The Story of Shazam: The Startup Days

The story of Shazam is one of pure innovation, foresight, and friendship. In 1999, Chris Barton dreamed of a seemingly impossible solution to ambient music recognition and created the team?—?including friend Dhiraj Mukherjee, classmate Philip Inghelbrecht, and engineer Avery Wang?—?to make it a reality. Shazam now has over 500 million users worldwide and has extended its service to other areas of business, including television and social networking.

 

“The number one determinant of entrepreneurial success is persistence. If you are not prepared to go to superhuman levels that are beyond rationality to realize your dream, then your chance of finding success is virtually zero.” Chris Barton

Newnham: Let’s start at the beginning. Can you tell me about your background and how you first got excited by technology?
Barton: My first interest in technology started at around eleven years old when my mother brought home a Sinclair computer, which was one of the very first personal computers. I taught myself how to write simple software programs on it and was absolutely fascinated by what computers could do. My parents were also big influences on me. My mother had been a Professor of Computer Science, and my father had been a Professor of Nuclear Physics. My father had always told me that his dreams as a child were to invent things, and I also loved the idea of inventing and creating.

Later, when I was a college student at UC Berkeley, I took lots of computer science classes and came close to pursuing a double major in the subject. However, I also loved business, and so I decided that one day I would pursue a career that would marry the two together. I actually graduated with a major in Economics and minor in Business, with a lot of coursework completed in computer science, and I knew that my career would one day have something to do with businesses that could be made possible by the new computer technologies.

Inghelbrecht: My first experience with technology was my 14th birthday. I got a top-line PC: 512kb RAM, boasting two 5? floppy drives (the version with a 10MB hard-drive was a little too expensive). It ran DOS, a program called MS Gemstar, which I think later became Windows, and I took an interest in writing small programs using BASIC. One of those programs allowed my dad to manage invoices?—?easy data entry, automated printing, and storage/search. My dad never took to it, probably because it took him ten times longer than pen and paper. I also learned to write in Pascal but sadly never ventured much beyond that.

I was fascinated by investment banking, and the first few years in my career were spent on the trading desks, mostly dealing in fixed-income derivative instruments. My decision later to attend Berkeley’s MBA program in 1998 was really driven by San Francisco’s wonderful location and access to nature. Little did I know that I had landed in Silicon Valley, smack in the middle of the dot-com boom (and after that, bust). I fell back in love with technology, and, in as little as a year, pretty much reinvented myself to that of a Silicon Valley junkie.

Mukherjee: I am from India and was born in Delhi, but I grew up all over the world. My father worked for Air India, and in his job as an airline manager, we moved every three years or so. I lived in Athens, Paris, Calcutta, Geneva, and Bombay growing up, with an unusual cocktail of languages and education systems. At heart, I am Indian; culturally, I am European; and professionally, I am largely American, since I lived and worked there for eleven years.

I remember my dad bought me a ZX Spectrum when it first came out in 1982 or 1983; I was 13 or 14 years old at the time. We were not wealthy, so it must have been a big expense, but it shaped the course of my life. I learned to program in BASIC and wrote little video games in my bedroom after school. Then, when I was 18, I went to university at Dartmouth College in the US, in beautiful, rural New Hampshire. Dartmouth was a pioneer in computer science, and the BASIC language was invented there. It was the first university in the United States to require every student to own a computer. I enrolled in 1987, and my first computer was a sleek Macintosh Plus.

At university, I took a couple of computer science classes but fairly quickly realized that I did not have the talent to be a software engineer. But I did enjoy fiddling around with computers and software programs so I got a job working at the University Computer Center as a consultant/ advisor to other students, which helped me learn to use applications like Microsoft Excel. I graduated with a degree in Mathematical Economics and a taste for technology.

My first job after university was as a management consultant with Bain and Company in San Francisco. I spent three years there and a gap year backpacking around the world. In September, 1995, I turned up at Stanford Business School, in the heart of Silicon Valley. Netscape had launched its web browser a year earlier in 1994 and had a spectacular IPO a few weeks before I arrived at Stanford. The internet revolution had started with a bang, and I just happened to be in the right place at the right time to be swept into the business of technology. After the first year of my MBA, I worked at Oracle, and when I graduated, I joined a startup called Viant in San Francisco, advising big companies on building digital businesses. My dalliance with technology had turned into a career.

“By my junior year, I was told not to bother coming to class and to just show up for the tests.” Avery Wang

Wang: To talk about my start in technology, I’ll have to go back to my early love for math and science. When I was in first grade, I saw a Star Trek episode where Captain Kirk was fighting the Lizard Man and figured out how to make gunpowder out of stuff he found lying around. They gave the recipe in the episode, so I had to go try it. Back then, you could get the key ingredient at the pharmacy just by asking for it. These days, you might get arrested. I’ll have to thank my parents for letting me obtain the materials and put together my own pyrotechnics. I was hooked and was always doing science experiments after that.

When I was ten, I was able to take a summer school class on computer programming. I’d then constantly pestered my parents to take me to the Lawrence Hall of Science in Berkeley so I could get some computer time on a 10 character-per-second teletype terminal that printed on a roll of paper. It blew my mind that on the other side of the teletype was a computer. By the time I was twelve, my parents bought me an Apple II computer, one of the earliest off the assembly line. It was a big outlay, since we were far from rich. That was an amazing geek machine. It came with schematics and source code for the firmware. I taught myself assembly language and did some cool things like calculating 16000 digits of pi and building an apparatus for making crank phone calls. That experience prepared me well for doing research in digital signal processing many years later.

