Home-loan pricing intelligence for Australian banks

Where does our home-loan pricing sit against our competitors today, and is our own CDR data right?

RateRadar reads the public Consumer Data Right (CDR) product feed of 15 lenders every hour. It applies each lender's discount rows the way a borrower sees the rate, checks the result against the lender's own website, and tells you where you rank and where your own feed is wrong.

Real example from the feasibility test

A customer-owned bank in our test: one variable home-loan rate in its own CDR feed

Rate a feed reader seesThe raw rate field, with no discount applied

Rate the feed impliesBase rate minus the matching discount row

Comparison rate in the same rowA comparison rate below the rate is a sign of an error

Why it happens: the bank publishes a base rate plus a separate discount row, and the two are paired by tier name. On this rate row, the tier name belongs to a different product family in the bank's range, while its discount row carries the correct name. The two no longer pair, so any comparison site or aggregator that reads the feed shows a rate about percentage points too high.

What the bank gets: the exact product, row and fix, sent privately to its CDR compliance team. We do not publish one bank's defects to others.

1. Competitor grid

Lowest advertised owner-occupier rate, principal and interest, by LVR

Each cell is the lowest rate any of the lender's products offers a borrower at that loan-to-value ratio (LVR). Select a cell to see the product, the raw feed value, how we reconciled it, and links to check it.

Show ranks for

Website-verified: this exact rate row matched the lender's website in the test CDR only: from the lender's CDR feed, not yet checked against its website 6.19 lowest in column

2. Data quality

How complete each feed is, and why raw CDR rates mislead

CDR feeds follow one standard, but lenders fill it in different ways. These checks run on every snapshot, for competitors and for your own feed.

Share of rate rows with an LVR tier

Without an LVR tier, a rate cannot be placed in the grid by LVR. Amber bars are below 80%.

Own-data check: how it works

We ran it on one bank's full home-loan feed in the test: all of its CDR price rows, compared with its website. All 53 website rates were found in the feed, and 2 feed defects turned up.

Read the raw rate field and the feed agrees with the website times out of . Apply the discount rows and it agrees out of .

further rows are extended-term, switching or bridging variants that the website does not publish.
rows carry a tier name from the wrong product family. In one, this breaks the discount pairing (the example at the top of this page).
Your CDR compliance team gets this list with the exact product ID, row and fix.

3. Change log

All 15 lenders read One lender failed, previous data kept Westpac returned an empty product list No run in that hour Full re-fetch of every product

4. How much to trust this

What was checked, and the results

35 / 35checkable rates matched the lender's website

15 / 50sampled rates could not be checked

53 / 53website rates of one bank found in its own feed

We drew a random sample of 50 rates across all 15 lenders and compared each with the lender's public rate page, captured within minutes of the CDR pull. All 35 that could be checked matched.
Of the 15 we could not check, 12 are on pages that only show rates after JavaScript runs (Teachers Mutual Bank, BOQ, IMB Bank, Heritage), 2 are switching-only rates that one bank's website does not publish, and 1 is a Westpac product its website does not name.
These checks were done by AI agents against the lenders' CDR feeds and public websites, not by a human auditor. The agent that built the pipeline also judged the 50 matches. An independent AI reviewer, briefed as a bank pricing head, then audited 6 rows. The evidence for every row is kept.

Known limits

No multi-year history yet. Hourly snapshots started on 24 Oct 2026.
Detection speed has not yet been measured across an RBA cash-rate move.
"Lowest advertised" can be a conditional offer (online-only, minimum loan size, package). The product name is shown in every cell.
With 50 samples, one mismatch would drop verified agreement to 97%.

5. What a pilot looks like

Three months, cancellable

A$1,500per month for the 3-month pilot

A$2,500per month after the pilot

Daily competitor grid for 15 competitors you name, by product type and LVR, with every cell traced to its source.
Monthly own-data CDR accuracy check for your compliance team: every price row compared with your website, with the defects listed by product ID.
Change alerts when a named competitor moves a rate, with the time we first saw it.
You can cancel at any time during the pilot.

To start a pilot or ask a question, reply to the email this link came in.