[TREE – LendingTree] Lead Gen, Marketplaces, and Scale Economies
Less than four years after the its founding and two years after the nationwide launch of its lending exchange, LendingTree raised its first round of capital in the public markets. A month later, in March 2000, the market began its brutal slide, creating two favorable conditions, collapsing advertising rates and declining mortgage rates, that LendingTree, serendipitously armed with cash, exploited to realize positive unit economics for the first time in its history, as the revenue that LendingTree realized from pairing home buyers with loan officers doubled even as the company’s marketing costs declined. With cash flowing and profits swelling, the company was bought by IAC in May 2003. At some point during IAC’s ownership, LendingTree thought it wise to originate mortgage loans (as “LendingTree Loans” or LTL)1, including subprime and Alt-A loans and loans in bubbly property markets like California and Florida.
At the 2006 peak, income from mortgage operations comprised nearly half its total revenue. Sixty percent of this business went away over the next 2 years, leaving behind a trail of restructuring charges, asset impairments, and class action lawsuits. Moreover, the in-house mortgage lending operation was claiming for itself leads that might have otherwise gone to LendingTree’s marketplace. So in May 2011, nearly 3 years after spinning off of IAC2 and just 3 months after declaring LTL’s intent “to be big”, LendingTree announced it would be selling its lending operations to Discover Financial Services. At around the same time, the company shut down its perennially money-losing proprietary real estate brokerage business, RealEstate.com, clearing the final debris off the online loan marketplace that it started with.
But is LendingTree a marketplace, really? Management refers to itself as such in its filings and makes comparisons to Expedia and Booking:
“So think of LendingTree as a marketplace or an exchange. Put us in the camp of other marketplace businesses, whether it’s an Expedia or Priceline in travel or a Google in generalized search or an Uber or an eBay, et cetera”.
Douglas Lebda, Founder and CEO (Goldman Sachs Financial Technology Conference, 9/7/2017)
With nearly the same conviction as Live Nation’s Mike Rapino, LendingTree describes its business model as a “flywheel”. The mutually reinforcing supply/demand pull that characterizes a classic marketplace seems at least halfway apparent in LendingTree’s model – consumers visit lendingtree.com to comparison shop for loans, which in turn draws lenders looking to acquire borrowers – but unlike the thick marketplaces of Booking or eBay or Uber, LendingTree does not benefit from an explicit and direct two-way feedback loop between consumers and suppliers that push it towards a winner-take-most outcome. This assertion is clarified through several key differences between LendingTree and Booking.
First, LendingTree’s supply base of 450 lenders, spread out among many different loan categories and particularly concentrated within credit cards, is not nearly as fragmented as Booking’s base of 1.9mn properties, constraining the scope of the coordination problem whose resolution may serve as a deep moat, as the marketplace plays host to many engaged but increasingly commodified constituents. Second, the way in which Booking and LendingTree match supply and demand differ. A hotel operator isn’t deciding which customers get to stay in which rooms; all room inventory is made available to all travelers. A lender, on the other hand, is seeking borrowers who match specific underwriting constraints and LendingTree must at least superficially screen a borrower before sending the lead to a lender. You obviously can’t present the same lending terms to different borrowers the same way you can show the same hotel room to different travelers. So when you visit lendingtree.com to shop for a mortgage, before seeing any offers at all, you’re required to answer a series of questions: what type of property you are purchasing, how the property will be used, where it is located, how must it costs, how much you are putting down, your estimated credit score, etc. That information is then passed on to lenders looking to fund someone with those specs and who, for the next several days, spam you with emails and phone calls. [The company is in the process of changing this process so that instead of being hounded with calls, you are pre-qualified by LendingTree before being handed off to a single lender. It’s an important move that I’ll talk more about later].
Per LendingTree’s motto, “when banks compete, you win!”, but you also lose your sanity and your temper in the process. Pungent complaints regarding unsolicited hard credit checks and incessant outreach, even after the borrower has chosen a lender, pollute the web (I used LendingTree years ago and can personally attest to the deluge of unwanted inbounds following a loan request). The bombardment of serial offers confounds the relationship between the number of offers provided and the utility received. Does the prospective borrower really feel better informed or less stressed out after fielding offers, one at a time, from 3 mortgage lenders vs. 4 or 5? Does she even want to entertain the second sales pitch after spending 20 minutes on the phone with the first lender?
I’m not saying that LendingTree doesn’t benefit from supply/demand aggregation. Of course it does. The more consumers that LendingTree attracts, the better the odds that a lender’s underwriting criteria will be met; the more lenders that advertise with LendingTree, the greater the chances a consumer, with his idiosyncratic credit profile, will be paired with a loan. Rather, I’m saying that the motion by which supply and demand is synchronized differs from a classic marketplace and that it is more appropriate to think of LendingTree as a sophisticated marketing operation than as a classic marketplace that scales through cross-side network effects. LendingTree’s matchmaking model is linear: lenders ink deals with LendingTree to deliver quality leads; LendingTree fulfills this mandate through precisely targeted doses of direct marketing and brand advertising.
Chicken/egg coordination dilemmas and the challenge of carefully balancing the needs of suppliers and consumers, don’t really apply here. The matching process simply begins with the lender specifying the type of borrowers they want to contact (for instance, a Prime borrower looking for a mortgage in such-and-such Zip Code or a subprime credit card customer seeking a low balance transfer) and what they are willing to pay, and proceeds with LendingTree spending just the right amount in the right channels (display, search) and on the right keywords to maximize the chances of producing that borrower. Basically, the company re-sells online ad inventory to banks, toggling its marketing spend in real-time based on the demand coming from lenders. In that sense, ad spending is a “cost of goods sold”, and it is no surprise that Bankrate3, a close competitor to LendingTree, treats it as such.
Advertising costs account for close to 80% of LendingTree’s cash P&L costs and for the most part, the company spends consistently more on advertising as a percent of revenue than other companies with two-sided business models.
Moreover, according to June 2018 data from SimilarWeb, among these two-sided businesses, LendingTree derives a disproportionate amount of traffic from paid channels.