Speaking of math, I was not an especially good student until seventh grade. I have never been good at mental arithmetic, so don’t ask me to calculate the tip. However, by high school, I started to notice that the more abstract the topics, the better I did. When I was a freshman, I saw the seniors in Chess Club doing their calculus homework. I was curious, so I bought a book called Quick Calculus and read the whole thing and did all the problems over the Christmas break. I was inspired to do competitions and did very well. By my junior year, I was told not to bother coming to class and to just show up for the tests.

I was never driven by grades but studied the sciences for the love of it, all the way to my PhD and beyond. It was hard to decide what exactly to specialize in. I was a Biology major at first but switched to Math and eventually ended up in the graduate program in Electrical Engineering. I took a three-year detour into neurosciences, including a Fulbright scholarship to Germany. When I got back to Stanford, I ended up at the Center for Computer Research in Music and Acoustics, where I did my dissertation under Professor Julius Smith. After that, my plan was to do a post-doc at UCSF [University of California, San Francisco] doing Auditory Neurophysiology research, but at the last minute, I was recruited by a multimedia chip company to work on digital signal processing algorithms. So, I ended up not being in academia, but I don’t regret it. My passion is to solve hard problems, and there are plenty out in the real world.

Newnham: Chris, can you tell me what moment led to the idea of Shazam?
Barton: After making the decision that I really wanted to start my own business, I spent the summer of 1999?—?in between first and second years?—?brainstorming different business ideas. That summer, I was halfway through my MBA, and I had moved to London for my summer internship at Microsoft’s internet search portal group MSN. I would describe all kinds of crazy ideas that I had to my good friends Perry Blacher and Dhiraj, who were my two best friends in London. In fact, Dhiraj and I would even schedule brainstorming sessions at local cafes on the weekends where we would literally make lists of numerous business ideas and discuss their feasibility.

I wanted to have co-founders for my business, and so I verbally agreed with Dhiraj?—?as well as with Philip Inghelbrecht, who was a Berkeley MBA classmate of mine?—?that we would start a business together once we had a good idea. I then introduced Dhiraj (who lived in London) and Philip (who lived in Berkeley) to each other, and fortunately, they really liked each other! So, now we were three co-founders who shared a passion to start a business; we just needed to pick the idea we were going to go forward with. At the time, in the summer of 1999, it was the height of the dot-com bubble. It seemed like there was so much opportunity for a startup company to thrive.

I think the idea that people wanted to identify the music that they heard in various places was one that many people had thought of. I later learned that there were actually several companies trying to solve this in different ways, but they all shared the same severe limitation that they only monitored the music that radio stations played. This meant their services could not be used in bars, clubs, cafes, retail stores, and so on. In addition, the user experience was poor, since you had to enter which radio station you were listening to.

However, there was a reason that all these startups took that approach; it seemed to be the only technically feasible way of solving this problem. I had also been thinking about a similar concept of enabling people to find out what was playing on the radio. I originally thought that I was going to build software that would help DJs keep records of the songs they played. Then, I would collect this information in real time and thus would be able to provide a great consumer mobile phone service that would tell you what song was playing on the radio. I was particularly excited about the fact that since my software would likely be the only software in the radio station, I would be the only party that would know which songs were playing on the radio. Then, the real idea for Shazam was about to be conceived.

The real A-ha! moment occurred one evening when I relaxed in my London flat in Ladbroke Grove. I had initially moved out to London for six months for my summer internship at Microsoft and now had moved on to continue several more months studying abroad at London Business School on exchange from my Berkeley MBA program. On this October evening, I had come home from a class called Strategic Innovation taught by Costa Markides. One of the principles that Professor Markides taught was the importance of thinking outside the box in order to be able to truly come up with innovations in the world of business.

That evening, I decided to try this type of thinking, so I was racking my brain to try to think outside the box about my radio-monitoring software idea, and I thought to myself, “What might a competitor be able to do to compete with me once I had this software installed in radio stations across the country and once users were delighted by being able to find out which songs they were playing on the radio? What risk was there of someone trumping me or circumventing me?” Suddenly, I thought, “What if someone could identify the song using the actual sound of the music captured over the mobile phone? Then, they would not need to know what the radio station was playing at all because they would actually identify using the sound of the music itself.”

I knew immediately that this was going to be magical! Users would not have to input the radio station they were listening to. They would not be limited to identifying songs only on the radio. They would simply hold their phone up to music and then receive the answer. The reason I had never thought of this approach before (nor had any other person or company to my knowledge) is that the reality was that no such pattern recognition technology existed. Inventing such a technology would be an incredibly challenging task that at the time seemed virtually impossible (and, in fact, many Stanford and MIT PhD’s in audio signal processing would later tell us that it was impossible).

At the time, voice recognition had very poor reliability over mobile phones even though the speaker’s voice was right at the mouthpiece of the phone and the vocabulary was constrained to a limited set of words. As we found when we captured sound samples of music captured via a mobile phone microphone in ambient music environments such as cafes, pubs, and restaurants, the resulting recorded sound contained almost zero audible music; it was lost in a sea of noise from chattering people and overriding blends of both noise and also silence that was caused by noise-reducing technology on mobile phones. These vocoder (voice encoder) algorithms were engineered to emphasize the human voice and block out other sounds coming from a distance. We would thus have a very challenging time deciphering what the music was.

Newnham: What was the original vision for the technology? What were you hoping it could achieve, and how did you think people would use it?
Barton: The original vision of the technology was to enable someone to use their mobile phone to identify the ambient music playing in any location. Since this was 1999, mobile phones were very basic; beyond making phone calls, the most sophisticated things that people did with mobile phones at that time were to install ringtones, send text messages, and, if you were really cutting-edge, possibly subscribe to sports score updates via text message. The coolest phone in 1999 was the Nokia 8210, which was small and had a black and white screen.

