管理層發言
Hello, and thank you for being here. My name is Lacy, and I will be your conference operator today. I would like to welcome everyone to the Better Home & Finance Holding Company Second Quarter 2025 Results Conference Call. Thank you. I will now turn the conference over to Tarek Afifi. You may begin.
Welcome to Better Home & Finance Holding Company's Second Quarter Earnings Conference Call. My name is Tarek Afifi, Corporate Finance at Better. Joining me on today's call are Vishal Garg, Founder and Chief Executive Officer of Better; and Kevin Ryan, Chief Financial Officer of Better. In addition to this conference call, please direct your attention to our second quarter earnings release, which is available on our Investor Relations website. Also available on our website is an investor presentation. Certain statements we make today may constitute forward-looking statements within the meaning of federal securities laws that are based on current expectations and assumptions. These expectations and assumptions are subject to risks, uncertainties, and other factors as discussed further in our SEC filings that could cause our actual results to differ materially from our historical results. We assume no responsibility to update forward-looking statements other than as required by law.
During today's discussion, management will discuss certain non-GAAP financial measures, which we believe are relevant in assessing the company's financial performance. These non-GAAP financial measures should not be considered replacements for and should be read together with our GAAP results. These non-GAAP financial measures are reconciled to GAAP financial measures in today's earnings release and investor presentation, both of which are available on the Investor Relations section of Better's website and when filed in our quarterly report on Form 10-Q filed with the SEC. Amounts described as of and for the quarter ended June 30, 2025, represent a preliminary estimate as of the date of this earnings release and may be revised upon filing our quarterly report on Form 10-Q with the SEC. More information as of and for the quarter ended June 30, 2025, will be provided upon filing our quarterly report on Form 10-Q with the SEC. I will now turn the call over to Vishal.
Thank you, Tarek, and welcome to our second quarter 2025 earnings call. We appreciate everyone joining us today and for your continued support as we advance our mission to make homeownership better, faster, and easier for our customers by building an AI-native technology platform that revolutionizes the entire homeownership journey, helping consumers go through the entire mortgage and home equity process in as little as one day. We continue to drive progress toward this vision in which every customer can seamlessly buy, sell, refinance, insure, and improve their home online instantly at a competitive price. These objectives are: one, leaning into growth in AI to drive increased volume and revenue; two, continuously improving our efficiency driven by ongoing advancements in our technology; three, diversifying our product and go-to-market strategies along with our distribution channels; and four, reducing our corporate costs as a percentage of our revenue, all with the goal of achieving profitability.
Better is the first scaled mortgage tech platform built to empower consumers and more recently, empower local mortgage bankers and financial institutions with Tinman technology to serve their end customer needs. This sets us apart from the rest of the mortgage industry and in the face of market challenges creates tremendous greenfield opportunity for Better. We remain focused on driving towards profitability by continuing to lean into Tinman technology and AI with Betsy executing approximately 600,000 consumer interactions in Q2, our AI underwriting growing to over 43% of locked loans with a clear path to 75% in the near future and our loan officer productivity in terms of funds per month increasing to over 3 times the mortgage industry median. We've been talking about our path to profitability in the medium term for some time. With all of the advancements in our AI platform and the progress we are seeing in bringing on not just mortgage advisers that are local market professionals, but also other large-scale enterprises that want to enter the mortgage business through our platform, I am pleased to share that we now have the visibility and expectation to achieve adjusted EBITDA breakeven by the third quarter of 2026, basically within the next year.
Kevin will speak more on this shortly. We have built a platform that is AI-first. While consumer adoption behavior of AI takes time, it is increasing at an exponential rate. We are one of the few mortgage companies, if not the only mortgage company in the U.S., with a full-scale tech stack in one place, all in one flow and entirely APIable to Agentic AI. Betsy has been built from scratch on our knowledge base that has been developed on 12 million-plus recorded phone calls, 500,000-plus funded loan files with the entire consumer financial graph and property graph matched to the investors' detailed line item fact-based criteria and 6 million pre-applications over the past eight years, and it is continuing to learn every day from every customer interaction. Since we launched Betsy AI, our lead-to-lock conversion rate has increased by over 30% from 3.3% to 4.4%, which is massively meaningful to drive incremental volume and revenue and squeezing profitability out of each loan.
