The Culture of Failure

Failure has become de rigeur, particularly in software start-ups that initially require little capital and small teams. The idea seems simple enough: you start with an initial venture concept, put together a team, and launch the venture. You develop minimally viable products, keep testing different market and product hypotheses, and pivot based on the market feedback you get. You expect to fail repeatedly and hope to eventually get to a product–market fit.

This approach is successful for some ventures—mostly for software-related companies with modest initial operating expenses. Here are some common examples of this type of venture: one venture creates new icons for your smartphone screen; another is a service that connects to your calendar and makes calls on your behalf; another helps put shopping lists together.

But our book is not for entrepreneurs who want to create ventures of this type of impact or restrict themselves to software products alone. If you want to go after the world-changing opportunities, the required expenses and level of expertise can be high. For these ventures, there aren’t many chances to survive if you get your value proposition wrong. This book is your guide to increasing your odds of success.

Often, the “pivot until you succeed” approach appeals to technologists who are in love with their tech. You get your technology into the market and see who wants it. There’s nothing wrong with being in love with tech; we’ve all been there, because new technology enables innovative ventures. But there is a problem with basing your business solely on your love of a piece of technology (sometimes called “tech push” or “tech transfer”). Technology is almost always best applied as a solution to a defined market problem. It enables the product to function. It is almost never a product in itself. For example, natural language was Siri’s solution to giving users to access web services by voice without clicks. Natural language alone was not a product.

The hit-or-miss approach makes sense only when and if you have struggled with your existing value proposition, and it’s failing. Then, of course, you can and should change course from your original vision. Many great companies have achieved success after reaching a point of failure in their existing approach and then finding a new market to attack. But we agree with Mike Moritz, a leading venture capitalist at Sequoia Partners, who told us, “Pivot means you’ve failed. It’s not that you shouldn’t have to move on sometimes, but it shouldn’t be a strategy.” He went on, “‘Fail fast, fail often’ is marketing rubbish. Nobody likes failure. Constant pivoting is like having a compass without a bearing. You need to know true north.”¹

It’s also worth considering to whom this failure approach appeals—more to the investors than the entrepreneurs. Many venture capital investors like to have a broad portfolio containing lots of small bets, some of which have a greater likelihood of taking off but most of which will fail. This approach allows them to pick and choose the ones that show traction, and to spread their risk over many ventures. It also means they don’t have to invest too heavily in any one venture at the earliest stages. This approach makes perfect sense for them, but not for the entrepreneurs who must bear the cost—financial, personal, and professional—resulting from the culture of failure.

Four Ingredients for Start-Up Success

Our process is fundamentally different by being a constructive approach, rather than a “fail fast, pivot often” approach. It guides entrepreneurs to building a breakthrough venture by combining four ingredients:

• A large market opportunity with potential for rapid growth
• An outstanding team capable of execution
• A differentiated technology or business solution that trumps the competition
• A value proposition and business plan that articulate the company’s value, strategy, and plan and attract the required capital²

Without every one of these elements, the probability of success is nearly zero.

Successful entrepreneurs know this. One serial entrepreneur we know had four successful start-ups to his credit (all of which were acquired by big companies), including one that provided key technology to enable wireless systems. He said, “Let me start with my goal in life—I want to leave a legacy where my endeavors have changed in a big way how people communicate and do business, and in this manner change the world of commerce and human interactions. Sure, I want to continue to be financially successful, but that I could do in other ways. It is the lure of world-changing ventures that keeps me going and motivated to work around the clock.”

He had chosen to sell each of his earlier companies after their products’ market potential had been validated but before they became full-fledged companies positioned for the public markets. Why? “I really never had the kind of investors who were interested in that,” he said. “I started these ventures, recruited excellent people to execute on my vision, but the kind of investors I had chosen were more interested in a short business cycle than the longer term. I never had a road map for going from the early-stage entrepreneurial company to a company that would be a long-term winning player and able to access the capital that would be needed.” Our book gives you this road map.

What do we mean by “world-changing” products or services? We mean those that meet great market needs in a way not done before. The enterprises that offer them reach a market of sufficient size to be perceived as major players, either in established markets or in new markets pioneered by the enterprises. Such companies start with a vision that reaches beyond the current market offerings with the aim of building a world-leading venture.