Once we invented the algorithm, we realized that our technology was not actually a music recognition technology. It was a “recorded sound identification technology” and in fact would work just as reliably in identifying the recorded sound associated with a television show or advertisement, a movie, a radio show, and so on. For this reason, we knew that one day we could broaden the use cases of Shazam, but we continued to believe that the killer app would be to identify music.

We felt that if we could enable people to identify music, this would become the entry point to a large number of things that people could then do with the music that inspired them. We believed there was a time of inspiration from music that we called “the moment,” when a piece of music moved someone and there was no way to take action at that moment even if you knew what the song was.

Our original PowerPoint presentations to investors painted a picture of a service that would enable people to do things such as buy the music, watch music videos, see lyrics, send music greeting cards to friends, find out which songs their friends were tagging, organize their favorite music into playlists, and so on. Many of these features are available today within the Shazam application but only became feasible after the advent of smartphone applications, which was years later. Shazam was so far ahead of its time in 2000 that it was actually three years before iTunes, seven years before the iPhone, and eight years before the App Store.

Newnham: What made you choose London as your base for Shazam?
Barton: At the time we were starting Shazam, in 2000, the most advanced mobile markets in the world were Europe and Japan. Apple and Google had not yet begun to look at mobile. Nokia was the mega-power in mobile devices. While US mobile operators did not even have interoperable SMS, European mobile operators were more advanced and offered “Premium SMS” as a mobile payment mechanism available to third parties, such as ringtone companies. Premium SMS was not available at US mobile operators. If Shazam had launched in the US, we literally would have had no way to generate revenues.

In Europe, we could charge users using Premium SMS for the music recognition service that we were planning to build thus Europe allowed for a business model. In addition, the UK was the most music-savvy
country in the world as measured by music purchased per capita. Consumers had a passion for music, and thus what more perfect test bed to launch Shazam?

Unlike in the US, this mobile environment in Europe had fostered the development of a mobile startup community with consumer mobile propositions. One big acquisition at the time was IOBox by Terra Lycos.

Newnham: Chris, can you tell me more about the process of finding Avery?
Barton: The beginning of our search for the right technical co-founder was an exercise to understand what kind of skill set this person had to have. We conducted a huge amount of research on the web to learn about the computational research related to the behavior of sound and music. We soon discovered that the most advanced academic research in this area was called audio signal processing, which was a sub-category of digital signal processing or DSP. The folks who were in music-related DSP worked on such things as algorithms to create music synthesizers (e.g. Yamaha), music similarity matching for purposes of music recommendations, and various forms of music recognition that were in a clean rather than noisy environment (e.g. radio monitoring).

During our research, we built a list of names of people who had expertise in this very niche area. We found that a very large portion of these people were studying or had completed PhDs in music-related DSP within the Electrical Engineering Departments of MIT’s Media Lab and Stanford’s Center for Computer Research in Music and Acoustics (CCRMA). We met with numerous people in this area and described our Shazam concept, but everyone said that what we were trying to do was impossible due to the high level of noise around ambient music. There were music recognition technologies that were used commercially at the time, but none would be able to work when the music signal was mixed with lots of overriding noise where the search must take place across a much larger music fingerprint database.

We kept hearing about a key professor in this field who was named Julius Smith and taught at Stanford CCRMA. After a few attempts, we finally were able to get in touch with him and met him in his house in Palo Alto. We hit it off immediately, and he took the optimistic view that it might just be possible to invent a technology that could identify music in such a noisy environment, although he did not know how it could be done. He agreed to join as an advisor to the company.

Soon after Julius joined as an advisor, we brought the list of names we had gathered who specialized in music DSP to Julius and asked him to “rank the five smartest.” He knew virtually everyone, since the DSP community was small, typically attended the same conferences, and referenced each other’s publications. We asked Julius to rank the people not only on who was the smartest, but also factor in whether the candidates had strong theoretical mathematics and statistics skills, since we believed those skills would be important for someone who had true inventor potential. Finally, we also wanted our key candidate to have strong software development skills so we could build a working algorithm.

Julius ranked the top five people, and at the top of that list was Avery Wang, who had completed his PhD under Julius years before. We reached out to Avery and met him for lunch at Café Brioche in Palo Alto. We were very lucky because he just happened to be an entrepreneur who was in limbo and really wanted to build a startup from the ground up. We hit it off, and he jumped on board as the fourth co-founder of Shazam.

Newnham: Avery, what did you think when you were first approached with the idea? At what moment did you think this was something you wanted to work on?
Wang: I remember getting a random email from Chris outlining his idea. He was an MBA student with a crazy idea. I was slightly busy and just ignored him. Being Chris, he didn’t let that discourage him and escalated until he could get a meeting with me. I figured one way to get him off my back was to have lunch with him.

So, I met Chris and Philip for lunch and heard them out. I was delighted at their energy and passion, but beyond that, I was amazed at the huge amount of due diligence research they had done on all aspects of the problem space. They had detailed analyses on state-of-the-art technology, the business landscape, possible competitors, the mobile phone ecology, music labels, business model, operational details, etc. These guys had really done their homework, to put it mildly.

A minor factor standing in their way was that they had no idea of how to actually do the music recognition. They told me that they had done extensive due diligence on who could work on this and asked professors from MIT, Stanford, and UC Berkeley whether 1) they knew how to do it and 2) whom they would recommend. Nobody had the answer to 1), but Chris and Philip followed up with each recommended expert and asked the same questions. They kept doing this until the list stopped growing, and they thought they had a pretty comprehensive list of candidates.

