Canadian Mortgage Rate Data: Sources & Methodology
Ratellow publishes a live Canadian mortgage rate dataset covering the best available 3-year fixed, 5-year fixed and 5-year variable rates across 6 tracked lenders, alongside the posted 5-year benchmark, the Prime Rate and the 5-year Government of Canada bond yield. This page documents exactly where those numbers come from, how often they change, what we do when a source fails, and how to cite the data. It is written to be checkable: every figure below is rendered from the same dataset that powers our rate pages, not retyped by hand.
Dataset as of . Underlying rate pages refresh twice daily.
1. What we track
The dataset carries six headline series plus a per-lender rate table. The distinction that matters most is street rate versus posted rate: a street rate is what a qualified borrower can actually transact at today, while a posted rate is the published sticker rate that mainly survives as an input to penalty calculations. Ratellow leads with street rates.
| Series | Field | Definition | Refresh |
|---|---|---|---|
| 5-Year Fixed (street rate) | streetRate5Yr | The lowest 5-year fixed rate a borrower could actually get from the lenders we track — not an advertised posted rate.Median of the "best market rate" 5-year fixed figure reported by each source in the run. | Twice daily |
| 3-Year Fixed (street rate) | streetRate3Yr | The same measure over a 3-year term, the fastest-growing term choice among 2026 renewals.Median of the "best market rate" 3-year fixed figure reported by each source in the run. | Twice daily |
| 5-Year Variable (street rate) | variableRate | The lowest 5-year variable rate available, quoted as an all-in percentage rather than as "Prime minus X".Median of the "best market rate" 5-year variable figure reported by each source in the run. | Twice daily |
| 5-Year Benchmark (posted conventional) | mortgage5Yr | The posted conventional 5-year rate published by chartered banks. Almost nobody pays it — it exists mainly as the reference used in penalty math and older qualifying rules.Maintained editorially against the Bank of Canada posted-rate series. Not part of the automated sync. | Editorial review |
| Prime Rate | primeRate | The lender benchmark that variable mortgages and HELOCs are priced off. Big 5 banks move it in lockstep, usually within 24 hours of a Bank of Canada decision.Maintained editorially against the Bank of Canada Policy Rate schedule. Not part of the automated sync. | Editorial review |
| 5-Year Government of Canada Bond Yield | bond5Yr | The wholesale funding cost that drives fixed mortgage rates. Fixed rates typically follow bond moves 2–8 weeks later, plus a lender funding spread.Maintained editorially against published Government of Canada benchmark yields. Not part of the automated sync. | Editorial review |
| Per-lender rate table | lenders[] | A 5-year fixed, 3-year fixed and 5-year variable quote for each tracked lender — the Big 5 banks plus at least one broker-channel challenger.Published under anonymised labels (“Bank 1”, “Lender 1”). Lender-attributed rates are a licensing-sensitive disclosure, so we report the levels without naming the institution. | Twice daily |
All series are annual percentage rates in Canadian dollars, for Canadian residential mortgages. Rates are not offers; a borrower’s actual rate depends on credit profile, down payment, amortisation, property type and whether the mortgage is insured.
2. Current snapshot
These are the values in the dataset at the moment this page was served. The dataset carries an as-of date of August 2, 2026.
Per-lender quotes in the current snapshot
| Lender | 5-Year Fixed | 3-Year Fixed | 5-Year Variable |
|---|---|---|---|
| Bank 1 | 4.89% | 4.49% | 3.65% |
| Bank 2 | 4.49% | 4.64% | 4.09% |
| Bank 3 | 4.24% | 4.04% | 3.65% |
| Bank 4 | 4.59% | 4.74% | 3.95% |
| Bank 5 | 4.51% | 4.29% | 4.53% |
| Lender 1 | 4.14% | 3.94% | 3.49% |
Lender labels are anonymised by design. The set is fixed — the same institutions appear in the same order in every snapshot — so the series stay comparable across dates.