We think of the process of building great businesses as climbing a staircase with frequent exit points. Each exit point represents a milestone where the entrepreneurs and investors can choose to realize the value of the business by a sale. But to continue climbing the staircase to a billion-dollar market value, the business requires new resources, people, talent, courage, and commitment. It also often involves a public offering of stock (initial public offering, or IPO) to access capital. This book discusses those requirements for continued success. If you keep going up the staircase, you can realize more value at the next exit level—but only if you have the resources and talent to continue building the business.

The Role of DARPA

Remarkably, in parallel with our Vanguard vision, DARPA (the Defense Advanced Research Program Agency) had developed a vision to create a program that would lead to breakthroughs in artificial intelligence. DARPA is the primary US agency responsible for funding breakthrough technologies, such as the internet and many other world-leading innovations. Tony Tether, then director of DARPA, and Ron Brachman and Zach Lemnios, director and deputy director respectively of the DARPA Information Processing Technology Office, were seeking to create a program to answer the question of whether artificial intelligent software systems could assist people in their daily lives. The fictional model for the program was the character Radar O’Reilly on the TV series M*A*S*H*. Radar, the ultimate personal assistant, who always knew what his colonel wanted before the colonel knew what the colonel wanted. SRI won a project under the DARPA program and called it CALO (Cognitive Assistant that Learns and Organizes).⁴ CALO, years later, helped inspire the creation of Siri.

CALO developed into a massive program under the leadership of Bill Mark; Ray Perrault, director of the artificial intelligence center; Adam Cheyer, David Israel, Karen Myers, and Tom Garvey, program directors in the artificial intelligence center; Tom Dietterich, professor at Oregon State University; and many others. DARPA funded the program from 2003 to 2009, and it included the participation of more than twenty-three universities (including Stanford University, Carnegie Mellon, UC Berkeley, and MIT) and labs from the who’s who of the artificial intelligence world. At more than $180 million, CALO was the largest artificial intelligence program in the history of DARPA. Concepts from the CALO program contributed to the basis of Siri and subsequent ventures.

In 2007 the Vanguard team decided that the market opportunity we envisioned would best be pursued as a new venture rather than as a project and license with an existing company. We had developed a few commercial projects with carriers and handset manufacturers, but the projects were difficult and time consuming, and they executed only a small part of our vision. Why? In almost all our discussions with existing telecom players, they responded with one of the following reasons:

• It wasn’t possible—the technology was twenty years away.
• It was too expensive—we were seeking $5 million to $10 million in development funding plus license.
• It wasn’t part of their business model.
• It wasn’t within twelve months or so of being a product.
• It wasn’t an early source of revenue.
• They were already doing it themselves.

This list illustrates another important point about our approach: if we had started a venture and had been constantly developing products that served the narrow goals and vision of each of these companies, none of the products would have served the broader market we intended to address, and the venture would likely have failed.

To begin the venture, we formed a stakeholder team of SRI business and technology leaders to help develop a value proposition for our venture. A stakeholder team is a team that SRI forms in order to create a venture concept, with the goal of converging on the value proposition. It is composed of internal SRI staff members as well as external members recruited to join the venture. You can think of the stakeholder team at SRI as the founding team, with the external members and some SRI members joining the venture when it is formed (usually, SRI members prefer to stay at SRI and continue to work on their research).

The Siri Team

Once we had reached a stage where we believed the value proposition for the venture was feasible, we needed leadership with creativity, passion, and domain experience. We were fortunate to recruit Dag Kittlaus as entrepreneur-in-residence (EIR) at SRI, a position that indicates a member of the stakeholder team who has come from outside SRI and who will likely join the company in one of the executive roles. At SRI we often recruit EIRs who can be a CEO or CTO or VP of Engineering. Our agreement with Kittlaus was that if and when SRI created the venture, he would likely become CEO.⁵

Kittlaus had just left Motorola, where he had demonstrated his entrepreneurial skill by creating a mobile internet portal and launching dozens of innovative mobile applications. He was highly knowledgeable about the mobile market. Tom Gruber joined a few months later. Gruber had been a research scientist at Stanford Knowledge Systems Laboratory and was a leading innovator in intelligent user interfaces. The founding team was now led by Kittlaus as EIR and future CEO; Gruber as EIR and future CTO; and Adam Cheyer as future VP of engineering, together with Bill Mark and Norman (who both planned to remain at SRI). We met almost daily, constantly discussing our concept and product possibilities, all working together in SRI’s venture wing.