When they described the challenge to me, I also had no clue how to do it or even where to begin. Even though Julius Smith had recommended me, I felt I wasn’t necessarily the right guy, since I had no background in content recognition and search engines. Contrary to many misconceptions floating around out there, this problem had nothing to do with my PhD thesis. I think they then got me together with Julius and had him convince me that it was maybe not impossible.

Their timing was good. I had been working on a two-man boutique audio hardware company that was struggling to get traction. I was floating on savings and income from trading the markets, but I was kind of bored and open to doing something new and interesting. Chris and Philip were persuasive, and I liked their energy, enthusiasm, and ethics, so I said I’d give it a try. A small detail was they didn’t have anything to pay me, but I was used to that, so it wasn’t a problem.

Newnham: How did you begin looking for a possible solution?
Wang: I started out by listing all the challenges that we had to overcome. These included being able to search a massive database of millions of songs; being able to search quickly; having high recognition rate; having a low false positive rate; and being tolerant to noise and distortion, room reverberation, as well as voice compression.

These considerations at least narrowed down the possibilities, though it was down to a nearly empty list. As Chris and Philip already knew, there were no known techniques that could satisfy all these factors. I knew it had to use log-time (or better) fast hashing database search techniques. I knew that techniques in which a sample was compared against every segment of every track in the database could probably be ruled out as being too slow. The technique had to be highly fault-tolerant, able to withstand noise being injected into the signal, as well as having parts of the target signal missing due to obstruction by noise or audio compression. We set up Julius Smith to be an advisor, and I met with him regularly to brainstorm. As the story goes, I was optimistic at first, but after about two months, I still didn’t have a good clue on how to make this work. Around the third and fourth months, I started to get concerned that my co-founders were out pitching to investors, getting office space, and hiring people, all moving on a project plan where my piece needed to be invented by a certain date.

Along with brainstorming with Julius, Philip was tasked with project managing the invention process, and sometimes the three of us would sit down and Philip would take notes and encourage us to think of different approaches. Meanwhile, Chris had moved to London to work with Dhiraj; they had got an office and started building a team.

As I saw the momentum and earnest diligence the guys were putting forth, I realized it wasn’t just a toy problem but that a bunch of real people were depending on it. So, the pressure was on and increasing.

Newnham: Setting up a business with anyone, let alone friends, can be fraught with difficulties (as well as having many benefits). How did you complement each other, and what roles did you each take on?
Barton: As you might imagine, choosing the right co-founders is a critical decision that impacts everything. The co-founders you work with impact everything from your happiness to your ability to fulfill your vision and live by your values and much more. For entrepreneurs who might think intelligently about the skill sets of their potential co-founders, I would emphasize the importance of the fuzzy, emotional characteristics over the easy-to-define ones.

I wanted to build a business that was based on cutting-edge technology and had the aim to revolutionize the way we discover music by using the mobile phone. Yet in retrospect I never once thought I must have co-founders who were experts in cutting-edge technology (with the exception of Avery, of course, who was critical to inventing the algorithm) or who understood the music industry or who had experience in the mobile industry. The fuzzy things were more important.

Philip and Dhiraj came from two different places in my life. Philip was my business school classmate, and Dhiraj was my London social buddy. Both were friends of mine. Both were people whom I had tremendous respect for. They both had passions and the desire to build a startup company from the ground up. They each had different strengths and skill sets that complemented mine. They were high-integrity guys that just felt like people I could start a company with. Starting a company is much more than just a job. It’s a life experience, and so you want to be sure that you do it with people that you want to share the experience with.

Both of them were very smart and could figure out how to solve difficult problems. Both of them could tackle all kinds of things that they had never done before, ranging from writing up patents to crafting critical partnerships. Philip had incredible productivity. He seemed to have the output of ten superhumans. Dhiraj always maintained a zen-like calm and had a very high EQ that was very valuable in challenging times.

Avery was the easy choice once we found him. When Julius Smith told us that this guy named Avery Wang was not only a DSP guy but also a mathematician and statistician at academic levels, I believed he was the one who had what it would take to fundamentally invent new approaches to the world of audio signal processing. It turns out he did.

Newnham: Can you take me back to what the early days in the office were like?
Mukherjee: When Chris and I first started brainstorming business ideas, I was working with Viant in London, where I had a happy and successful career as well as a busy schedule. In the beginning, it was hard to make time on weekends to work on the new business idea. Chris went back to business school at the end of 1999, and he and Philip were able to put more energy towards Shazam, since they had more flexibility as students and were able to point some of their academic classes to startup-related topics.

I had taken quite a few courses in entrepreneurship and venture capital at business school, so I had a pretty good idea of what I was getting into. I had also worked at a startup for three years, had set up an office and grown it from four to forty employees in a year, and enjoyed every minute of it. Still, I remember having a call with Chris where he pushed me hard to resign so that I could focus on Shazam. Luckily, I had saved up some money and practiced living frugally, knowing that I would need to survive for several months with no income the day I became an entrepreneur.

I left Viant in March, 2000, which is exactly when the NASDAQ peaked. I had no idea at that time that the internet bubble would burst over the next few months, making our lives as entrepreneurs harder and harder with each passing week. Back then, it was the beginning of a new adventure; I was ready for it and loved the excitement, the intensity of learning, and the unpredictability. I can remember the sensations from many of those moments in sharp detail even though I normally have a terrible memory!

