top of page

Models, Good and Bad

Among the investors betting early on Wix in 2007, just a year after its founding, was Bessemer Venture Partners. In his deal memo, Adam Fisher (the lead on the deal) concluded with an expected value calculation for the investment:

Source: BVP investment memo on Wix

A few things are remarkable about this memo. First, in 2007 it was brave to invest in Wix, even at a low initial valuation of $10m. All the company had was a demo video and beta site; it wasn’t even clear that it would settle on a website-building tool.

Four years is a remarkably short period of time to look for a return in this situation, but even here, the 1% probability of $200m valuation ended up being conservative: by 2012 the company was valued at over $300m. A year after that, at the IPO, the company was worth over $750m. Seven years later—today—Wix is worth over $16 billion. By comparison, Adobe is worth $235 billion. The market for creators is vast and growing. If Wix continues to execute, the opportunity is enormous.

Wrong Models

Was Adam Fisher’s model wrong? Andy Jassy, CEO of Amazon Web Services, gave an interview recently in which he noted that in the early days, Amazon would use a net present value (NPV) analysis for deciding which internal projects they should invest in. This is similar to the discounted cash flow (DCF) models we use internally at Heller House to evaluate investment opportunities and conceptually similar to the internal rate of return (IRR) calculation Adam Fisher made in his Wix memo: the goal is to figure out what types of returns one can earn from a dollar invested in a given opportunity.

When I’m asked about how we value the companies in which we invest—some of which don’t yet produce accounting profits—my answer is always the same: we use DCFs for everything. I know that my models are wrong because I cannot forecast a company’s revenue growth, profits, and margins ten years out. But it’s useful to have guardrails to Fermi-ize our assumptions: do they make sense? Is the outcome of the exercise reasonable based on what I believe is the market size of this opportunity? Is it within a realm of possible futures?

What Jassy and Bezos realized eventually, however, is that some of the most exciting projects they dreamt up weren’t getting funded. It was hard—if not impossible—to assign an NPV to them (in this regard, Adam Fisher did a very good job: it was hard to see the future for Wix, but he took a very good stab at it!).

Amazon ditched the NPV approach and moved to a decision-making process involving five questions:

1. If we build it and it’s successful, can it be really big and move the needle?

2. Is it being well-served today?

3. Do we have some kind of differentiated approach to it?

4. Do we have some competence in the area, and if not, can we acquire it quickly?

5. If we like the answers to the four questions above, can we put a group of single-threaded, focused people on this initiative?

If these questions sound like the team at amazon read Clay Christensen’s Innovator’s Dilemma, it’s because they did. From The Everything Store by Brad Stone:

“One day in 2004, Bezos called Kessel into his office and abruptly took away his impressive job, with all of its responsibilities and subordinates. He said he wanted Kessel to take over Amazon’s fledgling digital efforts. Kessel was skeptical. ‘My first reaction was that I already had the best job in the world,’ he says. ‘Ultimately Jeff talked about building brand-new things, and I got excited by the challenge.’ Bezos was adamant that Kessel could not run both the physical and digital-media businesses at the same time. ‘If you are running both businesses you will never go after the digital opportunity with tenacity,’ he said.”

This is what Jassy meant by “single-threaded”: in computer speak, a thread is a computer command; a single-threaded program executes one command at a time. A multithreaded program executes—you guessed it—many commands in parallel. A team cannot pursue a new business with the tenacity required for real creation if it’s preoccupied with an existing business. Stone continues:

“By that time, Bezos and his executives had devoured and raptly discussed another book that would significantly affect the company’s strategy: The Innovator’s Dilemma, by Harvard professor Clayton Christensen. Christensen wrote that great companies fail not because they want to avoid disruptive change but because they are reluctant to embrace promising new markets that might undermine their traditional businesses and that do not appear to satisfy their short-term growth requirements.”

This is what Jassy was referring to: projects where the NPV is clear and calculable are the obvious, short-term ones. By definition, new markets are undefined, and that’s where NPV fails:

“Sears, for example, failed to move from department stores to discount retailing; IBM couldn’t shift from mainframe to minicomputers. The companies that solved the innovator’s dilemma, Christensen wrote, succeeded when they ‘set up autonomous organizations charged with building new and independent businesses around the disruptive technology.’”