Now, as you can tell, we've still got a really long way to go to keep on improving that lead-to-lock conversion rate. But as we scale Betsy at near zero marginal cost because it's all built on our own internal proprietary Tinman platform, we believe that we can further improve our unit economics as the mortgage market stabilizes over time with relatively minimal increases in fixed CapEx. Betsy has led to an even better customer experience. For customers who opt to speak with a human, Betsy will help find the most talented individual on our sales team for that customer to interact with for their specific needs. Since we've adopted Betsy, our Net Promoter Score has increased from 39 to 64, putting our score in line with companies like Google and Apple and far superior to that of traditional mortgage companies and financial institutions. Betsy is built on top of our machine learning pricing and eligibility engine, which we internally call DE, which uses the consumer's financial graph and the property graph, and each specific data element is matched to the criteria set by each individual mortgage investor directly within our platform.
It is able to run hundreds of thousands of pricing permutations across the now over 45 mortgage investors who buy the loans on our platform as they are funded on a loan-by-loan basis. These loan purchases include GSEs, banks, REITs, and increasingly private credit funds, all of whom bring different guidelines, risk-return preferences, and other criteria to our platform. They represent over $1 trillion in demand, bidding on loans each day that match the criteria and portfolio needs, and with our platform fulfilling those criteria at an error rate that is substantially lower than the industry's. Over the next one to two years, we expect to significantly grow our investor network, particularly as the private credit market moves further into consumer asset-based finance. Within consumer lending, mortgage and home equity is over 85% of the total addressable market and is particularly attractive to private credit as it has a robust track record in recessionary periods on a basis relative to unsecured consumer debt like personal loans, credit cards, and BNPL loans.
Our credit and underwriting quality are best-in-class. Yet aside from the short window that we hold the loans due to settlement, we do not take any credit risk as we have forward commitments to sell each loan to investors the moment they are locked and funded. This model remains unique in the fintech landscape, where a very large number of the originators that started out as marketplaces have now moved on to actually holding credit risk on their books. Better is doing exactly the same thing today as it did when it opened its doors in January of 2016. We do not hold credit risk on the assets that we originate. And we believe this will serve us well if we enter into a recessionary climate. As we look to the second half of 2025 and beyond, our strategic priorities remain focused on what lies in our control. Our first priority is to continue to propel growth opportunities independent of broader economic and mortgage market conditions.
In the second quarter of 2025, on a year-over-year basis, we grew funded loan volume by 25% to $1.2 billion and revenue by 37% to $44.1 million, driven by funding more loans, both through our D2C and Tinman AI platform channels at higher gain on sale compared to the period past. During the quarter, year-on-year funded loan volume growth was driven by HELOC and home equity loans increasing by 166%, refinance loan volume increasing by 109%, and purchase loan volume increasing by 1%. Our overall growth in the quarter is attributable to the strategic investments we've made in technology, product innovation, and distribution expansion, including Betsy AI, our Tinman AI platform strategy with NEO powered by Better being proven out, and the efficient expansion of D2C driven by the efficiencies we are gaining implementing AI throughout the entire platform. These strategic initiatives have positioned us to capitalize on a broader set of market opportunities, enhance our operational efficiency, and continue to drive sustainable growth with revenue growth outstripping loan origination volume growth, both in the quarter that just passed and what we expect to be in the quarters to come.
Our second priority on the path to profitability is to continue to reduce our expenses and improve our operational efficiency with the goal of reaching adjusted EBITDA breakeven by Q3 2026, essentially within the next 12 months. With our Tinman AI platform, we have been able to automate time and labor-intensive components of the mortgage process and continuously reduce our cost to originate to approximately half of the industry average. Over the coming 12 months, we expect to drive that even further to have it reach approximately one third of the industry average and then continue to keep growing as the consumer adoption of AI and interaction with AI continue to improve our ability to take costs out of the manufacturing process of mortgages. While we expect loan origination expenses will increase as we grow volume, we believe our continued investments in AI with our product and engineering roadmaps well on track will significantly drive down costs further, resulting in improved operating efficiency and unit economics.