Members of the stakeholder team worked together over several months, proposed many possible markets, applications, and value propositions to launch the venture, and finally converged on the market opportunity that addressed the highest need: people wanted to perform all sorts of tasks with their smartphones but were frustrated by the need to repeatedly click the keypad. When they tried to make a restaurant reservation or buy a movie ticket on their phone, for example, they had to keep clicking on those impossibly small keypads to enter the required information on the website. It might not seem like a big deal, but it was. Although smartphones had more computational power than the original PCs, their popular applications were limited to simple functions requiring few steps, such as choosing ringtones and text messaging. In fact, market research found that each time users needed to click on a screen on their smartphone (in 2007 clicking on smartphones was not yet a natural task), 20 percent of them abandoned the application or purchase intent due to the cumbersome steps. Having to click through multiple stages and screens to perform and execute tasks was too annoying for most people. As a result, businesses were losing revenue and opportunities. This was the market problem we were seeking to solve with a revolutionary new product.

The breakthrough idea behind Siri was simple and powerful: in contrast with search engines, we decided that Siri would be a “do engine” that would allow people to use their natural spoken language to get answers to their queries rather than by clicking on the smartphone; all the effort to get the answers would be done by Siri. (Only later would we call it a virtual personal assistant.) Siri would allow people to buy tickets, make reservations, get the weather report, and find a movie by speaking into a smartphone. Siri would give them answers, not links.

But how would Siri make money?

We decided that Siri’s revenue model was dependent on CPA, or cost-per-action, basically getting a fee for helping execute a transaction. We could get substantial revenue from the customer leads Siri would generate for hotels, restaurants, and more.

The Siri Value Proposition

We created a broad outline of Siri’s value proposition:

• Target the consumer pain of too many clicks.
• Provide a differentiated and breakthrough technology solution of zero clicks by deploying speech recognition, natural language understanding, and artificial intelligence.
• Give the consumer a “do engine” that provides answers to queries rather than just links.
• Help consumers save time.
• Surprise and delight consumers.
• Create a new market category of virtual personal assistant, avoiding head-on conflict with competitors like Google, which dominates search.
• Have a business model of CPA, charging the web services for leading additional consumers to their services.

This was our eureka moment, when our value proposition for Siri became clear. We could solve a major problem for millions of consumers with a powerful product that could generate billions of dollars in revenue.

There’s an important point to make here: we didn’t start with the cool technology of natural language and artificial intelligence and then create a venture, build a product, and try to find a market and product fit. Instead, prior to forming a venture, we iterated on the venture concept until we arrived at a value proposition that addressed the market pain with a differentiated technology solution.

The Technology Challenge

The technology to solve this market problem was daunting even though it had decades of development. Speech-to-text was the easy part, because SRI had a great deal of experience in that technology. Years before, it had launched the company Nuance, which had tackled speech-to-text, and Nuance was the world leader in that market. The hard part was having the computing capability to analyze the words in the text to understand the intent the user was trying to express in the utterance, and then to reason about and provide an answer to the request. This requires the computer to identify concepts that humans talk about and to associate groups of words with those concepts. This process of analysis, representation, and association constitutes the subfield of artificial intelligence known as natural language understanding. Humans perform these tasks easily, but for computers it has been extraordinarily difficult, and, in general, most people believed it was impossible.

Natural language and reasoning systems had not been in wide use in the commercial world. To implement even the most primitive systems required PhD-level computer scientists specializing in artificial intelligence and natural language. Beyond that, the systems were brittle and difficult to use. Because we were talking about a product that would be available to millions of consumers, it had to be robust.

The broad basis of the technology had been developed under programs with DARPA by the SRI Speech Lab and the SRI Artificial Intelligence Lab, as well as by SRI’s internal investments. Adam Cheyer and Didier Guzzoni executed the specific technological implementation that allowed us to make Siri a product that could be deployed to millions. For almost two decades Cheyer, one of SRI’s most visionary computer scientists, had designed and implemented a vision of delegated computing and “agent-based systems” that enabled humans to interact with networked programs and devices. With Guzzoni, his PhD thesis student, Cheyer developed approaches for natural language understanding and reasoning that simplified the task of understanding and responding to queries. These were the specific approaches that were the basis for Siri’s intelligence.

In the existing state of artificial intelligence and natural language understanding, even with Cheyer’s and Guzzoni’s solution, it was unrealistic to expect the computer to understand everything a user might possibly say. A computer might do reasonably well with taking spoken words and turning them into text, but even so, Siri needed to understand what the user wanted so that it could provide an answer, such as making a reservation or reminding you of an appointment. We decided to restrict Siri to the vertical market domains of travel and entertainment, thereby circumscribing the kinds of general requests it could be expected to understand.