This is why Amazon has, famously, two-pizza teams (a small team that can be fed with two pizzas) for new initiatives. A new, undefined market is meaningless to Amazon the behemoth, but it’s huge for a single-threaded, two-pizza team.

“Drawing lessons directly from the book, Bezos unshackled Kessel from Amazon’s traditional media organization. ‘Your job is to kill your own business,’ he told him. ‘I want you to proceed as if your goal is to put everyone selling physical books out of a job.’ Bezos underscored the urgency of the effort. He believed that if Amazon didn’t lead the world into the age of digital reading, then Apple or Google would. When Kessel asked Bezos what his deadline was on developing the company’s first piece of hardware, an electronic reading device, Bezos told him, ‘You are basically already late.’”

Jassy noted that at Amazon they use the phrase “you can’t fight gravity”: when it’s clear that technology is evolving in a certain direction, you can either embrace the trend and try to shape it, or you can stay in denial and struggle to keep up—or worse, become roadkill.

In the case of books, it was clear that someone would invent a great digital reading experience; it might as well be Amazon.

But here’s how all this ties back to Wix and our investment process: is there a better way of evaluating investment ideas where there is new-market creation? Can we add another tool to our DCF-based toolkit?

New-market creation is happening in all our investments today. What is the market for online presence? (Wix) How about the market for online shops and entrepreneurs accepting payments? (Shopify, Square) And the market for enterprise messaging? (Slack). The list goes on.

The fact that these are poorly defined markets does not make them unattractive—quite the opposite! As we’ve seen with Amazon’s internal process and the teachings of Clay Christensen, it is precisely these undefined markets that make for asymmetric opportunities. We can take Jassy’s five-point list and ask the same questions of our companies:

1. If they build it and it’s successful, can it be really big and move the needle?

2. Is this company producing a product or service where the customer is well-served today?

3. Does the company have some kind of differentiated approach to it?

4. Does the company have competence in the area?

5. Does the company have founders or managers solely focused on this opportunity?

Surely, we should keep our DCF models; after all, our companies are—from the point of view of ideation on a whiteboard—quite mature, having gone public with audited historical financial statements. But the most exciting investments we’ve made are ones in which the opportunity turned out larger than we envisioned initially. The most successful companies use internal research and development, as well as acquisitions, to extend their businesses and deliver growth for many decades.

This is certainly what happened with Amazon. Imagine having a DCF model of Amazon’s retail business in 2005 as one’s sole decision-making tool, unaware of the incredible revolution it would unleash with the launch of AWS in March 2006?

AWS was funded after Amazon ditched NPVs. It was dreamt to solve Amazon’s internal constraints. It was hard for various internal teams to compete for IT resources. Rather, wouldn’t it be better to have hardened interfaces between the organizations (application programming interfaces, or APIs), abstracting those resources into building blocks? Need more storage? Just do it through the API.

As Jassy and his team formulated their idea for AWS, they realized it might be an attractive service for outside developers. After all, Amazon’s struggles managing its IT infrastructure was far from unique.

Larry's Asymmetric Motivation

Three years before AWS launched, Larry Ellison, the driven founder of enterprise software behemoth Oracle, was certain he had the future pretty well locked-up. His goal was to become the world’s most important software company, beating the foes of the day, Microsoft, and SAP. He had already developed versions of the Oracle database and the E-Business Suite that could be delivered over the internet to customers, rather than having customers install the applications locally on their servers.

He also believed that product suites—not point solutions stitched together (“best of breed” software)—would win the day, and with these ingredients, Oracle would turn into the leader of the “utility” model of computing:

“Best of breed products are just transitional technologies; the customer saves a lot of money. We buy the computers, backup their data, run their applications, and upgrade their software much more cheaply than they could ever do it themselves. It is simply a matter of financially applying specialization of labor and economies of scale to the running of computer systems.