Lastly, our third priority is to continue diversifying our product and platform distribution channels. We serve the end consumer now, both through our direct-to-consumer model and through our AI platform model, which includes Tinman as a platform and Tinman as software. In our D2C business, we serve the consumer directly on better.com. We were founded on revolutionizing the consumer experience for the home finance process. And as such, our D2C business has always been at the forefront of pushing the envelope on what technology can execute in the mortgage industry. We have funded over $100 billion in loans on the D2C platform, which has served as a basis for training the AI and continuously creates a positive feedback loop, improving each and every day. Our D2C business allows us the opportunity to roll out and test new AI features at a scale and speed that others simply cannot match. And within our D2C unit channel, our unit economics continue to improve as we continue to drive AI throughout the entire process.
As you can see in the earnings deck that we've distributed along with our materials, our contribution margin or per loan profitability has continued to increase as the operating cost of fund has continued to decrease, driven by both conversion gains and the implementation of AI in the sales and operations workflows. More specifically, in Q2 2025 for D2C, revenue per loan was $78.86 per loan. Our cost per fund was $68.22 for a contribution profit of $1,064 and a contribution margin of 13%. Again, we have continued to optimize our pricing so that we remain competitive with the major players in the market. But as you can see, our total cost to originate is about half that of the mortgage industry average, which includes our customer acquisition costs. And we expect to continue to drive that down as we both increase conversion and lower CAC, and improve labor costs and thereby lower the cost to fund significantly.
Next, we serve the customer through our Tinman AI platform, powering local retail loan officers across the United States, for which we continue to see early rapid growth. In this model, we are effectively serving as a platform provider and P&L partner in the mortgage origination process with nearly zero customer acquisition cost because the mortgage retail loan officers bring their customers, their relationships, and their transactions to bear on our Tinman AI platform. We are quickly disrupting the traditional retail mortgage origination market by onboarding loan officers and branches onto our Tinman AI platform, empowering them to do more loans than they've ever done before, remove friction from their fulfillment process, and expand their capacity to help more customers through the lead flow that we generate from. We expect over time that these loan officers will be able to compress a staggering 80% of their back-office costs by using our platform.
The market for the Tinman AI platform business is massive, and we are just getting started. For context, over $1.2 trillion of mortgage volume in 2024 was originated by retail loan officers and mortgage brokers on antiquated technology and high operating costs. Just 1% of that market would translate to $12 billion in new loan volume, with again, nearly zero customer acquisition cost to better. We continue to make great progress on the Tinman AI platform with our first and now well-proven launch of NEO powered by Better. We began production with NEO at the start of 2025 and have high aspirations for the road ahead. In Q2, we funded $429 million loans for 1,009 families with NEO powered by Better, an increase of 164% and 176%, respectively, compared with the prior quarter. The unit economics of our Tinman AI platform are quite strong. Specifically, in Q2 2025, for every loan funded on the Tinman AI platform on NEO powered by Better, we generated a contribution profit of $6,172 on a revenue per loan of $15,538, resulting in a contribution margin of 40% for the Tinman AI platform.
As we are able to increase penetration of the Tinman AI platform's processes for the retail loan officers, we expect that margin to increase even further, enabling us to either compensate these retail loan officers more handsomely or help drive more profits for them and their branches over time, which we believe, combined together, create a kind of holy trifecta. The Tinman AI platform enables retail loan officers to do more loans, serve more customers while working the same hours, and with a lower cost to fund, resulting in dramatic increases in profitability for their business versus being on a traditional retail platform or mortgage broker platform. It is this trifecta that is allowing us to recruit additional local loan officers from highly successful retail mortgage companies, including loanDepot, Nationwide, and Movement Mortgage, just to name a few. Just this last quarter, we onboarded loan officers from these companies that funded over $180 million last year in loan volume.
And going forward, we expect to attract even more of these talented high-volume mortgage loan officers in the retail channel due to the superior technology and model offered to them by our platform. Furthermore, the specialized nature of these loan officers is broadening our reach into new loan types that are more nuanced or complex but come with higher margins. These loan types include non-conventional FHA, VA, and jumbo loans. As we continue to expand with NEO powered by Better and the Tinman AI platform at large, we expect to do more of these specialized loan types with higher gain on sale margins and deeper efficiency through our tech. As we have proven the Tinman AI platform with NEO powered by Better with the entire mortgage industry watching, we have been inundated with other mortgage teams and companies wanting to move their business to the Tinman AI platform. We see massive opportunity in the road ahead with other traditional mortgage originators and are making solid progress executing on a robust pipeline of future clients and partners.
