Why brokers lose clients at refinance, and what retention software actually does

The mechanic behind the loss

Very few clients leave a broker because of the service. They leave because a number moved and they heard about it from someone who was not you.

The ACCC put figures on that in its Home Loan Price Inquiry final report, published in November 2020. As at September 2020, borrowers with variable rate owner occupier loans between three and five years old were on average paying around 58 basis points above the average interest rate for new loans. Between five and ten years old, around 71 basis points. Older than ten years, around 104 basis points. The inquiry also found that as at December 2019, almost half of all variable rate loans had been originated at least four years earlier.

The ACCC's own worked example is the one to keep in your head. On a loan of around $250,000, moving to a rate 58 basis points lower saved over $1,400 in interest in the first year, and over $17,000 across the remaining term in net present value terms. On a $500,000 loan, over $2,800 in the first year and over $34,000 over the remaining term.

Those specific figures come from a 2020 rate environment and should not be quoted as current pricing. The structure they describe has not changed. Front book pricing is sharper than back book pricing, and the gap tends to widen with the age of the loan. Every month a client sits on an untouched rate, the case for somebody else's phone call gets stronger.

Refinancing is not a spike, it is the weather

ABS Lending Indicators for the June quarter 2026, released on 14 August 2026, recorded 66,449 owner occupier external refinances worth $41.9 billion, down 0.9 per cent on the quarter and 1.1 per cent on the year. Investor external refinancing was 36,597 loans worth $25.2 billion, down 2.3 per cent on the quarter and up 5.3 per cent on the year. Total new housing loan commitments were 134,225 loans worth $97.6 billion.

The ABS reports new loan commitments excluding refinancing, so those are two separate flows rather than a total and a subset. Set side by side, external refinancing in that quarter was worth close to seventy per cent of all new housing lending. It moves with the rate cycle, but it does not stop. Planning your book on the assumption that a quiet refinancing period will protect it is planning against the data.

The ABS also records, in the quality notes attached to that release, that data quality concerns remain in how lenders are reporting the value of internal refinancing of housing finance loans. That matters here, because internal refinancing is the repricing and product switching that happens without the loan leaving the lender, and it is the outcome you usually want. It is also the flow the published numbers describe least well.

Record broker share raises the stakes on your back book

MFAA data reported in June 2026 put brokers at 81 per cent of all new residential home loans in the March 2026 quarter. That is up from 76.8 per cent a year earlier and 74.1 per cent in March 2024, and it beats the previous record of 76.7 per cent set in the December 2025 quarter. The channel settled $124.88 billion in the March 2026 quarter, an increase of $25.51 billion on the same quarter a year earlier.

MFAA chief executive Anja Pannek noted the channel has grown from 55.3 per cent in March 2018 to 81 per cent in March 2026.

The implication for retention is direct. When the broker channel writes four fifths of new lending, most refinancing is broker originated, which means most refinancing away from you is another broker's win rather than a bank branch's. Your trail book is somebody else's prospect list, and they have the same tools you do.

The three ways a client actually leaves

The first is the lender's own retention team. The Adviser reported in April 2026 that ANZ was offering retention payments of around $2,000 plus rate discounts of 15 to 20 basis points to borrowers who had already lodged discharge forms to refinance elsewhere. Brokers quoted in that article described putting 10 to 15 hours of unpaid work into a refinance file that then does not settle. Matt Turner of GSC Finance was quoted saying it shows the retention system is flawed. Whatever your view on the practice, note what it tells you about timing. The lender's best price often appears only at the point of departure.

The second is another broker. Outreach against an aggregated trail book is cheap and increasingly automated, and it does not require the other broker to know anything about your client except their approximate loan age and lender.

The third is the client themselves, in an app, comparing their rate to an advertised one and drawing a conclusion.

In all three cases the trigger is identical. The client learned that their rate was uncompetitive from a source that was not their broker. Retention work is, at bottom, the practice of being that source first.

