Case study — Technology · 12 min read · 5 questions
Why Opendoor and Zillow failed at flipping houses
Published · Updated
The thesis
In 2014 Opendoor was the answer to every complaint about Silicon Valley. Here was a startup taking on an industry as old as the Middle Ages — buy or sell a home in a few clicks — using Stanford PhDs and proprietary models to price any house more precisely than an agent, then buy, repair and resell within three to four months. Not flipping houses: a once-in-a-lifetime market maker. The numbers were shown only under NDA, and the hype pulled Zillow and Redfin in behind it.
The business it built is narrower than the story. Opendoor buys below market, takes a 5% service fee, then adds a non-negotiable repair fee it is not obliged to spend — in one case deducting $20,000 for work it never did before relisting ten days later. Its own filings say margins come from the service charge to sellers, not the spread between purchase and resale. It is a fee business dressed as a market maker, appealing to a small group: Americans own a home for eight years and most sellers want the maximum price.
What makes it a house of cards is the leverage. Debt went from $271 million to $6.1 billion in two years, and a single point of interest costs $40 million more. Against $1.3 billion of operating expenses at roughly $22,000 of profit per home, Opendoor needs about 60,000 sales a year to break even. It sold 21,000. Zillow ran the same play and lost $400 million in a quarter before firing 6,000 people — and its CEO gave the category its honest description: a leveraged housing trader, not a market maker.
How do you think about this? 5 strategy questions this case raises and does not answer.
Read the comments, or add yours
The statistics
By the numbers — swipe or use arrows
Revenue, homes sold, market counts, inventory value and average home value, gross margin, net losses, operating expenses, contribution profit after interest, and senior and mezzanine debt facility terms and balances from Opendoor Technologies annual reports and 10-K filings, 2018 through 2021, with Offerpad Solutions figures from its own filings; Zillow Offers losses, wind-down and layoffs as announced by Zillow Group; funding rounds and valuations as reported at the time; seller experience statistics from industry survey data cited in the episode
Key takeaways
The pitch was everything Silicon Valley wanted to believe about itself. In an era criticized for Instagram clones and petty mobile apps, Opendoor took on an industry as old as the Middle Ages — and combined every hot trend of the 2010s: online-to-offline platform, data science, machine learning, algorithms, AI and automation.
The technical claim was precision at scale. Where an agent weighs a few comparables — bedrooms, bathrooms, square footage, lot size — Opendoor claimed to quantify hundreds of variables: floor types, stainless appliances, pool sizes, granite countertops, proximity to transit and schools, even road noise — continuously repriced as local sales occur, rather than a one-time fixed calculation that goes stale.
But the offer is deliberately below market, because it has to be. Opendoor's all-cash offers come in 0.2 to 2% under fair market value, and with the flat 5% service fee the seller is leaving at least 5-7% on the table. It must buy below market to resell at a profit in three to four months.
The repair fee is where the real margin is, and it is non-negotiable. After the in-person assessment the offer drops, with a repair charge from half a percent to 5% or higher — and sellers cannot dispute whether the repairs are needed or the percentage fair. Net proceeds end up 5-10% below fair market value.
And it need not do the repairs at all. Opendoor has no obligation to perform them or show proof. In one case it deducted $20,000 for HVAC, flooring and foundation work, then relisted the home ten days later with fresh paint and none of it done. That $20,000 was pure profit. This is pricing power exercised on people with no alternative.
The company admits where the money comes from. In its own words: "our business model is designed to generate margins from our service charge to sellers and not from the spread between acquisition price and resale price." It is a fee business, not a market maker.
The appeal is far more niche than the company admits. The average length of home ownership is eight years, and most sellers want maximum price and treat the traditional hassle as a small cost of appreciation. Real seller conversion has sat at 35% for three years — roughly 4 in 10 engaged sellers — across only 44 markets.