I'm proud of the independent achievements we have made with both our D2C and Tinman AI channels. And now to take it a step further, we are working to bridge the two together. I'm particularly excited by the testing we are now conducting, whereby our AI selectively matches pre-approved D2C purchase customers based on a full set of parameters about them and the property that they're buying with localized NEO powered by Better loan officers who are experts in their particular geographic areas. These loan officers also have terrific relationships established over decades with the realtor community in these markets, and so are able to easily integrate with the existing workflows that those realtors prefer. Once implemented at scale, we see the potential for a significant increase in conversion and incremental volume and revenue at very healthy margins for both the D2C business and the NEO powered by Better business.
While we are still in beta mode, we believe this approach can further improve the unit economics of our D2C business over time by generating revenue from customers who would otherwise have not converted on our online platform. In other words, these pre-approved leads would have come in through D2C self-serve, realized that they prefer a high-touch experience with a local market expert, and thus been at risk of falling out. Now the AI is selectively determining on an individual basis if that lead would have been more likely to convert with a local market expert on the Tinman AI platform. And our matching algorithm will continue to learn and improve over the next six to twelve months, just in time for the 2026 purchase season. Just to give you a reminder, Better generates over 250,000 pre-approved mortgage customers every year. When applied to our average dollar loan volume of nearly $300,000, this implies a total volume potential of $75 billion.
And yet our total market share of actually funded purchase originations is currently less than 50 basis points. Our market share by home shoppers per year as a function of the amount of loans that we are preapproving is actually over 10 times that, nearly 5%. This showcases the massive opportunity we have by opening a different method of conversion for the customers that come to us for a preapproval. And coming back to our multi-pronged distribution, we are serving the customer by also powering banks, credit unions, and other larger mortgage originators that are seeking to license our Tinman AI software to become more efficient and customer-centric. We have built a highly fine-tuned platform for our own business and customers, and we are now seeing demand from others in the industry to directly license our software to use in their own businesses. A lot of banks and credit unions are taking a refreshed look at the mortgage space as the regulatory environment is becoming more favorable.
However, bank origination of mortgages has largely been unprofitable given their higher cost to originate. This is where our Tinman AI software comes in. Our Tinman AI software essentially provides mortgage in a box, enabling banks to not only use our software, but also gain access to underwriting resources and sales resources if they so desire. By using the Tinman AI software, banks and credit unions no longer need to invest in a variety of different systems, pay millions of dollars in system integration costs, upfront costs, implementation costs, seat licenses, or any of those sorts of things. Gone are the days of eight different software platforms sort of speaking to each other, stitched together by middleware, by consultants that charge hundreds of thousands of dollars just to integrate these systems, and with no unified data set for any type of machine learning to operate on. We are changing the industry altogether.
What's more, our pricing model is fundamentally different. It is what is now coming to be in vogue amongst AI companies as outcome as a service. We are simply charging our customers on a funded loan basis, so that their cost event for processing, fulfilling, underwriting, even selling a loan becomes directly tied to the revenue event and takes the risk out of the transaction for these partners, which is a massive differentiator to the traditional software pricing model in the mortgage industry and in financial services in general. We are pleased that our first bank partner on the Tinman AI software platform has begun funding loans on our platform. In this partnership, we are powering their entire mortgage platform from a software perspective, from click to close with their sales and operational personnel across the full range of products that they offer, including non-QM and other niche products entirely on Tinman.
To get them up and running, not just for conforming FHA VA but also for niche non-QM products, took just under three months from LOI signing to loans flowing through our software. And we believe this can ramp to over $4 million a month in monthly revenue in the near medium term, based on the pipeline that we have. We believe an even larger addressable market exists within the mortgage ecosystem for a holistic one-stop software solution powered by the industry's leading AI engine, Tinman. To put the opportunity into context, over 5 million mortgages were built on the Encompass platform in 2024. To the extent that we can achieve even 1% penetration of the Encompass customer base based on our current pricing, we believe that could drive an incremental 50,000 new loans and $75 million of revenue at software margins to better per year. We have an extremely strong pipeline of partners who want to execute either their entry into the mortgage business or their growth in the mortgage business by leveraging the Tinman platform, which provides them the ability to scale nearly infinitely in terms of sales and underwriting without having to scale people, which is what the entire mortgage industry hurt from the bust that came in the period of 2022 to 2024.