What retention software actually does

Strip the marketing away and products in this category do some combination of four things. They hold a current picture of every loan in your book with its rate. They compare that rate against a reference point, usually what the same lender is writing now or what the broader market offers. They rank or score which clients are most exposed. And they either trigger a repricing request with the existing lender or hand you a refinance opportunity.

Sherlok is the most visible Australian example and is positioned as an automated repricing and refinancing retention platform that predicts loan book churn and reprices on the broker's behalf. AFG announced a preferred partner arrangement with it, giving AFG brokers integration and discounted subscription pricing.

Sherlok's own Home Loan Pricing Spotlight, reported by The Adviser in April 2024 and drawn from 36,546 processed reprices between July and December 2023, published an average rate reduction of 0.46 per cent, annual interest savings of $2,070 per home loan on an assumed $450,000 average loan, a 32 per cent reduction in client churn compared with brokers who did not use the technology, and an additional $50,816 of revenue per broker per year. Those are the vendor's own figures on its own users. Finance OS, which publishes this article and is building a comparison product for broker software, has not tested Sherlok or any other tool named here.

On churn generally, Sherlok founder Adam Grocke was reported in August 2023 putting broker churn at around 17 per cent, against CBA at approximately 20.7 per cent, or around 18 per cent excluding amortisation, and predicting broker churn would rise toward 20 per cent. In March 2023 he put average loan life at 37 months, based on Sherlok data covering brokers with $70 billion in mortgages. Again, one vendor's dataset rather than an industry statistic.

What it does not do

It does not make the call. Every product in this category produces a list. The outcome is decided in the conversation, not in the scoring model. A broker who does not action the list gets a more precise view of clients they are about to lose.

It does not create pricing that is not there. If the lender will not move, the tool has told you something useful and then stopped. What you do with that, whether you refinance or hold and revisit, is judgement.

It does not build your best interests duty record. An automated repricing request is not evidence that you considered the client's circumstances. RG 273 expects records showing the options and recommendation you gave and why, including your decision making process, and it favours notes taken contemporaneously. If the software sends the email, you still own the file note.

And it does not solve clawback. Knowing a client is at risk only helps if you reach them before they lodge a discharge form somewhere else. The tool buys you notice, not immunity.

What to check before you buy

Where does the rate data come from and how often does it refresh. A comparison run against stale advertised rates produces confident wrong answers, and confident wrong answers sent to a client under your name are worse than silence.

How does it connect to your book. Aggregator commission file, direct CRM sync, or Consumer Data Right access. If the tool uses CDR data, understand the basis on which you receive it. Mortgage brokers are one of the trusted adviser classes named in the CDR Rules, and once CDR data is disclosed to a trusted adviser the OAIC is explicit that it is no longer subject to the protections and safeguards of the CDR system unless that adviser is separately accredited. From that point the data is yours to protect under the Privacy Act.

What actually goes out under your name. Read the repricing template and the client email templates before you switch anything on. A repricing request submitted to a lender under your credit representative number is your communication, and so is an automated message telling a client their rate is uncompetitive.

What happens to the client list when you leave. Ask about export in a usable format, and about deletion.

And check whether your aggregator already provides a version of this. Several run rate alert or book health programs. Paying separately for an alert you already receive is a common and avoidable cost.

The version you can build without buying anything

Export your book. Add a column for settlement date and a column for current rate. Add a rule that says any loan sitting more than a set margin above what that same lender is writing today gets a call this month, starting with the largest balances and the oldest loans.

That is the entire mechanic. Software makes it continuous rather than quarterly, and it removes the manual comparison step, which is genuinely the tedious part. What it does not do is change the underlying activity, which is a broker contacting a client about their rate before somebody else does.

If you are not running the manual version now, the honest expectation to set is that software will not rescue the book on its own. It will give you a longer, better ordered list of clients you are not calling. The tooling decision is worth making after the habit exists, not instead of it.