Mortgage lead conversion rates: which ones have a source

The mortgage funnel has good public data at the bottom and almost none at the top. That is the single most useful thing to know before you plan against any of it, and it is the opposite of how the numbers are usually presented, where a contact rate and a funding rate sit in the same table in the same typeface as if they were equally solid.

So before the numbers, a sorting rule. Every figure in this market falls into one of three tiers by where it comes from, and the tier decides what you are allowed to do with it.

Green: application to funded

This is the only part of the funnel where an outside party counts the same thing twice a year and publishes the method. Two sources matter.

ICE Mortgage Technology, December 2021. Closing rate of 78.4 percent across all loans, 79.8 percent on purchase, 77.5 percent on refinance. The denominator is applications that reached a loan origination pipeline, measured on a 90 day rolling cycle across roughly four in five applications started in the origination software this report is drawn from.

Source: ICE Mortgage Technology (formerly Ellie Mae) Origination Insight Report, December 2021. The report is no longer publicly hosted, so there is no primary link to hand you and the year has to travel with the figure instead.

CFPB and HMDA, 2023. An origination rate of 57 percent, from about 5.7 million originations against roughly 10 million applications. On a narrower base, first-lien applications on owner-occupied single-family homes, the 2023 application mix was 73.9 percent purchase, 15.9 percent cash-out refinance and 6.9 percent rate and term refinance, with home improvement making up the rest. That last split is worth sitting with: cash-out was more than twice the size of classic rate and term refinance that year. The two counts are not the same population, so do not divide one into the other.

Sources: CFPB, "2023 Mortgage Market Activity and Trends", December 2024, for the application and origination counts, which are on page 7. files.consumerfinance.gov · Urban Institute Housing Finance Policy Center, "Housing Finance At A Glance", August 2024, page 39, for the application mix, from HMDA 2018 to 2023, first liens on owner-occupied single-family properties, shares by loan count. urban.org

Two things that go wrong with these two numbers

The denominators are not interchangeable. HMDA counts every application filed, including the ones a lender kills at intake. The ICE figure counts applications that already made it into a pipeline. A third number in circulation, a pull-through rate of roughly 75 percent attributed to trade association reporting, counts closings against applications submitted at a sample of independent mortgage banks. We cannot hand you a year for that one, because the underlying report sits behind a paywall and every version we found was a paraphrase of a paraphrase, which is reason enough to hold it loosely. Put any two of these three in one table and the table is wrong no matter which way it leans.

The ICE series appears to have stopped. The last data point anyone can produce is December 2021. Every "ICE closing rate" quoted for 2023 or later that we have chased has led back to that same 2021 report, or to nothing at all. If a deck hands you a 2025 closing rate with that name on it, ask for the issue date.

Amber: how many calls, and how fast

One dataset is large enough and mortgage-heavy enough to be worth using, and old enough that you have to say so out loud. Velocify studied 3.5 million leads and found that 93 percent of the leads that eventually converted were reached by the sixth attempt. It is the best available number for setting a call cadence, and we take it apart in how many calls it takes to reach a mortgage lead.

Source: Velocify study of 3.5 million leads, published around 2013. The original paper is no longer hosted; every version in circulation today is a secondary citation, so treat the figure as directional and dated.

The other study everybody in mortgage quotes is the 2007 lead response work out of MIT and InsideSales, which measured how sharply the odds of reaching somebody fall between a five minute and a thirty minute callback. It is a real study with a real population. It is also not a mortgage study, and it is quoted in this market as though it were.

Source: Oldroyd, MIT and InsideSales lead response management study, as presented at the MarketingSherpa Demand Generation Summit on 16 October 2007. The five minute against thirty minute comparison is on page 21. Cross-industry, not mortgage-specific. content.marketingsherpa.com

Red: contact rate, booking rate, preapproval rate

Here is the part that will cost you money if you plan against it. There is no independent, audited measurement of contact rate, booking rate or preapproval rate for internet mortgage leads. Not one. Every specific figure in circulation was published by a company that sells leads, sells software to people who buy leads, or writes pages designed to rank for the search you just made.

We are not reprinting those ranges here. Quoting one properly means naming the seller who published it, and quoting it without that attribution is the exact mechanism that turned these numbers into folklore in the first place. What is worth describing is how to spot them.

• Agreement between those pages is not confirmation. Several of them publish structurally near-identical benchmark articles with overlapping ranges and no primary citation. That is shared derivative origin, not independent replication.
• The stage rates do not reconcile with the headline. Take any of those pages, multiply its own stage by stage rates together, and compare the product against the same page's own lead to funded claim. On leads sold to a single buyer the multiplication comes out materially higher than the headline. On leads sold to several buyers it lines up. The exaggeration sits exactly where the premium is charged.
• Precision without method is a tell. A table of tidy ranges with no population, no date range and no definition of "contacted" is a design decision, not a measurement.

Figures with no primary source at all

Two claims deserve their own section because they are repeated by people who would never repeat a vendor statistic.

"78 percent of buyers go with whoever answers first." There is no study. Every citation leads to another article citing another article, with no population and no year anywhere in the chain. It is folklore that acquired a decimal point.

"The MBA says internet leads convert at 3 to 5 percent." The attribution to the Mortgage Bankers Association is everywhere and the underlying report is nowhere. We could not find it, and neither has anyone we have asked to produce it.

The measurement gap nobody has filled

Two conversion rates simply do not exist in public: contacted to preapproval and booked to preapproval. Not disputed, not contested, absent. Nobody publishes them, which is remarkable given that they are the two numbers a branch manager actually needs to size a lead budget, because they sit between the stage a desk controls and the stage the public data starts at.

The second gap is by product. There is no published funnel for DSCR at all, despite CoreLogic putting DSCR at 28.7 percent of non-QM production in December 2024, ahead of bank statement loans for the first time. The same reporting put non-QM borrowers at an average 776 FICO and about 75 percent LTV in 2024. A product that large with no published funnel is not an oversight anybody is in a hurry to correct.

Source: CoreLogic non-QM production analysis, December 2024. We could find no openly hosted copy of that analysis, only trade coverage of it, so there is nothing to link here either.

What to do with this on your own desk

• Plan the bottom of the funnel against the green tier, name the year every time, and keep the denominator written next to the number.
• Measure your own contact rate rather than adopting one. Yours is knowable from your own dialer inside a week; the published ones are not knowable at all.
• When a vendor quotes you a stage rate, ask for the population, the date range and the definition of the stage. The answer to that question is more informative than the number.

Two related pages: the note on the trigger lead ban that took effect in March 2026, and our FAQ, which answers the buying questions rather than the research ones.