Most mortgage CEOs do not want to repeat what they had to go through during that time. And thankfully, Betsy enables them to have nearly infinite scale featured parity with a traditional loan officer doing refinances and the ability to flex up or flex down their marketing spend to meet the balance sheet needs that they have. That is unique to the mortgage industry. It is unique within consumer lending, and we are bringing that power across the board to some of the largest financial services companies in the United States. And again, we are excited to share with you as these partnership discussions mature and we execute and launch with these partners. So to recap, while our D2C business has always been at the forefront of pushing the envelope of what technology can do in the mortgage industry at its core, we are making great advancements in substantially broadening the use of Tinman, which is like years ahead of the industry through diversification on both the Tinman AI as a platform for other mortgage originators and Tinman AI as a software service to solve the mortgage industry's broken tech stack.
Looking ahead to the second half of 2025 and beyond, the opportunity ahead of us has never been more exciting. We remain focused on enhancing our go-to-market with growth being our North Star, along with continued expense management, channel diversification, all with the goal of getting to profitability on an adjusted EBITDA basis in the next 12 months. While we will continue to invest in building the leading AI platform in the mortgage industry, Tinman, to improve our customer experience and further drive down labor costs and make our platform more efficient and scalable, ultimately, the goal that we have is to now drive the business to profitability, and we hope to achieve that on an adjusted EBITDA basis within the next 12 months. Let me now turn it over to Kevin Ryan, our Chief Financial Officer, who will discuss the quarterly performance and our financial strategy going forward.
Thank you, Vishal. As we've discussed on prior calls, even with the continued challenging market environment and heightened macro volatility weighing on our industry, we continue to make great progress towards our goals of driving increased volume and revenue, balanced with ongoing expense management and improved efficiency. Our goal has been to reach profitability in the medium term. We now have the pathway and visibility to guide to adjusted EBITDA breakeven by Q3 2026, driven by volume growth in our direct-to-consumer and Tinman's platform channels, per loan contribution margin continuing to improve, the continued expansion of higher-margin channels, including Tinman AI platform and Tinman AI software, pricing improvements, and continued corporate and vendor cost reductions. We would like to note that these growth opportunities come with varying levels of expansion and profitability profiles and will change based on the broader macroeconomic trajectory.
As a result, our path to adjusted EBITDA breakeven is unlikely to be linear on a quarterly basis, and we do not anticipate the same level of burn reduction each and every quarter. In the second quarter of 2025, on a year-over-year basis, we grew funded loan volume by 25% to approximately $1.2 billion and revenue by 37% to $44 million, driven by funding more loans to both our D2C and Tinman AI platform channels. We had an adjusted EBITDA loss of approximately $27 million. By channel, second-quarter funded loan volume was 64% generated through direct-to-consumer and 36% generated through Tinman AI platform, along with B2B. By product, funded loan volume was 67% purchase, 20% second lien, and 13% refinance. On a sequential quarter-over-quarter basis, Q2 funded loan volume was approximately up 39% and revenue was up approximately 36%. This revenue growth was driven by increased volume from NEO powered by Better, which is higher gain on sale margins.
Our continued push towards increased pricing and a tailwind from a loan loss reserve release. We expect to continue to drive growth through Tinman AI efficiencies, distribution channel diversification, and optimized marketing while balancing these growth expenses with further corporate and fixed vendor cost reductions. Turning to expenses. During the quarter, total expenses decreased approximately 3% in Q2 compared to Q1. We continue building our Tinman AI platform and Tinman AI software channels, leaning into productivity-driven savings through AI deployment across the mortgage business and driving costs down further in our corporate functions. We are excited about using AI to drive the business towards growth and profitability, similar to the advances we experienced in 2016 to 2021, when we grew originations by over 100 times. Now to touch briefly on our balance sheet and capital positioning.
As we discussed on our last call, we closed a major debt restructuring with our partners at SoftBank in April. The accounting entries are a bit complicated and are laid out in the release. Big picture, we increased our GAAP equity by over $210 million, and we meaningfully reduced our corporate debt. We ended the second quarter of 2025 with $241 million of cash, restricted cash, short-term investments, and assets held for sale. In addition, we continue to maintain strong relationships with our financing counterparties with three warehouse facilities for a total capacity of $575 million as of June 30, 2025. We are particularly excited that the Tinman AI platform loan volume is continuing to grow in line with forecast, and we expect over $500 million of AI platform originations in Q3, which is growth of over 16% versus Q2. For the full year of 2025, we expect funded loan volume to increase year-over-year, driven by tailwinds from growth initiatives, including Tinman AI platform, offset by continued macro pressure and the loss of our Ally business, a roughly $1 billion headwind.