The first few months were discovering entire new worlds?—?lawyers who helped to register the company, accountants to do the book-keeping, researching how mobile phone microphones worked, investigating interactive voice response (IVR) systems, and worrying about the threat from DAB (Digital Audio Broadcasting) technology.

I remember walking into HSBC’s offices with Chris and setting up our first business banking account. We were helped by a young blonde bank officer who was courteous and friendly but a shade skeptical about our business prospects as we went through the process. She asked us how much capital we planned to put into the company, and Chris and I looked across at each other, shrugged and guesstimated “a million dollars?” Forty-five minutes later, we walked out with our bank account set up with the minimum balance of ￡100 in it, somewhat short of the million-dollar mark.

While we worked on raising our angel funding, we were also talking to prospective partners and vendors. One meeting I remember in particular was with an IBM salesperson who sold IVR systems for tens of thousands of pounds. We made an appointment to meet him, and he arrived impeccably dressed in a pinstripe suit and carrying a briefcase. Our offices at the time were my flat in Bayswater, so we sat around my dining table for an hour and a half, Chris and I dressed in jeans and t-shirts and knowing virtually nothing about IVR technology and sitting on ￡100 in our business bank account. To the IBM person’s credit, he never let his professionalism slip, although I wonder if he thought this was some sort of hoax call or Candid Camera episode. At the time, we were very secretive about the Shazam idea, so we wouldn’t even reveal what we were trying to build!

By far, the most exciting part was meeting with prospective angels and trying to land an investment. Our tactics were unorthodox, but sometimes they worked. We raised some early capital from well-regarded friends and family, which helped to get the ball rolling. However, we really wanted some high-profile angels from the music and mobile industries, so we set about cold-calling, cold-emailing, or otherwise buttonholing them in any way we could.

Eventually, after three months of intensive fund-raising, we closed our angel round in August, 2000. We had done it?—?one million dollars! I still wonder what our bank manager at HSBC thought about that…

Newnham: What was the mobile landscape like when you started the business?
Barton: Back in 2000, the mobile industry was in its first days of the great data revolution. Mobile operators were preparing for a world where voice revenues would become commoditized and data revenues would be the great new growth area. In order to prepare for this, they were making very large investments to acquire 3G licenses and to build out their 3G networks. While they had a vision for rich media experiences on mobile phones, I don’t think it was clear yet that the real solution would be the true smartphone (e.g. iPhone or Android) as we know it today.

In 2000, the great innovations that were around the corner included cameras within the phone, enabling people to send each other photos with MMS messages, using the SMS text message platform to enable one-to-many communications and interaction with information and entertainment services, and enabling a web-surfing like experience using the rudimentary WAP browser. Mobile applications (Java and Symbian) were on the horizon but would ultimately prove to be inadequate for mass application distribution.

When you look at what Shazam does today, most things were not possible and in fact were not even within a few years of becoming available when we started the company: You could not write a single application that could run on a very large number of phones (this was finally changed by iPhone). There was no “App Store” for phones to get your application out to users. Without an application, Shazam could not provide a rich experience that included album cover artwork, YouTube videos, audio samples to listen to, and an ability to navigate through an experience, including your previous tags list. Lastly, digital music downloads did not exist for mobile phones except for ringtones.

Furthermore, the phones and mobile technologies had severe constraints in relation to our vision for Shazam: the only way to record sound and send it to a server was by making a voice phone call. The only way to push information to someone’s mobile phone was by text message. Mobile phone screens at the time consisted of two to three lines of black and white display, making visualisation such as album art a non-starter. It was impossible to play music samples from a server on mobile phones without making a phone call. And even ringtones were limited to “monophonic” and later “polyphonic” sounds that were not true music recordings.

We wanted Shazam to work for most mobile phone users, and we found a way to make that happen. We designed a service where users would simply have to dial a four-digit phone number (which we secured through a partnership with mobile operators) that routed into our IVR system, and then we would use text-messaging to send the information to the user’s phone with the name of the song and artist. In certain cases, you could text back if you wanted to buy the ringtone when it was available for that song. We made money by charging users each time they used Shazam, much like calling directory enquiries.

The problem was that without millions of regular users, our revenue would not be sufficient, and so we remained reliant on venture capital funding for many years. Without the frictionless experience of app stores, it would prove very difficult to attract millions of users.

Newnham: Talking of investment, can you tell me how you chose your investors?
Barton: We pursued our investor financing in two initial stages: (1) a round of $1 million from angel investors so that we could build a demo of our product and figure out the business approach so that we could later raise (2) a round of $7.5 million from venture capital investors so that we could build and launch a full commercial business.

In pursuing angel investors, we decided to aim for people who had built careers in related areas to our business for three reasons: (1) They were more likely to be attracted to our business and want to be part of it; (2) They would be able to provide valuable advice and introductions for us; and (3) They would add more credibility as investors in our business when we approached the venture capital investors. Most of our angel investors were people that we had never known or met, so we often just reached out to them “cold” to get an initial meeting with them.

Eventually, we were able to raise money from an impressive group of people. As we got more of them committed, we were able to make note of this in order to gain credibility with other prospective angel investors. Our angel investors included the former chairmen of EMI and BMG, the inventor of the 56k modem, the founder of Amazon Europe and the former CEO of British Telecom, among others.

About half of these people were found through reaching out broadly in our personal and professional networks (we, of course, created a dazzling PowerPoint presentation that illustrated our plan, team, technology, and progress to date). The other half were people I found by scouring the internet for names associated with big companies and simply emailing them. Some of these were accidental: for example, when mighty Napster was hard-balling us to license the Shazam technology (late 2000), Philip reached out to Brent Townshend for advice. Brent was the sole inventor of the 56k modem and, a bit like David and Goliath, had been able to enforce his patents with the big modem manufacturers. We never got Brent’s advice but did receive a $250,000 cheque in the mail a few days later.

