What this platform is — and is not
Countability combines expertise in human, machine, and artificial intelligence to generate reliable insights from data, in a reproducible manner, without exaggeration or hallucinations.
Data first
Countability states what the record supports today. We do not make projections unless explicitly requested. Institutional projections appear as published — cited benchmarks, attributed to their issuers and kept distinct from Countability’s own analysis. Where projection work is explicitly commissioned, its assumptions and its confidence are stated with it.
No unsupported verdicts
Where institutions disagree, both positions are shown side by side — never averaged, merged, or adjudicated between, and policy merit is never graded. Where the record supports a bounded conclusion, Countability states it and shows the chain.
No causal overreach
Correlation is reported as observed co-movement over a stated window, never as cause. Derived readings print their assumptions with the claim.
Calculations you can check
Every number is official source data or arithmetic disclosed with it. Nothing is generated, estimated, or smoothed by a model.
How a claim is produced
Public data and evidence enter a deterministic analytical engine — a system built so that the same inputs produce the same record, every time. Sources are checked on the way in and preserved as evidence; conclusions are tested against explicit, stated rules. Every published claim can be traced from its final form back to the source evidence and the logic that produced it, and carries its evidence strength, its confidence, and its citations.
Large language models are extremely useful for exploration and drafting. However, they can produce different answers from the same prompt, making it difficult to know which result should be trusted for an important decision. For high-impact questions, that uncertainty can have dangerous or costly consequences. Countability separates language generation from evidence generation: a deterministic analytical engine, not a model, produces every published conclusion, from preserved sources, disclosed formulas, and stated assumptions. The same inputs and rules always produce the same result, so every conclusion can be independently checked rather than taken on trust.
Internal methods evolve; public claims remain reproducible. Corrections and revisions stay on the record, dated, rather than replacing it. Where machine tools, statistics, and mathematics do the calculating, every step runs under disclosed rules; AI-assisted tooling used elsewhere in Countability's work never originates a published figure or claim. Confidence and limitations are stated through the evidence tier — unsupported claims do not enter the record.
Vocabulary
- Vintage — one dated build of the record. The same inputs, rerun, produce the same vintage; a later vintage supersedes an earlier one and never overwrites it.
- Tier — the evidence class every claim carries: OBSERVED is source data or arithmetic on it, CORRELATED is co-movement over a stated window (never cause), INFERRED is a derived reading whose assumptions are printed with it.
- The record — everything published here: claims, commitments, disagreements, gaps, and corrections. Each entry cites its source, and the evidence stays visible when conclusions change.
How to read the clock
Every reading on this site carries more than one date, and each means something different. The reference period is what the data itself describes — the quarter, month, or fiscal year a figure is about. The official release date is when the issuing institution first published that figure. The evidence cutoff is when Countability's own record last checked and incorporated it — an official release newer than that date is shown as unresolved until the next complete update. The assessed date is when Countability's own reading, not just the underlying data, was last affirmed, changed, or corrected. The next trigger is the specific named release or event that would cause the reading to be checked again.
For releases on the tracked calendar (see /now/), Countability aims to acknowledge a new release by the next business day, and to adjudicate a reading it bears on within five business days or state plainly that it remains under review. Everything else is incorporated at the next complete monthly update. This is an operating target the institution holds itself to, not a published guarantee.
Evidence tiers
A restatement of official source data, or arithmetic on it. No interpretation.
Two official series move together over a stated window; the coefficient and window are shown, and no causal claim is made.
A derived reading that requires stated assumptions; the assumptions and their limits are printed with the claim.
Forward-looking or weakly evidenced. The platform deliberately publishes none.
What the record's state labels mean
These are the record's state labels: Measured, Contested, Blocked, Unmapped, and no label at all. Other small labels near them — whether a reading is new or has changed, how far out a question reaches, which part of a Monitor a card belongs to — are explained where they appear, not here.
The record has a specific, numbered target for this question, or for one of the questions inside it, and can say whether progress is keeping pace with it.
Two or more official sources answer this question differently. The record shows every position, side by side, and does not choose one.
Something the record needs for this question is missing, or a registered target can't yet be checked against the record. Where something is missing, the record names it.
The record does not yet reach this question through a target, a dispute, or a known gap. That's an honest starting point, not a mistake.
Sometimes nothing is labelled at all. That means the record touches this question some other way — a data series, an institutional projection, a dated event — just not through a target, a dispute, or a gap.
