No single ledger tracks Canada’s public money — but four federal dataset families come close
Asked:
“Which Canadian open datasets about government spending, procurement, grants and contributions, or lobbying are the most complete and machine-readable, and what gaps remain?”
This report scores a reasoned shortlist of 12 leading official datasets — not an exhaustive census of every Canadian open dataset — across spending, procurement, grants & contributions and lobbying, and distils 30 official and independent assessments into a gap taxonomy. Six of the 12 state all four machine-readability features; coverage starts range from 1996 (lobbying registrations) to 2017 (proactive contract records).
Machine-readability scorecard, grouped by category
Each dataset is scored 0–4 on whether its official documentation explicitly states: a machine-readable format, bulk download access, an update schedule, and a historical coverage start. The score measures stated machine-readability only — not substantive completeness. These are complementary datasets, not rivals; there is no single winner. Click any row for scope, formats, thresholds and remaining gaps.
Spending
Procurement
Grants & contributions
Lobbying
Feature not stated in source
What the scorecard cannot say
CanadaBuys award notices are the most machine-readable procurement feed — complete bulk CSVs with an XML dictionary, coverage from 2012, and daily refresh of current files — yet the files include only federal organizations, not the multi-government awards shown on the portal.
open.canada.caFederal proactive contract disclosure is broad but late and soft at the edges: records generally start in 2017, contracts of $10,000 or less appear only as annual aggregates, publication can lag four to five months, data are unaudited, and values can be ceilings or options rather than actual spend.
open.canada.caThe federal Lobbying Registry is an unusually strong bulk source — complete registration files from 1996 and monthly communication-report CSVs from 2008, both with data dictionaries — but it covers only registered paid lobbying and prescribed contacts, by design.
lobbycanada.gc.caPublic Accounts transfer-payment files give the longest reconciled annual series on grants (from 2003) but name recipients only at $100,000 or more; smaller payments appear only as class totals.
open.canada.ca
Seven gaps that remain, with sources
Synthesized from 30 official and independent assessment rows. Near-identical federal contract-search caveats are folded into single entries. Official documentation is cited in the scorecard above; the entries below mix official reviews (Procurement Ombudsman, OGP reporting mechanism, parliamentary evidence) with third-party assessments, labelled by host.
Provinces: strong examples, weak coverage
An independent cross-government scan rates Nova Scotia’s open-data portal the only perfect score of fourteen governments reviewed — API, multiple export formats, and 31,700+ contracts including submitted-bid information — while most jurisdictions offer no downloadable winners list at all. This is an assessment-source finding, not an official inventory.
blog.govconweekly.caThe same scan identifies Québec as the only Canadian jurisdiction publishing procurement in the international Open Contracting Data Standard, with bulk JSON exports — though SEAO’s French-first titles and weekly multi-file catalogue demand programmatic handling.
dev.toAlberta’s sole-source service-contract file is clean OData/CSV with a complete export endpoint and quarterly refresh — but covers only sole-source service contracts of $10,000 or more, a narrow slice of provincial buying.
open.canada.caStrong provincial examples do not solve the national problem: CanadaBuys is not a single portal for federal, provincial and municipal procurement, and provincial systems do not share data with each other or with the federal portal.
duke.yt
The 12-dataset shortlist
| Dataset | Category | Start | Formats | Refresh | Threshold | Score | Source |
|---|
Method: a reasoned shortlist of 12 leading official Canadian open datasets on spending, procurement, grants & contributions and lobbying, scored 0–4 on whether source documentation explicitly states machine-readable format, bulk access, update schedule and historical start — a statement count, not a completeness measure — plus 30 official and independent assessment rows synthesized into seven gap themes. Research conducted October 9, 2026. 42 source-backed rows span 21 hosts; the largest host backs 23.8%, so no single source dominates. Long scope and gap texts are truncated for space; full detail sits behind each source link.