It takes a special combination of ingredients to land in my “flops” file. You have to lose your parent company at least $50 million. You probably started out with big ambitious goals, all-at-once funding, leaders personally committed to a given approach and most of all, a boatload of assumptions relative to facts. It’s a recipe for totally unnecessary disaster.
Big corporate flops have been a fascination of mine for ages. And they are a puzzle – why do otherwise smart, sensible, executives who know their business so often get into big trouble when they tackle a challenge that has even a touch of uncertainty to it?
One big reason is that people in “thoughtland” underestimate how many assumptions they are making, especially if they are used to a long track record of success. Take Quibi, the short-form movie company that promised to fundamentally alter how young people consume content. When it still had the working name of “NewTV” in 2018, even veteran entrepreneur Steve Blank wondered whether the fact that the Katzenberg – Whitman venture had successfully raised over $2 billion meant they didn’t have to test their assumptions as they launched the business. In other words, was “lean startup” or what I would call “discovery driven”thinking not necessary if you had all that cash?
Quibi launched in April of 2020, and three short months later articles were asking, “How did a starry $1.75B Netflix rival crash so fast?.” I would argue that there is a recipe, which, when followed, nearly always will land you in the flops file. Let’s explore that with the recent flameout of Zillow’s iBuyer program, Offers, as our case study.
The Zillow Offers Concept
The original idea for Zillow in 2005 was to use the Internet to make the process of buying and selling homes more transparent – similar to what Expedia (which CEO Rich Barton had been part of) did for travel.
At that point, the idea of not just listing homes, but buying and selling them as well, was definitely underway. Spencer Rascoff had stepped down in favor of Barton in 2019. They announced:
“We created Zillow Group in 2005 to make the real estate shopping and purchase process easier,” Barton said in a statement announcing the management shuffle. “Much of our original dream is just now becoming possible. We are at an inflection point in this quest, and the time is right to shuffle leadership seats.” Barton, who also founded Expedia in 1994 and cofounded Glassdoor in 2007, was Zillow’s CEO from 2005 until Rascoff took over in 2010.
As Rascoff also said, “The world is finally ready for the seamless real estate transaction, and no company is better positioned to deliver.” The embodiment of this seamless experience was a thing called Zillow Offers, which was launched in April of 2018. They would buy up homes, make necessary repairs and list the home on the open market with the goal of having a sale within 90 days of acquisition.
The activities that year were not promising – an analyst estimated that each home generated only $1,723 in profit for Zillow after renovation and other costs were accounted for. Nonetheless, the company maintained high hopes for the business. Indeed, it projected transaction volumes of 5,000 homes per month, with revenue potential of $20 billion within five years.
This is where the math becomes head-scratching – even if they succeeded at generating 60,000 sales per year, each transaction would have to generate over $300,000 in revenue to get to a $20 billion figure. According to Zillow’s own figures, the median price of a home in the United States in 2020 is $269,039. Say what???
The context: Beware a boss’s pet project
It is worth considering the context here. The company went public in 2011, with a nice “pop” in its stock price, although, like many Internet companies of the day it was not profitable. It’s performance continued to be uneven. But, before returning to the CEO role, Barton “doubled down” on commitment to Zillow stock, reportedly buying 700,000 shares worth more than $19 million in November of 2018.
In other words, the guy who was about to take over the company had already made a big personal investment in the home-buying strategy and clearly believed in it. Note: when the boss clearly believes in anything, it is really hard for others in the company to challenge those beliefs (see also Amazon Fire phone).
Untested assumptions, taken as facts
So let’s look at some of the key assumptions that underpinned the company’s belief in this idea.
Faith in the power of the algorithm
The first was that an algorithm would be able to identify properties suitable for flipping at scale because it would have inputs from across the nation, giving it an edge. The problem is that unlike, say Internet advertising, home values are influenced by many factors an algorithm can’t identify. Specific location, neighbors, school districts, walkability and the like all influence prices.
This was a knowable issue, people. The pricing algorithm, reflected in Zillow’s popular “Zestimate” estimation of a home’s worth has been shown to often be wildly inaccurate. It doesn’t pick up price spikes in competitive markets, or really understand the subtle factors that influence what buyers consider to be an attractive property. Since this was the basis for Zillow’s pricing decisions, it isn’t surprising that they were on average paying too much for properties.
Faith in the ability to forecast future prices
For a “flipping” model to work, the entity doing the flipping needs to be pretty confident that after they have assumed the risk and spent the capital to buy a home, that it will appreciate rapidly once it’s been fixed up. On the company earnings call discussing the end of the business, Barton ruefully noted that “volatility” meant the business was looking too risky for the capital it would require. Again, this was a knowable fact – one can compare real-world prices with asking and offered prices to see what the differences are.
