The startup economy of today is eerily similar to the banking sector of 2007 right before the financial crisis. In testimony to Congress about the causes of that crisis, Ben Bernanke said: “The propensity for excessive risk-taking by … interconnected firms must be greatly reduced.”
We’re seeing that kind of dangerous interconnectedness again — this time in the startup sector. Slack, WeWork, even giants like Amazon are seeking to make more money faster off of other startups that are rich with cash and known for profligate spending. Startups are encouraged to go after other startups as customers, thanks to a cocktail of strategies like Lean and land-and-expand, making for an interconnected ecosystem where much of the risk is hidden.
New companies provide increasingly specialized services to each other, trying to become “monopolies” in a special niche, following the philosophy Peter Thiel laid out in Zero to One. Even more so, investment money that was once fearful of business models built selling to “SMBs” (small and medium sized businesses) — in saner times referred to as “The startup death zone” — now chases those dollars as the land of endlessly promised “growth.”
Like banking in 2007, easy money has bloated the technology sector, leading to riskier investments. Startups provide services to each other, rather than the larger market, as banks once generated massive sums on trading by swapping new financial derivatives back and forth. And the B2B market, once the darling of venture investing, has become an “S2S” model with much more limited possibilities, and perhaps larger ramifications for the economy at large..
The tyranny of small numbers
Part of the growth of the startups-chasing-startups has come from the new strategies of selling software, primarily based on the Lean methodology, which came into vogue in the early 2010s. Lean coined the acronym MVP, “Minimum Viable Product,” based on the idea that a new company should find its first customer as quickly as possible rather than waiting to build the best possible product it can.
The pursuit of a quick sale, however, often results in building for the first customer to arrive. In the enterprise space that typically means newer, smaller companies that can make purchase decisions quickly thanks to the lack of bureaucracy or controls. Selling to a Fortune 500 is a long and painful process, requiring proposals and vendor approvals. To avoid that path, companies now start catering specifically to startups from the beginning. This mentality has transformed the whole B2B model (once seen by investors as safer, and therefore more valuable, than retail) into an S2S model obsessed with Lean, MVP, and customer count. Instead of changing the world, entrepreneurs seek to make a sale to their neighbors at the local WeWork.
Greenhouse, an HR and recruiting SaaS company, is a great example of this MVP route and its challenges. Its functionality is not as deep as older systems like PeopleSoft but has improved over time as it has grown. This “good enough” software is fine for small companies, which have a small budget and want something set up tomorrow — perfect for the Software as a Service (SaaS) delivery model over the Internet. Greenhouse, by cornering that part of the market, now sports a valuation estimated to be $300 million.
The challenge for SaaS companies like Greenhouse is that they monopolize one niche for smaller-sized customers but are never able to capture the big fish they need to actually become profitable. That’s because the biggest companies already have software, provided by bigger, stronger competitors — for Greenhouse, HR players like Workday can give more services (several startups I speak to are in the midst of switching from Greenhouse to Workday). Even Workday competes with “legacy” giants like Oracle that are backed by billions of dollars and armies of salespeople. These little monopolies then often end up stuck with a customer base made of only small to mid-sized companies instead. A fisherman cannot be very profitable when he only catches sardines.
This trend is multiplying. Other startups now build their own products on top of other startups, hoping to catch the same growth wave. One example is TalentWall, software that creates a “map” of applicants in Greenhouse as a visual tool. Companies like TalentWall fill in gaps when software is lacking. But from a business perspective, these startups have even more risk — a third layer of a startup based on a startup selling to startups, creating a deeply interconnected but likely highly risky ecosystem. Depending on startups for customers is tricky — its both less profitable and a big risk as most startups fail — and this multi-level startup creates even more risk.
Examples of this type of business abound. Geekbot is a tool used in Slack, the company messaging tool and another unicorn primarily serving startups. Geekbot creates automated reminders over instant message, which is an interesting feature but not much of a business. Hubspot is wildly popular with small businesses, and its partner list is full of other SaaS companies building on Hubspot to get those startup customers. Like barnacles on a whale, these mini-startups have grown up around the S2S unicorns, hoping to grow with them but also completely dependent on those unicorns to ever have any business.
That risk tends to be invisible during the good times. Companies like Slack and Greenhouse (and their dependents) appear to be growing like gangbusters, as their own customers are rapidly growing headcounts or revenues. This growth technique is an evolution of an old hat in software called land and expand, where an old customer is sold new products. Instead of cross-selling new products to an account, these new companies price their software by the number of users and so make more money as the startups they serve hire at a breakneck pace. Investors see this growth and assume the startup has hit the magic formula to become a unicorn. Moreover, this growth validates VCs’ theories about “niche monopolies” and “pricing power” that justify sky-high valuations.
The blockchain industry, which I currently work in, has no shortage of examples replicating these models. The original blockchain startups wanted to “revolutionize finance,” but finding actual users — even the most popular Ethereum apps have just a couple thousand daily users — or dealing with regulatory hurdles like whether tokens are securities proved too hard. Many startups in the space then pivoted instead to providing services to other blockchain startups — helping “crypto hedge funds” with their taxes, providing trade execution to cryptocurrency “whales,” and other services that Fortune 500s typically buy from investment banks’ prime brokerages or the Big 4 consulting companies.