In the U.K., we were pleased that Birmingham Bank grew its loan book by 90% in the second quarter sequentially versus the first quarter of 2025. While we continue to undergo efforts to exit our non-core U.K. assets, we expect the divestitures of three smaller non-core U.K. businesses to start being a benefit to our adjusted EBITDA in the second half of 2025 as a result of the dispositions. We expect further improvements to our adjusted EBITDA losses in 2025 compared to 2024 through a combination of AI-driven improvements in conversion rates, efficiency gains, and continued corporate cost reductions. With that, I'll now turn it back to the operator for Q&A.
分析師問答
Your first question comes from the line of Bose George with KBW.
Actually, when you talk about partners trying to enter the space using your technology, can you help characterize who they are? Are they financial companies who haven't figured out a way to offer the product effectively? Or just any color there would be great.
Sure. Within the industry, there are different types of partners entering the market. First, we have next-generation wealth management companies aiming to compete with established firms like UBS and Charles Schwab by becoming comprehensive wealth management platforms. One essential product for any wealth management firm is mortgages, and we've partnered with one of the leading robo-advisers to beta test a mortgage product. We are increasingly receiving inquiries from firms with customer bases ranging from 1 million to 20 million who want to include mortgage or home equity products in their offerings. Next, we have traditional fintech lenders. The major fintech companies that have successfully provided personal installment loans and buy now, pay later solutions, which also have significant customer bases of 1 million to 10 million, are now exploring home equity products. Traditional home equity software solutions have been scarce since the market largely collapsed after the global financial crisis, only recently showing signs of recovery.
There are very few third-party origination platforms that offer a seamless process from application to closing for home equity products, particularly those that allow lenders to hold loans rather than requiring specific delivery to a designated takeout. While a couple of platforms license their software, they impose requirements that may not fit all lenders' needs and often charge high fees. This has led to considerable interest from these fintechs in home equity options. Finally, we see larger fintech companies with customer bases exceeding 20 million. Many missed out on the mortgage boom of 2020 and 2021, but as interest rates decrease and demand increases, they are looking for scalable solutions. Some of these companies have integrated AI in their operations and seek technologies that can seamlessly adjust to changes in the rate environment. We are engaging with these companies as well, currently in beta with some and have established partnerships with others. As these initiatives grow, we look forward to discussing them further as they become significant contributors to our revenue.
And then when you talk about, Kevin, the cost to originate at the industry average and declining, how much volume do you need broadly on the overhead side just to have that benefit really kind of drop to the bottom line more meaningfully?
I believe we are already seeing that. If you examine Page 18 of our presentation regarding unit economics for D2C, it becomes clear that D2C faces a high customer acquisition cost (CAC). This is particularly true for purchase loans, as CAC is recognized immediately, while the loans take six to nine months to materialize. In Q2 2025, the labor cost per fund for D2C is under $2,500, which is quite favorable compared to the industry's labor costs of $6,500 to $7,000 per fund. Additionally, our data costs per fund are decreasing as conversion rates improve. Often, we gather data on consumers and provide preapproval while paying for credit income asset data, but there are times when they do not convert. As we enhance our conversion rate, which has increased by nearly 30% thanks to Betsy this quarter, we see that data costs per fund have dropped from $1,200 to $800. We anticipate that these costs will continue to decrease due to economies of scale and improvements in our data collection methods.
Furthermore, our gain on sale revenue has increased by nearly 10% from one quarter to the next, as we remain responsive to customers and align with their preferred communication methods. This responsiveness has allowed us to offer less of a discount on our products and to improve our average gain on sale. If we were solely a D2C business, we would need to double or triple our business to reach breakeven, given our contribution margin and the rate of monthly burn. However, with our B2B platform and the Tinman AI platform, margins are significantly higher, almost double or triple those of D2C. We simply need to grow those volumes to reach breakeven similar to what we have in D2C, which is why, for the first time in four years, we have clear visibility into achieving breakeven.
Your next question comes from the line of Brendan McCarthy.