To further simplify the reasoning part of the equation, Siri was designed to handle user utterances that were requests for web-oriented services. Siri would interpret the utterance in the context of one or more web services, input the correct information into the web service, and combine the results into a text or spoken answer to the consumer. For example, if a user asks for “hotels” that are “available tomorrow” “in San Francisco” using terms such as “top-rated” or “good deal,” Siri needs to access and consolidate the results from websites that handle hotel reservations (such as hotels.com) and have extensive written reviews (such as Yelp). As a result, Siri enables a smartphone to act as a (limited) personal assistant, allowing the user to make dinner reservations, check the weather, or find a movie with no clicks at all. Unlocking the promise of smartphones using invisible technology is Siri’s key function.

Seeking Investment

At this point, in 2007, we decided to seek outside investment for our venture. We approached a small number of venture capitalists who were familiar with SRI and who regularly participated in SRI venture reviews. Because Siri’s success depended on creating breakthroughs in the market and in technology, gaining venture funding was not going to be easy. These venture capitalists had seen the cycle of hype versus reality for artificial intelligence and were skeptical that it would work. They worried about every element of the value proposition and business plan, including market, technology, competitors, and more. Some specific issues, among many more, were as follows.

• Would we be able to grow a large consumer base?
• Would the artificial intelligence work?
• Would the processing power of the smartphone be sufficient?
• Would the latency (response delay) of communication and processing be too slow?
• Would the business model of CPA be sufficiently revenue generating?
• Would possible competitors, such as Google or Microsoft, be able to rapidly move with their own products and kill the young venture?

The team then addressed these concerns with an in-depth explanation of both the opportunity and the risks. They created a new venture deck and went back to the original venture capitalists as well as others in Silicon Valley with a compelling value proposition and business plan.

In the end, concerns can only be mitigated and never removed completely. Siri was going to be a bold but risky investment. Clearly it would impact the mobile software industry with its artificial intelligence technology. It would be a breakthrough resulting from a convergence of remarkable worldwide trends—the emergence of smartphones; the advancement of computing, storage, and communications speeds; the growth of web services; and the development of new artificial intelligence approaches. These trends were all converging in 2008. It was the right time.

We raised enough money from venture groups ($8.5 million) to fund the new venture for eighteen months. Our goal with these funds was to build the Siri application, bring it to market, and have at least six more months of funds to allow us to raise our next round of funding. We were also fortunate in that two of Silicon Valley’s top venture capitalists—Gary Morgenthaler of Morgenthaler Ventures, and Shawn Carolan of Menlo Ventures—invested and joined the board, in addition to Dag Kittlaus and Norman. There’s an important point here: outstanding venture capitalists like Morgenthaler and Carolan are both bold and brilliant. They clearly understand the risks, but also the breakthrough opportunities, and seek market and technology breakthroughs such as Siri for their investments. Most venture capitalists will be far less likely to be willing to take the kind of risks listed above. Morgenthaler’s wisdom from decades of experience and Carolan’s deep insights were essential from our first board meeting to our acquisition by Apple, constantly adding essential value to our strategy, plan, execution, and deal terms.

Many issues emerged. Some companies, such as Nuance, were providers as well as competitors, and we had to decide how to manage this. Nuance was one of two vendors providing Siri’s speech-to-text capability but not its core technology, which was natural language understanding and artificial intelligence. Google and others were building solutions that began to closely resemble Siri. Companies were making offers to acquire us, even before Apple made its offer. Deal terms with providers and web services companies were complex. Other opportunities emerged with major wireless carriers, but they would have caused significant distraction from our initial product.

From Idea to Product to Sale of the Company

The Siri team developed the application over the next eighteen months. Our difficulty was that the software originated with researchers at SRI and was never intended to be used for more than a prototype. It needed to be transformed into software that could be used by millions of people, with a compelling user experience, security, robustness, scalability, and all the other elements necessary for a commercial product.

There’s an important lesson here: if we hadn’t sustained a clear focus on our product, our market, and our value proposition, but instead had frequently pivoted, with market testing of different products, we would have burned through our initial investment and never reached our goal. As it was, we barely had enough money to reach this stage. In fact we were almost six months behind our original plan because of issues in making the speech recognition robust as well as issues of the latency time for Siri’s responses to queries.