A UK law firm drafted up a bridge loan document that would greatly ease the process of raising angel financing. With this approach, no price negotiation had to take place, and, in fact, we were able to trickle in the angel investments over several months under the same terms. All of the angels could rest assured under the terms that their money would convert to equity at a discount to the rate that the venture capitalists negotiated for valuation. Everyone understood that venture capitalists were the real market makers for startups and thus were in the best position to put the right valuation on the company.

Within weeks of completing the angel round, we began to seek our venture capital round. It was a tremendously difficult time to raise VC money (November, 2000, to July, 2001) since the markets had collapsed and, in particular, the internet bubble had burst. Many VCs told us that they were not making any investments at all or were only interested in B2B investments that they perceived to be less risky than consumer propositions. After all, who could really predict the fickle consumer?

After pitching a very large number of VCs, we finally were able to close our Series A venture financing. We closed it literally weeks before the tragic September 11, 2001. Our lead venture investor was a music loving business executive-turned-VC named Ajay Chowdhury from IDG Ventures Europe. He was very passionate about the potential for Shazam and could see the tremendous opportunity that it could have among consumers.

Newnham: How did you decide what you needed, and how was it distributed within the business?
Barton: They always say, “Raise more money than you think you need,” and it is so true. In our initial angel fundraising, we thought $200,000?—?$300,000 would be plenty to start with. Fortunately, we had excess demand from angels and our bridge financing structure allowed us to raise $1 million, which we ended up doing. This was really helpful since the venture fundraising process took much longer than we expected.

In order to raise money from venture capitalists, we built detailed models of our financing needs assuming a specific hiring plan and a certain amount of technology investment. This was completely new territory for me, so Dhiraj’s experience, as well as his network of contacts from his prior job, was extremely helpful for this. Philip was a star spreadsheet whiz and turned the whole thing into a beautiful Excel model.

We built a list of venture capitalists to approach. Most were in London, but a few were in Munich, Paris, Stockholm, and, of course, Silicon Valley. The US venture capitalists had absolutely no interest at the time in consumer mobile businesses. Europe was way ahead in this area, so we doubled down on European VCs. We refined and refined our PowerPoint presentation so that it answered every concern that we would hear from potential investors. We were able to get meetings with pretty much everyone that we wanted to, but getting them to make an investment proved incredibly difficult.

When a VC was somewhat interested, they would say to us, “We are interested but we don’t take lead. Let us know if you find a lead VC.” The lead VC is the one who makes the bulk of the investment and sets the price and terms. Ultimately, if the investment fails, they have taken the most risk in terms of capital and reputation. We found that many VCs love to follow (co-invest alongside a lead) rather than lead because, that way, they avoid these risks. The real hero VC is the one who takes lead. They are the first person on the dance floor. They have the guts and passion. To our great surprise, however, they are few and far between, but fortunately, we did find our lead (Ajay Chowdhury).

With the angel financing, we were able to actually start spending some money. We rented our first office in the middle of Soho and recruited three engineers who happened to become available as Viant closed down some of its offices. One of our angel investors was running Viant’s Munich office, so he helped us hire these three engineers who lived in Munich. We also recruited a superstar ex-Apple guy in Silicon Valley to help build our cluster search engine to identify music. So, now, we were made up of three people in Palo Alto, three people in Munich, and two people in London. Everyone worked for no more than half of their market-rate salaries and instead received equity incentives. Everyone was incredibly passionate.

Once we had the $7.5 million in venture capital funding in July, 2001, we immediately began to build the business so that we could build toward launch. Our budget told us we had enough money to take us a little past twelve months. We had a very expensive business to build. Everything in technology was more difficult and expensive back then. There was no Amazon AWS or S3, so we had to budget to buy our own server farm. We had to architect and build a cluster search engine by custom-crafting racks of PCs and creating a cluster search software to run on it.

Newnham: Avery, I understand that a working demo was shown at the investment pitches. Can you talk me through the actual process of developing the algorithm and what it felt like to have a working demo?
Wang: As I said earlier, there was a lot of pressure for coming up with a working algorithm. At first, I was kind of amused and was just going to give it a shot. But as the guys were raising money, getting office space, and hiring people, it became increasingly real and unfeasible to back out. The guys were doing OK going around showing slideware without having a demo?—?I admire them for that?—?but eventually, we would have to show something before going much further. After a few months of working on the problem, I was getting ready to give up and tell the guys that I couldn’t do it. Chris, however, wasn’t convinced that it was impossible.

It’s hard to describe exactly how to create on-demand an algorithm with radically different properties from everything that has been done before. Usually, a technology exists first before a team tries to commercialize it. Here, we did things the other way around. As I partially described earlier, working with Julius, we first listed the attributes that the algorithm needed to have. Then, we started looking at algorithms from other domains that had bits of the desired properties. I looked at hashing algorithms used in text search engines. I studied previous research in the content recognition but concluded that none of the known techniques would solve our needs. We looked at speech recognition algorithms but saw many shortcomings. I kept mulling over the list of requirements and let it stew for a while. I didn’t actually try implementing any of the known techniques in the literature to try out on the problem; I figured they weren’t worth trying, since they weren’t close to what we needed them to do.