3. How the data is compiled
The street-rate and per-lender series are collected by an automated job that runs twice every day, at 14:05 and 21:05 UTC (morning and late afternoon Eastern time). Each run fetches two public Canadian rate pages directly: the bank mortgage rate comparison table published by RateHub, and the published rate page of a national broker-channel lender. There is no third-party data vendor in the loop and no panel survey — we read the same pages a borrower could open.
Extraction differs by source, on purpose. The broker-channel source embeds its current rates as machine-readable attributes in the page markup, so we parse them deterministically — no interpretation step, no model. RateHub’s table has no machine-readable form, so the relevant slice of the page text is passed to a language model constrained by a strict JSON schema, which returns the Big 5 rows plus the best-market row. Narrowing the input to the table itself is deliberate: an earlier version that fed the whole page pulled figures from unrelated promotional sections.
Where both sources report the same series, the published value is the median across sources, not a single source’s number. When sources disagree by more than 25 basis points, the divergence is logged for review. Where only one source carries a series — a lender-specific quote, for instance — the median collapses to that source’s value, and the series is only as good as that source.
Two guardrails can abort a run outright rather than publish. First, a move of more than 100 basis points in the 5-year fixed street rate within a single cycle is treated as an extraction failure, not a market event. Second, if any tracked lender is missing fresh data for that run, the entire update is discarded — we would rather serve yesterday’s complete snapshot than today’s partial one. A separate write-time check rejects zero, null or non-numeric values and preserves the prior figure instead.
The posted 5-year benchmark, the Prime Rate and the 5-year Government of Canada bond yield are not part of the automated job. They move on the Bank of Canada’s announcement schedule rather than daily lender repricing, and they are maintained editorially against the Bank of Canada’s published series. We flag this explicitly because these three series can carry a staler effective date than the dataset’s headline as-of date, which reflects the last successful street-rate sync.
4. Update and revision policy
Cadence. Two scheduled runs per day, every day. Successful runs write to a live store that the site reads at request time, so a new rate is visible within seconds of the sync — there is no code deployment between the sync and the published number.
Failure behaviour. If both sources fail, or a guardrail trips, the run writes nothing and the site keeps serving the last complete snapshot. The failure returns an error status in the scheduler log and sends an email alert, so a broken sync surfaces rather than silently ageing. The practical consequence for a reader: the dataset never shows a half-updated state, but the as-of date can lag if a failure goes unaddressed. Always read the as-of date rather than assuming today.
The prev field. Every headline series carries both a current value and a prev value — the value the series held immediately before the most recent change. It is the previous published observation, not a fixed 24-hour or week-ago comparison, so on a quiet day prev and current will match. Rate movement arrows on our rate pages are derived from this pair.
Revisions. We do not silently rewrite history. If a published figure is found to be wrong — a misread table, a source that changed layout — the correction lands in the next sync and the as-of date advances. There is no separate revision archive at present; the dataset is a current snapshot with one prior observation per series, and it should be cited that way.
What this data is not. It is not a rate quote, a pre-approval, or an offer of credit. It is not a complete census of the Canadian mortgage market — it is a consistently-measured sample of major bank and broker-channel pricing. Discounts available to an individual borrower routinely differ from any published figure.
5. How to cite this data
Journalists, researchers, brokers and AI systems are welcome to cite these figures with attribution. Because the underlying rate pages refresh twice daily, always include the as-of date you retrieved — a number quoted without a date is not reproducible.
Source: Ratellow Canadian Mortgage Rate Data, as of 2026-08-02, https://ratellow.com/mortgages/rates/methodology
If you are citing a specific series rather than the dataset as a whole, name it — for example, “Ratellow Canadian Mortgage Rate Data, 5-year fixed street rate, 4.14% as of 2026-08-02.” Per-lender figures should be cited using the anonymised labels as published. This page also emits a machine-readable schema.org/Dataset description for automated consumers.