MAK Immigration Data Observatory: Release 2

Replaced, November 3, 2026. A corrected version is at https://doi.org/10.5281/zenodo.23927168. This version's description and README wrongly say the PR tables state months only on the wait-time column, and state the suppression_or_unavailable rule too broadly. The ZIP, the ledger file and the government values did not change. This version stays online for the record. Correction notice, November 3, 2026 (third correction that day; this version replaces 10.5281/zenodo.23915719). What was wrong in the description and README of version 23915719, and what changed: They stated the Release 2 approval as a plain fact. They now say that Usman Khalil, RCIC R709592, states that he approved these on September 9, 2026. The build files do not themselves record that approval. They gave July 6, 2026 as the data date of the build. Other files in the build carry other dates, so the sentence now names the PR intake file. They said three dates appear only on the canonical page. March 26 and January 16, 2026 also appear in the ledger files as listing or signature dates, so the sentence now says the three are effective or expiry dates. They said to filter on row_role before summing, but not that leaf rows alone can fall short of province_total. They now say so. Their Global Skills Strategy note left out the 14-day public commitment in the source footnote. It now gives it. They had no note on suppression_or_unavailable, on expected_wait_state or on the unit of PR rows in the processing snapshots. They now do. Other changes: the notice for version 23915719 is kept below, labelled as history. In public-policies-2026-source-page-qa.json, only the file-level correction_notice changed. It now says current states are on the canonical page and gives one example. No record in the file changed, so the same four records still differ from the copy inside the ZIP. Use this corrected copy instead of the copy inside the ZIP. The checksum file is updated. What did not change: mak-observatory-release2-build.zip is byte for byte the September 9, 2026 build, with the same SHA-256. No government value changed. History: earlier notice of version release-2-correction-2-2026-11-03 (10.5281/zenodo.23915719), kept for the record. It describes that version, not this one. That version replaced 10.5281/zenodo.23911007. What was wrong in version 23911007: in the corrected ledger file and in the copy inside the ZIP, the Francophone Minority Communities study-permit record (study-permit-franco-minority-communities-student-pilot-2026) gives August 25, 2027 as the expiry without saying it applies to Part A only. The source says Parts B and C remain in effect until the revocation or expiry of a companion policy, which expires on November 30, 2032. In the same files, the record sah-refugees-april-23-extension quotes the wrong sentence as its expiry evidence; its expiry date was correctly left blank. The description and README said IRCC revises recent months, and earlier years are revised too. They said MAK-derived or source-trigger-resolved dates and current-state labels appear only on the canonical page, but the ledger files hold 5 effective dates resolved from a policy's signature, and build-time status labels. They said the ledger keeps supersession dates distinct, and it has no supersession date field. They said annual totals are not in the normalized files, and the Express Entry candidates file is annual. Their eTA note did not say that the unit field reads "days", and they had no note on Global Skills Strategy rows, on the separate dates of the PR wait-time column, on program labels, on summing rows that include subtotals, or on the official_change_summary field. What changed in version 23915719: the corrected copy of public-policies-2026-source-page-qa.json now also corrects those two records. Four records now differ from the copy inside the ZIP, and each carries a mak_correction_2026_11_03 object: palestinian-territories-immigration-measures-2026-2 (blank effective and expiry dates; the source states August 1, 2026 and December 30, 2027), soccer-2026-biometrics-exemption (blank signature date; the source states November 24, 2025, amended February 6, 2026), and the two records above. Use the corrected copy instead of the copy inside the ZIP. That version's README, checksum file and description were updated. What did not change in version 23915719: mak-observatory-release2-build.zip is byte for byte the September 9, 2026 build, no government value in the normalized datasets changed, and the expiry date value August 25, 2027 is kept, because the source states it for Part A. MAK Immigration Data Observatory: Release 2 preserves the Release 2 build published by MAK Canadian Immigration Services on September 9, 2026. The package contains normalized CSV and JSON datasets, source receipts and captures, Express Entry round-event history, dated IRCC/CIMM processing and service-standard