The funding was a closed loop of reputation. Opendoor was hatched by Keith Rabois — PayPal Mafia, early LinkedIn, COO of Square — who assembled angel investors from his own circle: the CEO of Y Combinator and founders of PayPal, Affirm, Yelp, YouTube, Reddit, Box and Lyft. Their starpower supplied credibility by association before any results existed.
The debt structure is the actual business. Senior facilities lend 80-90% of a home's cost basis at variable rates tied to LIBOR, with $7.8 billion of capacity, repayable immediately on sale — and if a home sells at a loss, Opendoor must find the cash elsewhere or face foreclosure. Mezzanine facilities add $3 billion funding 100% of cost basis, at 10% interest.
That leverage compounds fast. Outstanding debt went from $271 million in 2019 to $6.1 billion in 2021, and at that balance a single percentage point of interest costs $40 million more a year — about 2,000 additional home sales just to stand still.
Growth was real and did not help. Revenue went from $1.8B in 2018 to over $8B in 2021, markets from 18 to 44, homes flipped from 7,000 to 21,000. Inventory tripled from 5,500 homes worth $1.3B at an average $233,000 to 17,000 homes worth $6B at an average $358,000 — moving deliberately into higher-value properties.
The losses grew with it. Even in the record 2021 rebound, Opendoor posted a record loss of over $600 million — and while revenue grew 3.5x from 2020 to 2021, operating expenses grew 3.2x. That is a linear business, which is precisely what a technology company is not supposed to be.
The metric leadership champions is Contribution Profit After Interest, which grew from $27M in 2019 to $482M in 2021 — about $5,800 per home in 2019 rising to over $22,000 in 2021. Impressive, and beside the point.
Because against $1.3 billion of 2021 operating expenses, $22,000 a home means Opendoor must sell roughly 60,000 homes a year just to break even. It sold 21,000. Nobody ever doubted there was money in flipping houses — people have done it by hand for decades. The question is whether there is enough to justify doing it with algorithms at scale.
Zillow ran the same play and it ended in public. It launched Zillow Offers in 2018 predicting a $20 billion run rate by 2024, believing its 15 years of Zestimate work made it the real expert. Its algorithms kept assuming a hot market, it overpaid across a sprawling inventory, lost more than $400 million in a single quarter, killed the division, apologized, and laid off 25% of the company — over 6,000 people into a looming recession.
And the copycat's numbers close the argument. Offerpad runs the same model slowly and selectively, buying only homes it is confident it can flip, and lands on an identical ~$22,000 contribution profit per home. By tightly controlling costs it achieved a net profit of $6 million — a 0.31% net margin. One profitable iBuyer, at a third of a percent, still does not justify the machinery.
Common questions
Why did Opendoor fail?
Because the unit economics never reached the scale the cost structure demanded. Opendoor makes about $22,000 of contribution profit per home against roughly $1.3 billion in annual operating expenses — meaning it needs to sell around 60,000 homes a year to break even, and it sold 21,000 in 2021 while posting a record loss over $600 million. It also carries enormous leverage: debt went from $271 million to $6.1 billion in two years, where a single point of interest costs $40 million more annually. It went public via SPAC at a $6 billion valuation and trades below its 2019 private valuation.
How does Opendoor make money?
Not from flipping, by its own admission — the filings state margins come from the service charge to sellers rather than the spread between purchase and resale price. Opendoor buys 0.2 to 2% below fair market value, charges a flat 5% service fee, then adds a non-negotiable repair fee of anywhere from half a percent to over 5% of home value. The seller ends up with 5 to 10% less than their home was worth, and Opendoor holds inventory acquired below market, which improves the odds on the eventual resale.
Is selling to Opendoor a good deal?