Ordinary people noticed too. As real estate publication “The Real Deal” observed, “Homebuyers took to Twitter to complain that Zillow’s much-vaunted estimates were wildly out of sync with market prices. Sellers were just as eager to crow about how much the Seattle-based company agreed to pay over asking prices. It was a big problem for a company that had bet its financial future on a computer algorithm to let it quickly buy and sell homes at a profit.”
In Real Life is not the Internet
We’ve seen this movie before: Internet companies making forays into the Real World, assuming it will behave like the controlled digital systems they are used to. The ill-fated startup MoviePass charged so little for unlimited access to movie tickets that it spiraled into bankruptcy. Google’s radio project, Google Audio Ads failed when it was unable to manage ad inventory for local radio stations and couldn’t close the loop on the link between the ad and a buyer action the way it can with Internet advertising.
Trading homes, in real life, is a really complicated and messy business. Homes are emotional. There is a lot of paperwork. There are a dozen characters from appraisers to title companies to lawyers who need to get involved. And the whole thing (as anyone who has done it can attest to) almost always takes longer and is more involved than one thinks at the outset.
And of course, although no one could have predicted this, the pandemic meant that suddenly millions of people who might have eased into becoming homeowners decided to do it all at once and right away, creating market pressures that no one could have anticipated.
Few opportunities for low-commitment testing
Before making an irreversible and significant commitment to a particular business approach, I’m a huge fan of doing as much low-commitment, de-risking testing as possible. As Michael Schrage says, a cheap experiment beats good ideas every time.
For instance, Zillow could have run a test in one or two zip codes that would tell it how close its algorithms came to actual values (in fact, they already had this information before they launched the iBuyer venture, but apparently managed to ignore the inconvenient fact that Zestimates are not the same as appraisals).
They could have partnered with an experienced in-real-life home-flipper to again test their assumptions about pricing and timing. Let’s try it with 10 or 15 homes before we leap in to buy hundreds (which they did in the first year of the program).
Given all the data Zillow has, they could easily have simulated how successful the business would be, contrasting what their algorithm-based decisions would have been with what happened in real life. Of course, they would have had to be willing to heed the results.
Leaders personally associated with the approach taken
It seems pretty clear, given Barton’s significant investment in the strategy, his personal association with it as a new CEO and the funds invested in it right from the beginning, that he wasn’t about to take an experimental approach to this thing. It was going to be “go big or go home.”
And the rosy story told by Barton lasted almost until … well, it couldn’t last. As housingwire.com reports, the announcement that the company was shuttering its i-Buying business and laying off about 25% of its workforce was a “Mind-boggling” about-face, given that up until that point Barton’s projections were extremely positive. When the moment came, it came, as I often talk about, gradually and then suddenly.
Respecting uncertainty, planning differently
High uncertainty situations with lots of assumptions lurking about are perfect for discovery driven planning.
1. Define success before you start. You want to earn revenue of $20 billion a year flipping houses? Sure. Let’s just see how realistic that is.
2. Market and competitive benchmarking. According to Statista, there were 6.5 million homes sold in the United States in 2020. So, let’s do the math. For Zillow to generate $20 billion in revenue, they would have to be able to clear $3,076.92 on every home sold if they sold every single building in the country that year. Say they’re able to get 50% share of the home buying/selling market, then they’d need to clear over $6,000 on each purchase. This is not rocket science, this is simple arithmetic. You can see without going to all the trouble of launching the business that it can’t work.
3. Operations specifications. Like so many businesses, this is where companies dreaming in Thoughtland fail to get real. Aside from price, what’s it really doing to take to increase the value of a home you purchased enough to move the needle – like, how many carpenters, electricians, gardeners, window guys, electricians….you name it. And who is going to manage all those people?
4. Document assumptions. This looks to me a lot like assumptions that completely took over from fact-based reasoning. We wanted it to be true, so in our minds it is!
5. No checkpoint planning. The heart and soul of a discovery driven plan is saying, “let’s not do this all at once, let’s plan to the next checkpoint and see what we’ve learned.” Clearly, not the approach taken here – it was much more “damn the torpedoes, full speed ahead!”
Resources for Learning
Do you want to get smarter about this? I’m delighted to say that the beta version of my learning module on the basics of Discovery Driven Planning is now available on-line!
You can sign up to order it here: https://learninghub.valize.com/offers/RaWGopHz/checkout
There are also all kinds of free resources on my web site (ritamcgrath.com) and our sister site, Valize.com. Articles like this one are in the “Thought Sparks” tab.
Lessons learned – again – managing innovation is not an undisciplined grab bag where you get to do whatever you want. It’s a disciplined process which can be de-risked and managed with the appropriate practices.