Three hundred registrations, one decision.

The Salvation Army's 333 registrations are all held under one business number, that of The Governing Council of The Salvation Army in Canada, the corporation Parliament created in 1909 to hold the Army's temporal affairs. In the Army's own words the local work is done by ministry units: a corps, which is a church led by a corps officer, or a social services unit led by an executive director. Together they report $1.4 billion a year. The median unit runs on $761,000. Only eight per cent run on under $150,000, and none has a board of three or fewer, because the boards are not local at all.

Seven people sit on more than a hundred of these boards each, and one sits on 318. They are the Governing Council: the constitutional governors of a body that holds every corps, shelter, thrift store, and camp as a unit of one organization. Beneath them a small tier of territorial and divisional officers sit on a handful each. The soldiers and adherents of a corps, and its unpaid local officers, are its congregation and its volunteers; they are not its directors. That is the design, and the record shows what it buys.

Ontario has more than a third, the Atlantic provinces more than a quarter, which is where the Army's history in Canada is deepest. The purposes are mostly religion on paper and mostly poverty in practice: the shelters and food banks and addiction programs that are registered as the corps that run them.

By annual revenue

By region

Region follows the riding a registration's mailing address falls in; 2 could not be assigned. Median revenue is $761,000 and the whole body reports $1.38 billion; the median board has 9 members.

The model flags fewer than five registrations in 333, under one per cent, and none at the stronger level. A flag means only that an organization's filings over several years resemble those of charities that later lost their registration, and filings prepared centrally by an office that does hundreds of them do not resemble that. The model cannot see strain in a unit whose books are kept at territorial headquarters.

The register shows the other half of the design. Since 2019, 66 Salvation Army registrations have ended: 54 by choice and 12 for failing to file. Twenty-eight of them landed in a single year, 2023, the year the territory announced it was merging its divisions to four, which is what a consolidation looks like when one office decides it, sequences it, and files it. The other years show the same pattern on a smaller scale: registrations ended in batches, by choice, filed and finished. This spring the Cornwall corps and its food bank closed, the Army said, to steward its resources responsibly; the decision was made above the corps, as the Army's decisions are.

Nobody in a corps in Corner Brook or Nipawin had to learn how to do that; somebody at territorial headquarters already knew.

Revocations on the record, by year

Revocations published in the Canada Gazette, matched by business number, 2011 to 5 September 2026. 90 in all.

An Army closure is decided above the unit and executed by people who have done it before. A corps has no separate legal existence; its property vests in the Governing Council; openings and closures are decisions of territorial and divisional leadership under the international Orders and Regulations, which are internal. The counsel is retained, the accountant is on staff, the successor is the next corps over or the division itself, and the assets never leave the organization. The soldiers of the local corps grieve the closure; they do not file it.

The Army is not better at closures because its people care more. It is better at them because the structure puts the knowledge, the authority, and the successor in one office, and the local unit inherits all three. Most charities have the local board and lack the office. The Army has the office and, in the legal sense, lacks the local board.

Each organization in this case study is incorporated in a province or territory, or federally, and it closes under that statute first and under the Income Tax Act second. The two have an order, and no government page sets it out.

The Critical Path is that order on one folded sheet, one edition per jurisdiction: each step from the first hard conversation to the last filing, who decides it, and the document it files, with its regulator, its fee, and a specimen of its first page. Each edition will come as part of Foundations, with the Guide and access to the Foundations Course. Foundations is in development.

AnglicanUnited ChurchLegionAgricultural societiesCommunity foundationsAll the case studies

The counts are arithmetic on three public files: the Canada Revenue Agency's list of charities and their annual returns, 2011 to 2024; the Canada Gazette's notices of revocation; and Elections Canada's postal code file, which places a mailing address in a riding. A "flag" is the mark our model, Canary v3 / Distress v3, puts on a charity whose filings over several years resemble those of charities that later lost registration. It is a weak signal and we say so: about 1.4 in a hundred charities lose their registration within three years, about 2.9 in a hundred flagged ones do, and the rest carry on. We use flags to count, never to approach anyone.

This body's registrations are picked out by the business number they share, which is exact. Regions follow the riding of the mailing address, and 22 of Canada's 343 ridings are not yet in the crosswalk. Where a count is under five, we print "fewer than five", so that no organization can be picked out of it. The method brief has the model's validation, the counting rules, and the limits in full.

Local record, as reported: Salvationist.ca, governance of The Salvation Army in Canada; Salvationist.ca, ministry units; Salvationist.ca on the territory's divisional boundary changes, 2023; CTV on the Cornwall corps and food bank closing, 2026.

Sources: CRA List of Charities and T3010 returns, 2011 to 2024; Canada Gazette, Part I, revocations 2019 to 2026; Elections Canada, 2023 representation order. Model: Canary v3 / Distress v3, 1 July 2026; out-of-time AUC 0.68; flag precision 2.5 to 3.3 per cent. Counting script: case-study.py; the riding-level counts it produces: case-studies-aggregate.csv. The organization-level score file is available to a reviewer under a data agreement, and the method paper on request. If you find an error in the counting, write to info@publicbenefit.ca; we will correct the case study and record it in the correction log.