Only if speed and certainty are worth more to you than 5 to 10% of your home's value. You avoid the three-month process, the $6,200 average preparation spend, the 35-day close and the roughly 20% chance an offer falls through — and you choose your own closing date. But you are giving up meaningful money, and the repair fee is not negotiable and not necessarily spent on repairs. The people it genuinely suits are those in a hurry or unusually stress-averse, which is why real seller conversion has stayed at about 35%.
Does Opendoor actually do the repairs it charges for?
It is not obligated to, and does not have to provide proof. In one documented case Opendoor deducted $20,000 from a seller for replacing HVAC, flooring and repairing the foundation, then relisted the home on the open market ten days later with fresh paint and none of the major work done. That $20,000 became additional profit. Because the fee is charged as a percentage of home value rather than as a credit for actual repair cost, the gap between what is charged and what is spent is the company's to keep.
What happened to Zillow Offers?
It collapsed publicly in 2021. Zillow entered iBuying in 2018 predicting a $20 billion run rate by 2024, confident that inventing Zestimate 15 years earlier made it the real expert in algorithmic valuation. When the housing market froze, exploded and then cooled, Zillow's models kept assuming conditions were still hot and it overpaid across a sprawling inventory. It lost more than $400 million in a single quarter, shut the division, apologized publicly, and laid off 25% of the company — over 6,000 people.
What is iBuying?
Instant buying: companies use algorithms to make immediate cash offers on homes, buy them directly, make repairs, and resell on the open market within three to four months. Opendoor pioneered it, and the hype it generated during its private years pulled Zillow and Redfin in behind it. Zillow's own CEO later gave the category its most accurate description, saying the business 'looks far more like a leveraged housing trader than a market maker.'
How is Opendoor financed?
Almost entirely on debt, secured against the houses. Senior facilities lend 80 to 90% of a home's cost basis at variable rates tied to global benchmarks, with $7.8 billion of total capacity, and must be repaid the moment a home sells — if it sells at a loss, Opendoor has to find the shortfall elsewhere or the home is foreclosed. Mezzanine facilities add $3 billion funding 100% of a home's cost basis at 10% interest. Equity raises fund operations; debt funds the houses.
Is Offerpad profitable?
Marginally, and the way it got there is revealing. Offerpad runs the same algorithmic model but slowly and selectively, buying only homes it is confident it can flip and optimizing for profitability over growth. Its contribution profit per home is essentially identical to Opendoor's at around $22,000 — the difference is cost discipline, which produced a $6 million net profit at a 0.31% net margin. That one iBuyer can clear a third of a percent by staying small does not establish that the model justifies the technology built for it.
Discussion
Opendoor's filings state that margins come from the service charge to sellers, not the spread between purchase and resale. If that is true, what exactly were the algorithms for — and what would you have built instead?
No answers yet — be the firstThe company can charge a percentage for repairs it is not obligated to perform, to sellers who cannot negotiate. Is that pricing power fairly earned from providing liquidity, or is it extraction from people without alternatives?
No answers yet — be the firstKeith Rabois raised from friends who were PayPal, YouTube, Reddit and Lyft founders, and that credibility unlocked billions in debt. Where does reputational underwriting stop being efficient capital allocation and start being a closed loop?
No answers yet — be the firstRevenue grew 3.5x while operating expenses grew 3.2x. What test would you apply to distinguish a genuine technology business from a linear one wearing the vocabulary?
No answers yet — be the firstOfferpad runs the identical model profitably at a 0.31% net margin by staying small and picky. Does that vindicate iBuying or bury it?
No answers yet — be the first
Related case studies
TechnologyWhy AI is just big data againBig data was sold for a decade and produced nothing. The same players are running the identical playbook with AI.15 min read
TechnologyWhy the scooter startups all went bankruptA Bird scooter grossed $2,000 a year, cost $1,300 to buy and $1,000 to maintain. It lost money on every vehicle, always.15 min read
TechnologyWhy Twitter never made moneyA Facebook user was worth $8 to advertisers. A Twitter user was worth $2.75. Everything else follows from that.12 min read