injured, arrested counts
• Source: description, source name, source URL

Taxonomy

12 research categories · 11 target dimensions with detailed group lists · 8 violence types · 9 conduct types · 6 motive categories · 5 cross-cutting themes · 8 evidence levels. The Criminal Code’s identifiable groups, Statistics Canada’s hate-crime categories and human-rights grounds are not identical; where a comparison with an official scheme is reported, the mapping is stated. Corrections follow the corrections and takedown procedure.

Access

A public JSON API is available at /api/incidents, accepting the same filter parameters as the UI (province, region, city, year, violenceType, conductType, targetedCommunity, motive, theme, verificationStatus, sourceType, hateMotivated, dateFrom, dateTo, q). A CSV export of any filtered view is available at /api/incidents/export.

Known limitations

• The dataset reflects only what the monitored sources publish. Incidents that go unreported, or are reported elsewhere, are not captured.
• Extraction and classification are performed automatically, so a record may occasionally misread detail in the original report. The source link is always preserved so any record can be checked against its origin.
• Sources report a municipality, never a street address. Markers are placed at a deliberately offset point within that municipality — they indicate the city an incident occurred in, not where it happened.
• The taxonomy is a fixed, simplified set of categories.