Transparency
How we collect the data
How we collect, check and publish bank interest rates.
The data pipeline
- Bank sources32 published pages
- Change detectionagainst the previous read
- Extraction3128 rows in the data edition
- Automated checksplausible range · real terms
- Source cross-checkconfirmed only when two sources agree
- Manual reviewoutliers wait for a human
- Publicationwith source and read time
A figure that fails a check stops at that step — nothing goes live automatically.
Data health
Official sources
32banks
read straight from the official site
Press sources
7banks
secondary reference · last read 24 Sep 2026
Data rows
3128rows
data edition 25 Sep 2026
Approved market rates
1447rates
multi-source survey, each with a read time
-
01 Published
3128 rows
Highest — opens the exact page we read, with the time we read it
The bank's own rate page, data interface or PDF rate sheet
-
02 Press sources
30 rows
Secondary reference — the product, channel or timing may differ from the bank's own schedule; not treated as a confirmed rate until cross-checked
News reports quoting the rate sheet, used when the bank's own page cannot be read
-
03 Market
1447 approved rows
Lowest — the bank does NOT publish this rate; always carries its source label
Public social-media groups, every row cleared by a human reviewer
The bars show how far the three layers can be trusted relative to one another, not an absolute scale. The row counts come from the data edition shown on this page.
Published rates
Taken from the bank itself in three ways, most trustworthy first: a direct query to the bank's public data interface, a reading of the fully rendered rate page, and a reading of the PDF rate sheet the bank issues. Every figure stores its source link and the moment it was read.
Market rates
The rate actually being quoted in the market, collected from public social-media groups. This is information shared by depositors and by bank staff — the rate bank staff quote, not the published rate. We always show it with a confidence label and the number of independent sources.
Checks before anything is published
Automated rule set
Every figure passes an automated rule set before it is shown: a plausible range for the product type, a term that really exists, the size of the move against the previous publication, and the distance from where other banks sit. A figure that fails any rule goes to manual review instead of going live.
Rates with conditions
Figures carrying the words "up to", "maximum" or "from" are always flagged as not fully verified and shown with the original conditions. We do not compress them into a bare number cut loose from the terms that apply.
Collected but not published
Foreign currency: collected, kept out of the comparison
We do collect foreign-currency deposit rates as well — 1,521 rows across EUR, USD, AUD, CAD and GBP — but we keep them out of the comparison table. The whole group sits at just 0–2.2%/yr, and USD is capped at 0%/yr, so every bank is level; a ranking of nothing but zeros helps no depositor decide anything. The figures stay in the archive and will be published once there is a way to present them that carries real comparative value.
What we do not do
We do not invent figures
Nothing is invented and nothing is estimated in place of a real figure. Where data is missing, the cell simply shows no number.
We do not mix in bonds
Corporate bond yields are never used in place of deposit rates — they are different instruments, and bonds carry no deposit insurance.
We do not mask digits
We never show masked figures of the "9.x%" kind — a number like that cannot be compared, so it does not count as data.
How often it updates
Published rates — daily
Published rates are checked every day. The system compares the content against the previous reading and processes only what changed.
Market rates — several times a day
Market rates are collected several times a day from the most active sources. The update date in every page footer is the moment the data was actually read, not the day the page was rebuilt.