The Mortgage Funnel Doesn’t Work the Way Most Lenders Assume

Most cycle-time and conversion assumptions come from pipeline reviews, loan officer feedback, and whatever the last dashboard happened to show. 

That’s useful information, but it’s built one loan and one branch at a time. It rarely matches what shows up when you pull twelve months of originations data across an entire lender base and look at the patterns underneath.

With 15 years of borrower data covering 50 million home loans, from application to close, Blend is in a position to test what the industry assumes. So we did.

We used the Blend Autopilot Analytics Agent to benchmark twelve months of device behavior, application timing, and conversion by institution type across lenders on Blend’s platform, then set the results next to what a panel of senior mortgage executives expected going in.

The two rarely lined up, starting with something as basic as where a borrower is most likely to actually finish an application.

Here’s what the data showed us:

More borrowers start on mobile, but that’s not where the loan closes

Mobile is where most applications begin. Borrowers are more likely to open a loan application on their phone than on a laptop. Look at which applications actually get completed, though, and desktop pulls ahead.

The applications that do get finished on mobile move faster once they’re underway. Across Blend’s origination data, mobile applications average around 20 minutes to complete, compared to 27 minutes on desktop. That gap makes mobile look like the channel to build around.

A strategy built around where applications start is optimizing for the wrong end of the funnel.

The friction isn’t the device, it’s moving between them

The bigger cost in a mortgage application comes from switching devices mid-application, not from which device a borrower picks first.

Looking at how long completion actually takes by device path, borrowers who start and finish on the same device move noticeably faster than those who don’t: about 23 minutes for mobile-to-mobile, and 30 minutes for desktop-to-desktop. Once a borrower switches devices partway through, that time increases substantially, and going from mobile to desktop costs more time than going the other direction. 

A borrower who picks up their phone to finish what they started on a laptop has to re-enter details the application failed to save, costing them minutes.

Borrowers aren’t applying at 3 a.m. as often as you’d expect

Late-night and overnight self-service is one of the more visible parts of a digital application experience. Loan officers see the timestamps. It’s easy to assume borrowers are engaging around the clock.

The aggregate numbers tell a narrower story, and it splits three ways depending on how the borrower got there. First, across the full portfolio, applications started between midnight and 6 a.m. on weekdays make up a small share of total volume, around 2.4%. Second, when a loan officer invites the borrower directly, the application activity clusters during business hours, which tracks with when that outreach actually happens. Third, applications driven by marketing skew toward evenings instead, with a meaningful share of that channel’s weekday volume landing between 5 p.m. and midnight. 

The overnight hours aren’t where the volume is. For marketing-driven applications, the evening, between 5 p.m. and midnight, carries a far bigger share than the middle of the night. That makes campaign send times and follow-up timing worth checking against when borrowers actually start.

Conversion and speed don’t sort by institution type the way you’d guess

The same “when” question applies to “who.” Independent mortgage banks have a reputation for moving fast and converting well, built largely on the assumption that a singular focus on mortgage translates directly into funnel performance. 

Blend’s benchmarking across its lender base breaks that assumption into three separate findings. Regional banks converted best overall. National banks moved fastest by median completion time. Credit unions trailed both measures. Independent mortgage banks led neither one. 

That ranking shifts again once you isolate the lead source. There’s no single segment that wins the funnel. There’s a different winner depending on whether you’re measuring speed, conversion, or channel, and most benchmarking conversations only ask one of those questions at a time.

The data shows what happened, not why it happened

That gap between what the panel expected and what the aggregate data showed points to a bigger problem underneath both findings above. Knowing where friction shows up in the funnel is not the same as knowing why it’s there. A lender can see that a given segment lags on conversion without knowing whether the cause is marketing, staffing, or a gap in the product experience itself, and building a fix on the wrong guess spends effort without closing the gap.

That’s the harder half of the work, and it’s also where the clearest example in the data shows what closing the gap actually looks like.

Pattern-level data tells a lender where to look. Finding out why still takes someone asking the next question.

One lender pushed conversion to 90% by changing who talks to the borrower first

That distinction between seeing a pattern and understanding its cause shows up clearly in one outlier Mukesh Jha of Blend identified while benchmarking across lenders. Among institutions with meaningfully lower conversion, one stood out at roughly 90%. 

The difference came down to timing, not technology or spend. Loan officers called borrowers early and walked them through the process before confusion had a chance to cause a drop-off. 

The highest-converting lender in the data won by answering the borrower’s confusion before it turned into an abandoned application, not by out-building anyone on technology.

Find where your funnel is losing borrowers

None of these gaps close by copying a device strategy or benchmarking against whichever institution type looks best this year. They close by asking, for each specific point where borrowers drop off, whether the fix is a channel problem, a timing problem, or a handoff problem, because those three get confused for each other constantly.

The key takeaway for lenders is understanding which moment in your funnel is losing borrowers, and whether the fix you’re funding is aimed at that moment.The real challenge is whether you have a plan for what happens in the 60 days after.

Curious what twelve months of your own origination data would show against these patterns?

• See how the Blend Autopilot Analytics Agent surfaces funnel gaps like these →