snapshots, and a bounded 2026 public-policy Change Ledger. Raw captures cover the eight workbooks and 17 policy pages. The Express Entry feed, CIMM pages and policy index are recorded by hash and retrieval date only. Coverage includes study permit holders and prior study-permit/PGWP status to permanent residence from January 2015 to June 2026; new study-permit applications processed/finalized and temporary-resident new applications and extensions processed/finalized (approved, refused or withdrawn; includes study permits, so not additive with the study-permit series) from January 2023 to June 2026; Express Entry round history through round 441 on September 4, 2026, with split identifiers preserved; processing/service-standard snapshots as of January 31 and February 28, 2026; and 17 source-reviewed 2026 public-policy Change Ledger records. Government source values and MAK-derived calculations are labelled separately. Suppressed values are not converted to zero. Partial periods remain partial. Historical Express Entry rounds are not forecasts. Processing snapshots are not interpolated into a continuous series. The Change Ledger keeps listing, signature, effective and expiry dates distinct and records supersession status and source wording separately. The deposited ledger files record source wording and date status at the source-page QA stage. They fill effective dates only where the source states a date or says the policy takes effect on signature, and they label the second kind. Three effective or expiry dates that need a calculation or another source appear only on the canonical page: March 26, 2026 (effective), January 16, 2026 (effective) and May 7, 2028 (expiry). March 26 and January 16, 2026 also appear in the ledger files, but only as listing or signature dates. The ledger files carry build-time proposed status labels from September 9, 2026. Current states as last rechecked and MAK summaries appear only on the canonical page. In the ledger's official_change_summary field, the first sentence is the source page's heading and the second is a fixed sentence added by MAK's build. IRCC revises its figures, most heavily in recent months but also in earlier years. Between this build's PR intake file (data as of July 6, 2026) and IRCC's later file (data as of September 1, 2026), every month of 2025 in the Total row changed, and most months of 2023 and 2024 changed by small amounts. June 2026 PR applications received was 19,485 in this build and 30,430 in IRCC's later file. Derived changes in the QA report apply only to this build's data. National total rows are in the raw workbooks, not the normalized files, except the PR Intake file, which includes a Total row: exclude it when summing countries. The monthly files omit annual-total columns. The Express Entry candidates file is annual, plus a January to June 2026 period. The province-based normalized files also contain subtotal rows, such as province_total: filter on row_role before summing. In the PR file and the two prior-status files, leaf rows alone can fall short of province_total, because some rows with no child rows are tagged aggregate, such as Protected Person in Canada and All Other Immigration - Total in the PR file and Provincial Nominee Program Total in the prior-status files. For eTA rows in the processing snapshots, the unit field says days, but the raw values are in minutes: use the raw fields. For Global Skills Strategy rows, the service standard is in the raw field only and reads 60 with a footnote marker. The source footnote, not held in the deposit, gives a 14-day public commitment for GSS. In the processing snapshots, the PR wait-time column carries its own date in the source: January 2026 in the March package and February 1, 2026 in the April package. Program labels keep source footnote text and differ between the two snapshots, so match them by hand. "Temporary Resident Visa" appears twice in each snapshot. In the processing snapshots, suppression_or_unavailable is true when any one cell in the row is missing, so most such rows still hold published processing times. expected_wait_state reads PUBLISHED on rows where the source shows "--". The PR tables state months only on the wait-time column; the unit on other PR columns is MAK's reading. Professional review provenance: Release 1 taxonomy and limitations were reviewed by Usman Khalil, RCIC R709592, and Omer Khalil, RCIC R710149. Release 2 underlying classifications, methodology and date/status/supersession resolutions: Usman Khalil, RCIC R709592, states that he approved these on September 9, 2026. The Release 2 build files keep their build-time status labels, such as RCIC_REVIEW_PENDING and NOT_PUBLIC, and do not themselves record that approval. This is descriptive research infrastructure, not individualized immigration advice, a prediction tool or government endorsement. Canonical methodology and limitations: https://makimmigration.ca/immigration-data-observatory/.