Just wanted to start off on the lead-to-lock conversion rate with Betsy. You mentioned that increased meaningfully. I guess what's really driven that? And what are some of the dynamics behind customer interaction to really lead that improvement?
The major development is the enhanced capabilities of Betsy, who previously focused on answering questions and gathering missing information. Betsy can now guide users through the locking process and ask about any necessary documentation. Additionally, she is re-evaluating all loans processed by underwriters, especially those that may have been suspended or denied after locking. This significantly improves the options available to consumers and creates smarter solutions with various investors in our marketplace, resulting in greater efficiency. We are now approving a higher percentage of customers and providing instant solutions. For example, a consumer who doesn't qualify for a mortgage pre-approval due to their debt-to-income ratio may have two debts that, if paid off with additional cash, would make them eligible. Traditionally, a human loan officer would find it challenging to calculate these scenarios due to their focus on sales rather than math. However, Betsy can quickly analyze these situations in our process, helping many individuals qualify in ways that other loan officers and platforms cannot. We are also gaining insights from our retail loan officers about what strategies help them succeed and achieve premium pricing, and we are incorporating these insights into both our direct-to-consumer team and Betsy to enhance her capabilities for consumers.
I appreciate your insight, Vishal. Now, regarding the B2B aspect, particularly Tinman AI as a software, could you provide an update on that pipeline? I recall you mentioned there was a small to medium-sized bank in the pipeline that might be considered for a near-term partnership. Any updates on that?
Yes. We're facilitating loans through that process, which is quite impressive. They're a bank of approximately $1 billion in the mortgage sector and have fully integrated our system. We managed to implement our new mortgage software platform within 90 days, a task that usually spans nine months and incurs substantial consulting and integration expenses. As their transaction volume increases, we anticipate it will become a significant revenue source for us. We also have several other major partners in development, including one of the top ten mortgage companies in the United States, along with numerous large fintech companies I've previously mentioned. We're very excited about the potential of this software business and its ability to help mortgage companies and fintechs originate loans more efficiently than the conventional multiple-system approach that many currently use.
Your next question comes from the line of Rayna Kumar with Oppenheimer.
This is Abigail Rutter on for Rayna. It sounds like the testing you guys mentioned is working pretty well. Could you guys provide any additional color here to expand on this opportunity? And then how do you expect it to work through the cycle?
We anticipate that our software business will have significantly higher margins than our platform business, which currently stands at a 40% margin. We have invested $1 billion in software development. The leading players in the market hold considerable market share and essentially dominate the competition, often starting as fee-based disk software. The top competitor in the market still only allows one user to access the loan file at any given time. It's clear that AI agents will prompt loan processors to "check out" of the file, similar to returning a library book, to allow for entries or changes. This presents a major opportunity for us. We have recently made strategic hires to expand this business, and we will soon announce some significant contracts that are being signed and implemented. Once these contracts are operational, we will share more details. Our model, which offers a complete solution, should encourage widespread adoption, ultimately leading users to transition entirely from their current systems. Initially, clients will use our solution alongside existing products, starting with home equity lines of credit before gradually shifting more products. We are actively defining the full value chain and its benefits. For updates, you can follow us on social media, particularly on LinkedIn, where we regularly share our software advancements and partnerships.
And then can you just talk a little bit more about the home equity business and that volume? And just any more color on kind of what drove or is driving that growth?
We've experienced significant growth in our home equity volume, increasing from $90 million in Q2 '24 to $240 million in Q2 '25, which represents a 260% increase. We believe we are currently the fastest-growing home equity lender in the United States. Our home equity products share many of the strengths that contributed to our success in refinancing, where we grew from $500 million in 2016 to $58 billion in 2021—a remarkable 100-fold increase in just five years. Our approach is centered on delivering a superior experience to both consumers and our business operations by making the process more cost-effective, efficient, and user-friendly. We achieve this through a marketplace model rather than simply funneling into securitization options, allowing us to offer multiple avenues for investor participation and maximizing approval chances for consumers. This contrasts with more traditional originators who have stricter criteria and are limited by rating agencies, which can hinder their ability to adapt.
As a result, those originators often face lower conversion rates from online leads. On the other hand, because our model facilitates a broader approval process due to our marketplace setup, we can achieve significantly better conversion rates. We are already on a $1 billion run rate for HELOC and home equity originations, having launched this product less than two years ago, and we are rapidly scaling its growth, where last quarter we achieved a 35% increase sequentially. We see a vast market potential in this area, and the product aligns well with our strengths.