Standing policies
Disagreements
Institutional conflicts are first-class objects, held open until an issuer resolves them — never hidden, never settled by the platform.
Evidence gaps
Where the data needed to answer a live question does not exist, is not public, or is not fit for purpose, the gap is published as its own object — with what would close it.
Corrections
Nothing is edited in place. If a figure needs correcting, the correction produces a new, dated version of the record — the old one stays on the record, not deleted, not hidden.
Disclosed convention: The Resolution
A record, not a ranking: this entry states what was published and how it compares to what followed. It never says who was right, which institution performed better, or grades a forecaster's competence. Every projection vintage an issuer publishes is preserved and dated; movement between vintages is shown as the signed arithmetic difference already present in the record, in the issuer’s own units. Before an audited or finalized outcome exists, a lineage is shown as unresolved rather than guessed at.
The public Resolution record →
Licensing and attribution
Every source below names the licence it is published under, and reproduces the attribution notice that licence asks for. Sources whose terms this project has not read verbatim carry no licence claim at all — the absence is recorded, not filled in.
Statistics Canada — Value-added Products
Every indicator, ratio, and rule finding this platform publishes is a Value-added Product under the Statistics Canada Open Licence. The notice that licence requires:
Adapted from Statistics Canada, Building permits, by type of structure and type of work, monthly (34-10-0292-01), 2026-05-01; Canadian government finance statistics for the federal government, annual (CGFS basis) (10-10-0016-01), 2024-01-01; Components of international migration, quarterly (17-10-0040-01), 2026-01-01; Consumer Price Index, monthly (18-10-0004-01), 2026-07-01; Credit liabilities of private non-financial corporations (36-10-0640-01), June 2026; Employee wages by industry, monthly, unadjusted (14-10-0063-01), 2026-07-01; Employment by class of worker, monthly, seasonally adjusted (14-10-0288-01), 2026-07-01; Estimates of the components of interprovincial migration, quarterly (17-10-0020-01), 2026-01-01; Estimates of the components of natural increase, quarterly (17-10-0059-01), 2026-01-01; Experimental estimates for business openings and closures (33-10-0270-01), May 2026; Experimental estimates for business openings and closures by employment size (33-10-0722-01), May 2026; GDP, expenditure-based, quarterly (36-10-0104-01), 2026-04-01; Gross domestic product, expenditure-based, provincial and territorial, annual (36-10-0222-01), 2024-01-01; Health indicator statistics, annual estimates (13-10-0905-01), 2024; Housing starts, all areas, SAAR (34-10-0158-01), 2026-06-01; Housing starts, under construction and completions, all areas, annual (34-10-0126-01), 2022-01-01; Housing starts, under construction and completions, all areas, quarterly (34-10-0135-01), 2026-04-01; Housing starts, under construction and completions, centres 10,000+, monthly (34-10-0143-01), 2022-12-01; Indexes of business sector labour productivity and related measures, quarterly, seasonally adjusted (36-10-0206-01), 2026-01-01; Infrastructure Economic Accounts, investment and net stock by asset, industry, and asset function (36-10-0608-01), 2025; Infrastructure average age and remaining useful life ratio (36-10-0611-01), 2025; Investment in building construction, monthly (34-10-0293-01), 2026-05-01; Job vacancies, payroll employees, job vacancy rate, and average offered hourly wage, monthly, seasonally adjusted (14-10-0432-01), 2026-05-01; Labour force characteristics of immigrants, 3-month moving average (14-10-0471-01), 2026-07-01; Labour force characteristics, monthly, seasonally adjusted (14-10-0287-01), 2026-08-01; Life expectancy and other elements of the complete life table, single-year estimates (13-10-0837-01), 2024; New housing price index, monthly (18-10-0205-01), 2026-06-01; Non-permanent residents by type, quarterly (17-10-0121-01), 2026-04-01; Physical flow account for greenhouse gas emissions (38-10-0097-01), 2023-01-01; Population estimates on July 1, by age and gender, annual (17-10-0005-01), 2025-01-01; Population estimates, quarterly (17-10-0009-01), 2026-04-01; Private households by size, July 1, annual (17-10-0159-01), 2025-01-01; Revenue, expenditure and budgetary balance, general governments, provincial and territorial economic accounts, annual (36-10-0450-01), 2024-01-01; Revenue, expenditure and budgetary balance, general governments, quarterly (National Accounts basis) (36-10-0477-01), 2026-04-01; Unmet health care needs by gender and age group (13-10-0836-01), 2024; Vacancy rates, apartment structures of six units and over, census metropolitan areas, annual (34-10-0127-01), 2025-01-01, data vintage 2026-05. This does not constitute an endorsement by Statistics Canada of this product.