We collected audio samples from mobile phones recording music being played in a variety of conditions, including noisy pubs and streets with traffic. When I listened to the audio samples, I was really worried because in many of them there was no audible music, just garbled sounds. A fundamental obstacle we had with mobile phones was that the recording system uses a cheap tiny microphone and applies low bit-rate compression that is good enough for the limited spectrum of voice and thus effectively chops off chunks on both ends of the music spectrum, sometimes along with squelch or background noise removal. I started to believe that it really wasn’t possible and that the researchers who said it wasn’t possible really did know what they were talking about. Discouraged, I still trudged on, motivated by not wanting to let the guys down.

As for the process of developing the algorithm, I reasoned that in order to have reproducible descriptions of audio features, there needed to be reproducible locations in the audio files called landmarks. An example of landmarks is energy peaks in the audio. Around each landmark there must then be a description of the local signal, which I called fingerprints. With Julius’s help, I focused my efforts on different formulations of landmarks, as well as fingerprints. On both fronts, we had various ideas and limited success?—?until, one day, I was plotting the time scatter plot of matching fingerprints. It was a straight diagonal line of data points in the plot. When I saw it, I knew we had a powerful statistical indicator of a match and that the algorithm could be made to be very fast. What I saw on the graph immediately suggested that the problem was solvable!

Newnham: On to the crucial partnerships. You made some great deals with big organizations early on. How did you secure the meetings, and how did you convince them to do such a deal when you were a relatively unknown startup?
Barton: The primary partnerships that we put in place in order to launch
Shazam commercially were 1) a partnership with an IVR provider; 2) a partnership with an SMS aggregator for premium SMS integration; and 3) partnerships with the four leading mobile operators in the UK?—?Vodafone, Orange, T-Mobile, and O2.

While the partnerships with the IVR and SMS aggregators were relatively simple supplier partnerships, the bigger challenges were the mobile operators. With the mobile operators, we wanted to secure a four-digit phone number that would direct phone calls from any mobile phone to our IVR system. It had to be the same four-digit number on every mobile operator, much like 411 or 999. Ours would be 2580, which were the only four digits that were in a straight line on the phone keypad. This idea was the genius of Simon Coley, who was the ex-Viant creative consultant that helped us. Only one company had ever secured such a four-digit number prior to us?—?and that was the traffic information company.

Mobile operators were very large and complex organizations that were not accustomed to working with startup companies at the time. They knew that we were an unknown startup and could disappear in no time; as a result, they were very reluctant to spend time and resources to create the custom integration that we sought. It was pure willpower and persistence that enabled us to succeed in securing these partnerships. We had dozens of meetings with numerous people at all four mobile operators and simply did not give up. We tried to create urgency with them by noting that they would not “want to be the only UK operator who did not have music recognition” when we launched. This was our only negotiating leverage, and it finally worked in time for our launch.

Newnham: Once you had everything and everyone in place, you launched the service in 2002. Can you tell me what the build-up was like and how you felt on launch day?
Mukherjee: I played a variety of roles at Shazam, from registering the company, writing the business plan, working with our lawyers on the legals for our angel round, fundraising along with Chris and Philip, and taking on marketing and product management roles as we built out the team. In the spring of 2002, we brought Jerry Roest on board as CEO, and he took charge of the business with admirable aplomb.

One of the first things Jerry asked me to do was to work out exactly what was needed to launch and to get us there. We had almost inevitably been hit by setbacks and delays?—?from the time it took to negotiate with mobile operators, the difficulties building a large enough music database, and the challenges of the core recognition service itself. I had managed medium-sized projects before in my previous job at Viant, but I had never coordinated something so big?—?40-something employees, plus 50 or 60 temporary staff feverishly working away.

Jerry told me to create a spreadsheet with every single task that needed to get done for launch, work out how long each one would take to complete, and come up with a date. I spoke to all the Directors in depth?—?Business Development, Music, Engineering, Marketing, Operations. The launch plan grew and grew; by the time I was done, it was about six feet long. I printed it up and posted it on a wall where everyone in the company could see and read it, and it stretched from as high as I could reach to the floor.

My original plan aimed to launch on Friday, August 16. I talked it over with Jerry, and he agreed with my estimates but said that Friday was not the best day for a launch, so we moved it to the following Monday.

Every day in the build-up to launch, I would walk around the office checking on progress with teams, updating the status and estimates, and tracking the critical path. Piece by piece, we crossed off what needed to be done. When we hit roadblocks, we found creative workarounds. Everyone in the company knew exactly what needed to be done and by when. I would get into the office by 9:30 AM and worked until about 8 PM. After dinner, I would get back to work by 9 PM and work until about 3 AM. On Saturdays and Sundays, we only worked eight or ten hours a day.

My last conversation about launch with Jerry was late in the evening on that Friday night. There were a few last-minute glitches but nothing fatal, in Jerry’s view. Then, on Monday, August 19, exactly on schedule and 13 months since our Series A funding, we launched Shazam to an unsuspecting British public.

The launch day itself was almost surreal. Our CEO, Jerry, was on a television interview, wearing a suit and tie. The lights on our recognition cluster started to light up as calls came in from real users. That evening, we all trooped down to a local bar to celebrate reaching this milestone. Our lawyer had not slept a wink the night before. By 6 PM on the night of launch, giddy from the euphoria and lack of sleep, he appeared again and again with a tray full of indecently sweet shots. That evening has passed into Shazam folklore, and years later, many of us still celebrate the occasion with a mix of butterscotch Schnapps and Bailey’s and recall the buzz we felt that summer night in Soho.