Yes. I want to emphasize that consumer acceptance of the product has significantly increased. It did slow down after the crisis, as Vishal previously mentioned. However, mortgage rates remain high and home appreciation since COVID has been substantial. Therefore, the $1 billion can easily grow to $2 billion. We intend to continue promoting this product because we believe it can truly benefit consumers during this part of the cycle.
And also every HELOC, every home equity customer that we're booking today is a refinance tomorrow when rates come down eventually. It's a zero customer acquisition cost refinance. And as you can see, on D2C customer acquisition cost is 45% to 50% of revenue for us. So when we take that out, as rates come back down, each one of these HELOC customers, we're going to be putting them into a cash-out refinance or rate to refinance, and that's going to cost us $0 in customer acquisition cost. So again, just onboarding consumers onto the Better platform, warming them up for the refis that come.
Can you talk about how the software maybe adjusts for or tailors around fluctuations in interest rates? Can the AI get smart about capital markets conditions to get ahead of the market, if you will? Like we all know that mortgage banks can be slow to respond to shifts in interest rates, but does the AI help augment the speed at which consumers are shown a rate based on what's going on in the market?
Yes, it provides instant updates. It is now generating alerts for consumers and allowing us to introduce features that make trading interest rates or mortgages similar to trading equities. It enables better execution and faster repricing of our products. Unlike traditional banks, which may take time to update rates, the AI efficiently evaluates all options across 45 investors with thousands of products to find the best match for consumers based on their needs. This innovation is a significant shift in the market, especially given the constant changes in credit pricing. We anticipate a transformative moment in this sector akin to what happened with Robinhood in equities.
Yes, that's really interesting. A follow-up on the conversion rate. I mean, can you share more specifics on how the Tinman platform helps retain the borrowers' attention and get it from the point of lead generation to the next phase? Like we know the platform is really good at lead gen, but how does it kind of systematically support a that stronger conversion rate, especially with the different entities that are using the platform?
Sure. The Tinman platform has typically excelled in catering to consumers with strong credit and good incomes who are comfortable making significant financial decisions online. While the overall market potential in mortgages is vast, the subset of consumers willing to handle these processes entirely online has often been smaller. Our ongoing goal has been to provide immediate human assistance wherever necessary, but staffing for that support can be very costly. For example, when someone abandons their application mid-process, we have set up triggers to alert us. However, given that all professionals in this industry, such as loan officers and salespeople, require state-specific licenses, it is challenging to maintain a large, ready workforce without incurring high payroll costs. This makes it difficult for our low-cost direct-to-consumer model to effectively connect with consumers at crucial decision-making moments.
Now, we can reach out instantly through our system, and interactions with our virtual assistant, Betsy, have significantly increased over the last few months. The integration between Betsy and Tinman has improved substantially and is becoming more seamless, with Tinman acting as the calculation and matching engine while Betsy provides support, sometimes taking the place of a human when needed. This has been key to our growth; we saw a 35% increase in unit volume in the last quarter, even as we reduced our sales team size.
That concludes today's question-and-answer session. I will now turn the conference back over to Vishal Garg, CEO and Founder of Better Home & Finance Holding Company, for closing remarks.
Thank you all for continuing to support us as we build America's leading AI mortgage platform and in doing so, help consumers get a better mortgage, a better rate, and a better process, leading to both improving their financial stability and living a better life for them and their families. The past five years have been extremely challenging for us given the state of the market, interest rates, but now we're playing offense hard again, aggressively pursuing growth, monetizing the platform that we've built and the AI we've developed on top of it, and doing so independent of the broader economic and mortgage market conditions. We now have clear line of sight in how we win in the current environment with our advances in our Tinman AI platform and our Tinman AI software business and the improvement in unit economics, all together contributing to allowing us to clearly tell you today that we're going to be able to be breakeven as a business on an adjusted EBITDA basis in the next 12 months. I think that's the focus that you all should take away from this call. We now have a path out of a really terrible environment, and we have nearly infinite scale if the interest rate environment changes. Thank you for all your support. Thank you for staying with us, and we look forward to sharing a lot more good news with you over the coming year ahead.
This concludes today's conference. You may disconnect.