In 2009 we finally were ready to test Siri in the real world. Between November 2009 and February 2010, we ran the test with a few hundred people. Norman was one of the testers and had an experience that profoundly affected him—because it was an indication of the massive success the company would have.

He was on an airplane and just getting seated. The departure had been delayed, and his neighbor asked him what time they were expected to arrive. Norman reached for his iPhone and asked in natural language, “Siri, what time is United flight 98 expected to arrive?” Siri responded with the adjusted arrival time, and Norman happily provided it to his fellow passenger, adding that he was part of the team that had founded the Siri venture. The man looked stunned. “I have only one question,” he said. “Why are you sitting in coach? You ought to be a billionaire sitting in first class!”

At this stage Siri raised the next round of funding, led by Horizon Ventures, a venture capital firm headed by Solina Chau, who also directs the Li Ka-shing Foundation. Chau has a natural brilliance as an investor, immediately grasping breakthrough ideas and seeing the risks and the opportunities. She understood the power of putting natural language and artificial intelligence into the hands of millions of consumers, and she became a passionate supporter of Siri.

In February 2010, the team launched Siri as a free application in the Apple App Store. The company had prepared for the launch of the Siri application with demonstrations and reviews by top bloggers from sites like Scobleizer and TechCrunch. The demonstrations were a great success, and the press created an avalanche of consumer interest. Robert Scoble was one of the first to break the news about Siri and to clearly understand its promise. His review was glowing and stated that it was “the future of the web.”⁶ Siri was being downloaded free at an astronomical rate. By the first weekend, hundreds of thousands of users had downloaded it. In addition, it was in the top fifty of all Apple apps, and the top lifestyle app.

Two weeks after the launch, Dag Kittlaus recounted to us that he received a phone call: “Hi, this is Steve Jobs.” As Kittlaus told us, he thought it was a joke and hung up. Then the phone rang again: “Really, it’s Steve Jobs.” It was.

The two men talked for a while, with Jobs congratulating Kittlaus on Siri’s capability. He invited Kittlaus, Cheyer, and Gruber to his house. We predicted that he wanted to discuss the purchase of Siri, but we were not looking to sell. We believed that the business value would almost certainly increase as we continued to develop new versions of Siri according to our road map.

Over the next few weeks, Jobs opened discussions with Kittlaus about a purchase price for Siri, with multiple calls per week. Finally, Jobs made an offer that was a sufficient return on investment. We cannot reveal the offer due to contractual obligations, but it was rewarding. The Siri executive team members were also deeply attracted to working with Jobs and Apple.

Would you, dear reader, have sold Siri, or would you have decided to continue to build it as an independent company? As everyone knows, we decided to sell. As a new and highly popular service on Apple phones, Siri helped motivate purchases of millions of iPhones.

Siri continues to be a part of Apple’s IOS mobile devices. Although it was a breakthrough in both market and technology, consumers in the early days learned that the experience of using Siri could occasionally be disappointing. Sometimes it wasn’t able to recognize words or intent. In our opinion (after Apple bought Siri, we had no further knowledge of Apple’s use of Siri) these experiences may have been due to the fact that Apple had broadened Siri’s original focus from travel and entertainment to any topic anyone in the world wanted to ask about. This made Siri a true breakthrough phenomenon with wide appeal, but also made Siri’s word recognition, natural language understanding, and reasoning ability more challenging.

Now Apple and many other innovative companies are racing to develop products that both advance the technology and serve new markets. There is much that can be done. Speech recognition, natural language, and machine learning are only in their infancy. New virtual personal assistants will be even better at word and language recognition. They will retain context, enable conversations (and not only questions and answers), and learn from their users about their personal preferences and needs.

The next generation of virtual personal assistants will also become virtual personal specialists, helping consumers access personal and sensitive information such as health records or bank accounts. SRI recently launched a new venture, Kasisto, which provides this next generation virtual personal specialist to the banking industry. There is no doubt that the future of virtual personal assistants is secure.

The Siri story is only one example of how our process can lead to success and change the world. Keep reading to find more examples from our experience that show how to go about it. Rather than restrict our approach to specific industries, we describe a broad framework that applies to companies targeting services and products of all kinds. We focus on issues ranging from identifying market opportunities that create great ventures, to selecting products for a venture, and choosing the team, all the way through to sustaining innovation within a successful public company.