“We’re now far enough ahead to win. The internet model of computing has given birth to the utility model of computing and software as a service. We couldn't be better positioned. Our fault-tolerant grids of Linux database servers provide the ideal low-cost, high performance infrastructure for utility computing. Our applications outsourcing is one of the largest examples of software as a service in the industry. I know it won't be easy to catch Microsoft. They are more than three times bigger than we are, but they make money on desktop computing—Windows and Office. They are nowhere in utility computing. SAP is nowhere in utility computing. IBM is advertising utility computing but delivering labor intensive custom systems. So if utility computing is the future, then the future is ours.” [1]

And then, a different future happened. On March 2006, Amazon Web Services launched its first product, S3. This was the outcome:

Source: SEC filings, Heller House analysis

From a standing start, over the last 14 years AWS has grown larger than all of Oracle as measured by revenues. Notice that the chart starts six years ago; Amazon didn’t break out AWS’s numbers until the first quarter of 2015, showing results a year back, so our trailing twelve-month period starts at the end of 2014.

Didn’t Oracle see this coming? They did. But they kept serving their customers’ needs: on-premises databases and suite applications aimed at large enterprise customers. Meanwhile, AWS started at the low-end, offering software delivered as a service over the internet to anyone with a credit card. It was textbook disruption.

While every vendor claims slightly different market share figures, AWS today is the clear leader. The photo above, from Andy Jassy’s 2018 keynote at Re:Invent, shows AWS with nearly 52 percent market share, almost three times the next three Western competitors combined (I ignore Alibaba here since no Chief Information Officer in his or her right mind would put their sensitive corporate data in a Chinese cloud).

See that head poking out of the tiny red sliver in the picture? That’s Andy Jassy poking fun at Larry Ellison, who claims to have a cloud computing business used by exactly no-one. Oracle’s recent acquisition of TikTok won’t change that.

Christensen notes that the process of disruption has a paralyzing effect on industry leaders. Why would they abandon their current, extremely profitable strategy, to attack a low-end market with few revenues aimed at much smaller customers? Rather, they are motivated to lean into their current strategy. “Let’s give our largest, most premium customers more of what they want. AWS is a toy.” This is what Christensen terms “asymmetric motivation” and it’s at the core of the innovator’s dilemma.

Jassy notes, of his competitors: “First, they said ‘no one will use it’, then ‘maybe only startups will use it—but they won’t use it for anything real’. Then it was ‘enterprises will never use it’, then ‘enterprises will never use it for anything mission critical’. Companies and developers voted with their workloads and now [competitors] are in this spot of trying to spin something up and you know, it’s six or seven years late.”

There are a few important lessons we can draw from this. First, even industry insiders like Larry Ellison can’t see the future. Even when he was right—that the utility model of computing would win—he still couldn’t execute on it because he didn’t create the right type of organization to do so.

Ellison also didn’t foresee that the utility model would usher a wave of point solutions that could be easily stitched through APIs. The old model of expensive systems integrators (see his dig at IBM above), where most of the cost of new software was in the labor required to install it, is gone. Most companies today have armies of software developers in-house, building apps on top of their best of breed solutions—Zoom, Slack, Atlassian, Twilio—which themselves can evolve much more rapidly since they’re delivered as a service.

Ellison also wasn’t well-versed in Christensen’s teachings. Today, there is a joke in tech circles that everyone has read Christensen and therefore there is no more disruption. I doubt that’s true. While textbook disruption might be less frequent, large organizations move sluggishly compared to upstarts. There are dozens of companies developing new collaboration tools today, for instance, under the shadows of Microsoft and Google, and many are winning.

The real trouble is for those companies that haven’t internalized Christensen, and that seems to be—still—most of them outside of tech.

In a zero marginal cost internet world, it costs Wix close to zero to serve an incremental customer. Why not have a free tier? This has the added benefit of preventing low-end disruption: nobody can undercut you on price when your price is zero. All our companies share this feature; they’ve learned that lesson. But that also happens to be a feature of the landscape (the economics of an internet world). What we—and our companies—must be vigilant of are new threats, and those will always exist, even if they’re unclear right now. As manager and major investor in our partnership, it’s a job I relish doing every day.

[Note: this post is an excerpt from one of our letters to investors, originally published on October 9, 2020.]

[1] Softwar, by Matthew Symonds


bottom of page