The second most valuable company in the world itself derives no small part of its supercharged growth from faster growing startups. Amazon Web Services is the biggest provider to startups, helping unicorns scale before they have the time and money to build their own data centers. Infinite scale is one of AWS’ core value propositions, and the appeal of well-funded tech startups’ money is strong enough that Amazon has dedicated an entire section of its website to them, promoting “hot” growth companies like Acorns and Lenda. Snapchat alone has committed to spending $1 billion on cloud services with AWS, even as it struggles to find a viable business model, burning cash at an unsustainable rate, and more worryingly unable to retain the users it currently has. Is Amazon still a trillion dollar company when so many of its unprofitable customers will inevitably run out of VC cash?
The original Lean methodology and land/expand models have become more of a “we’ll fix it later” attitude to both business models and software. Startups build their growth on top of each other, just as banks once traded derivatives back and forth to create illusory profit, and the growth drives a vicious cycle of funding, hiring, and more spend on services from other startups. But there is one key ingredient still needed to create the kind of this kind of explosive growth: easy money.
Making risk invisible
Getting funding is not as hard as it used to be, in part due to there being so much more of it today than there once was. Funds are competing for access, and the old number-cruncher VC is being replaced by story-driven investors who are letting go of the old unwritten rules.
A critical example: Customer bases made up of small companies used to be anathema to VCs who lived and died by the CaC-to-LTV ratio. VCs used (and many still use) CaC-to-LTV to make sure companies don’t waste their funding on endless marketing and promotions to juice their growth metrics and make themselves more attractive for more funding, creating a startup ponzi scheme. Instead, a startup had to justify the money it spent to get a new customer in terms of the return-on-investment (ROI) each customer would bring, and to prove that customer was profitable over the long-term. This means customers that stick around are the most valuable, as they keep paying over the years — unlike most people who sign up for Blue Apron, the meal-kit delivery startup, who cancel their subscription after a few months.
Blue Apron (APRN) is the ultimate example of how being lax on CAC-to-LTV killed investors. There had always been concern about customers unsubscribing (referred to in the business as “churn”), but investors had faith that Blue Apron would apply some marketing techniques to make users more loyal. This inherent flaw did not stop the best VCs from putting money into the business. APRN itself conveniently left out churn statistics in its pre-IPO filing. Sadly, when your investment thesis includes large numbers of humans changing basic behaviors, the risk can be hard to estimate, leaving public investors as the suckers for a stock that is down 90 percent from its peak $2 billion valuation.
Churn is particularly high for businesses with small customers. Recessions wipe out startups in much higher numbers than larger, established businesses for a host of reasons such as less cash, greater credit risk, and non-diversified business models. Investors have historically valued these lower, to compensate for the additional risk. With the powerful concoction of Lean and land and expand strategies described above, overfunded VCs have instead inverted this from a problem into an opportunity, tracking complex metrics like SDRR (subscription dollar retention rate) as key for how well the land and expand model is working.
By looking at metrics like SDRR instead of profitability, investors assume the company will be able to raise prices later (monopolistic “pricing power”) and someday become incredibly profitable. Combining plans to raise prices and find new customers paints a picture of infinite exponential growth, as if the company could raise prices forever while continuing to find more customers — a premise that on its face violates economics’ basic laws of supply and demand.
One poster child for today’s pre-IPO combines the more lax standards for “enterprise” businesses with the startup-chasing of today’s growth market: WeWork. WeWork has been extensively criticized for its business model, which effectively arbitrages long-term leases with short-term rentals. Documents for its $700 million junk bond sale in April showed only 23 percent of its customers are “enterprise” clients. Unlike VCs, however, the public markets tend to use traditional metrics for valuation rather than these newer, edgier substitutes, and investors were unconvinced that WeWork’s numbers made sense, quickly dropping the worth of the bonds to less than face value within days.
The bond markets, it would seem, still have a bad taste in their mouth from the financial crisis over what was called a “liability mismatch,” when banks used short-term loans to buy long-term risky assets (such as mortgage-backed securities). This model worked great until 2007, when the housing markets plunged and and credit dried up, inevitably driving nearly all of Wall Street into bankruptcy. It is not an exaggeration to say WeWorks’ current model is essentially a real estate version of this arbitrage.
Another callback to the crisis is lax credit standards as a business model. One of the Valley’s newest unicorns is Brex, a virtual-reality (VR) startup that pivoted to payments when its founders realized they actually knew nothing about VR. Brex’s model is based on providing credit cards to startups, which likely concentrates their risk and the exposure in a recession. Despite these risks, the startup has charged its way to a billion-dollar valuation in less than two years of operations.
Despite the many similarities to the risks taken by banks during the financial crisis, startups are, for now, a smaller problem: All 260 unicorns combined would add up to less than the market cap of Apple by itself, about $840 billion, while JPMorgan and Bank of America are over $300 billion each.
That $840 billion in value still represents a lot of jobs, though, and a significant portion of revenue for many Fortune 500 companies. That number is also growing every year, from $600 billion in 2016 and only $100 billion in 2013, and could be large enough soon to test the wider economy. The ecosystem has yet to be tested by any major crisis: such as a decacorn collapse, fund implosion, or general recession, and even if the rest of the economy withstood such an event, it likely would have serious ramifications for places like New York or San Francisco, which have seen real estate explode as the tech industry has grown the last few years.
The clearest signal is that these second tier startups, with business models based on other early ventures’ growth, are likely to go down in large numbers in their next economic stress test. Their investors and employees will likely outperform for the near term, but a messy ending is inevitable. And unlike the “too big to fail” banks of 2007, there very likely won’t be any bailout to save them.
Faisal Khan leads strategy and business development for PegaSys, an enterprise blockchain company focused on Ethereum. He previously worked at Strategy& and Manhattan Venture Partners.