CMHC Licence Agreement for the Use of Data
Revocable grant; reproduce accurately, retain attribution, do not imply endorsement or link data to identify individuals. The direct survey-table scope does not resolve other CMHC instruments.
Verified 2026-09-08. All thirteen sections read in the live normal browser; complete text extraction preserved and compared. Direct HTTP HTML capture returned 403; no raw HTML capture is claimed.
Statistics Canada Open Licence
https://www.statcan.gc.ca/en/reference/licence
The licence forbids merging or linking the Information with other databases to identify an individual person, business or organization. Countability's standing aggregate-only exclusions already satisfy this; the licence and the doctrine agree.
Verified 2026-07-10. Licence text read verbatim from https://www.statcan.gc.ca/en/reference/licence (HTTP 200): Licence Grant, Acknowledgment of Source (a) and (b), and the restriction against merging or linking the Information with other databases to identify an individual person, business or organization.
Open Government Licence – Canada v2.0
https://open.canada.ca/en/open-government-licence-canada · accessed 2026-07-10
Exemptions: the licence grants no right to use Personal Information; third-party rights the Information Provider is not authorized to license; the names, crests, logos, or other official symbols of the Information Provider; or Information subject to other intellectual property rights. Non-endorsement: it grants no right to use the Information in a way that suggests official status, or that the Information Provider endorses this product or its use.
Verified 2026-07-10. Licence text read verbatim from https://open.canada.ca/en/open-government-licence-canada (HTTP 200, 23,505 bytes, sha256 b44f112f64081895e3e58982746db64cd8490b02b94105a10f0aa15ce39821ec, read from raw bytes with no summarizing layer): Grant, Attribution, Exemptions, Non-endorsement, No Warranty, Governing Law, Definitions, and Versioning. Each source's licence was additionally confirmed against its own entry in the Open Government Portal catalogue.
Bank of Canada Terms of Use, Disclosures and Disclaimers
https://www.bankofcanada.ca/terms/
The Bank's logo and wordmark, and bank note images, may not be reproduced without the Bank's written authorization (Terms of Use SS2.1-2.2) -- not engaged by a numeric restatement, but recorded here rather than assumed away. Terms of Use SS3 also prohibits circumventing any limit the Bank imposes on the frequency or volume of requests to its services, naming the Valet API explicitly; Countability's capture of Valet data is a small number of manual, citation-time fetches, not an automated poll.
Verified 2026-07-10. Terms of Use text read verbatim from https://www.bankofcanada.ca/terms/ (HTTP 200): SS1 Copyright / Permission to Reproduce ("the Bank permits you to freely use, copy, distribute and transmit its website content"), SS1.1 Attribution, SS1.2 Accuracy, SS1.3 Use of Content: Notice Requirement, SS2 Exceptions to Permission to Reproduce (bank note images; the Bank's logo and wordmark; third-party content), and SS3 No Unlawful or Prohibited Use (including the Valet API retrieval-limit clause).
United States Government work (public domain, 17 U.S.C. §105)
https://www.govinfo.gov/about/policies
The same notice warns that Government documents may include third-party copyrighted material used with permission: 'Publication in a Government document does not authorize any use or appropriation of such copyright material without consent of the owner.' Not engaged here -- the cited executive orders and proclamations are official texts authored by government officers, and Countability restates their operative clauses with citation rather than reproducing any third-party inclusion.
Verified 2026-07-17. govinfo's Public Domain & Copyright Notice read from raw bytes (archive-at-citation capture us-govinfo-policies, sha256 5189ea6f00ac…): 'The intent of the section is to place in the public domain all work of the United States Government' (17 U.S.C. §105, defined via §101), and 'public documents can generally be reprinted without legal restriction.'
Countability restates each licence as its issuer publishes it, and reproduces the attribution notice each licence asks for. Nothing here is legal advice, and nothing here asserts what any party may do with a source beyond what the source's own licence says. Where a publisher's terms have not been read in full by this project, this page says so rather than guessing.
Every public claim on this site is reproducible from the public sources it cites. Commissioned work follows the same evidentiary standard; its scope, price, timing, confidentiality, and refresh terms are set out in the written engagement — see Commissioned Intelligence.