Newnham: What did the service offer on day one?
Barton: When we launched, 95% of people with mobile phones in the UK (everyone except Virgin Mobile customers) could simply dial 2580 when they heard music in a bar, café, restaurant, club, or on the radio and hold their phone to the music. Fifteen seconds later, the voice phone call would terminate and a text message would be sent to their phone with the name of the artist and song. There was no charge if we could not identify the song. If we could, then they were charged 50 pence on their mobile phone bill. The service was branded as Shazam from day one. We ran advertising in many places to let people know about the service. Soon, we would also enable people to buy ringtones related to the songs that had been identified.

The next generation of the service several years later would be mobile applications (BREW and Java) that had a richer experience and, in some cases, enabled unlimited use for a subscription of around $3 per month. However, it was not until the advent of the iPhone App Store that Shazam’s user base growth hit the hockey stick with astronomical growth.

Newnham: Chris, you earlier touched on some of the tough times the business went through. Can you tell me how you got through them as a team?
Barton: The toughest times for us were the years after our launch, when we needed continued venture capital investment since we were not profitable. We raised multiple rounds of venture capital during these years, which was testament to the fact that people still believed in the business but that the company had not yet reached rocketing user or revenue growth, so these were not easy investments to agree terms on.

We also had our share of layoffs in the early years. This was painful; we had no rotten apples in the organization, and many of our employees had become close friends. The most valuable thing we had was the high level of integrity and trust that we had for each other among the founders, Shazam employees, and Jerry. During times like this, there is inherent conflict between the interests of the venture capitalists and the founders.

Ultimately, everything is a negotiation, and we made it through these tough times by thinking rationally, finding compromise, and remaining passionate about our dream. The most dangerous outcome for a startup during times like these can be a fallout between the founders, but we never had this issue because there was such a powerful bond between the four of us.

I think it definitely helped that we started as friends, as it made it easier rather than harder during low points. It is during these low points that your friendships are often the bond that keep you together. If we had only pure business relationships with each other, there could have been really destructive outcomes that led to the company collapsing.

Newnham: When did you all leave, and how was it to leave a business you started together?
Barton: All four of the founders left the company at different points and for different reasons. Avery eventually rejoined and is an employee today.

We all left on good terms and went on to do very exciting things. I personally joined Google about six months before it went public, when it was only 2,000 employees. Philip stayed with Shazam for another year and built the B2B business. Some of the resulting partnerships brought in a lot of cash, which was extremely important to the company’s survival. Philip also eventually ended up at Google and plunged back into startups only a few years later (TrueCar and RockMelt).

I won’t lie: it was definitely hard to leave the company, but I have always retained my passion and remained dedicated to the company through continued involvement at the board level. It is a real pleasure to see the company thriving today, and I also have found that many of the learnings I have taken from Google and Dropbox have helped me in guiding Shazam to a similar level of excellence through the board.

Newnham: With hindsight, is there anything you would have done differently with Shazam?
Barton: In an ideal world, Shazam would have started about five years later and therefore missed many of the tougher times while waiting for mobile phones to catch up with our vision. Also, in an ideal world, we would have run more leanly to make our venture funds last longer, but this was not easy given the hurdles of what we had to build. I also think it would have been smart to begin building our B2B licensing business even earlier than we did because it was critical to helping us become more financially viable in those tough times. Google did that very well by focusing early on with its white label B2B business to bring in revenues (running search for Yahoo).

Newnham: What did you all take away from the Shazam experience? What did it teach you about business and friendship?
Barton: If you are going to do a startup, you must make sure you are truly passionate about the business because you are going to need that raw passion to power you through the really tough challenges. There were other businesses I considered other than Shazam that would not have brought that same passion for me. Selecting the right co-founders who share passion, drive, results, and integrity is critical, and your efforts could be in vain if you don’t pick the right people. We were very fortunate to have an incredible founding team chemistry.

Building a startup from the ground up is a life experience, and you must always think of it this way. There is a much higher chance it will fail than succeed, so you want to be sure you are benefitting from the experience in many other ways, including learning, excitement, and other life experience benefits. Don’t lose sight of those things, as they are 90% of the value.

As I have taken new roles, I have sought out companies that treasure and value entrepreneurial spirit. Both Google and Dropbox are places where entrepreneurial people who want to help build new things and just make things happen thrive. I learned from Shazam that working in that type of environment with those types of people is a night-and-day difference from most companies in terms of realizing tremendous reward in your experience and satisfaction.

The number one determinant of entrepreneurial success is persistence. If you are not prepared to go to superhuman levels that are beyond rationality to realize your dream, then your chance of finding success is virtually zero.

“A team is stronger when people think in different ways: it is much easier to work with people who are similar to you, think in similar ways, or agree with your views. However, an effective team will bring together complementary skills and strengths and perspectives.” Dhiraj Mukherjee

Mukherjee: Have fun: I believed it before; I believed it during; I believe it now. As long as you are having fun, you will make good decisions, manage stress, not take things too much to heart, and remember that there are more important things in life than a business: true friends and family who will support you regardless of your success or failure.

Act on your values: this is hard but is eminently worthwhile. Values speak in a quiet voice that can get drowned out by inexperience, uncertainty, fear, greed, bad advice, or group-think. Being able to tune into one’s values and act according to them scrupulously is an infinite source of strength.

Never take no for a final answer; never accept defeat; never give into fear. When you do a startup, everything is stacked against you. You have no money. You have no product. You have no customers. You have no track record. You have no proof of anything you claim. Everything you do is met by resistance, skepticism, and rejection. If you give in to it, your career as an entrepreneur will be short and unfulfilled.

Never drop your standards: if you are trying to move quickly with limited resources in a competitive market, it can be tempting to take shortcuts and let standards slip. Don’t. Quality is a mental model, a habit that should not be compromised even under extreme pressure.

A team is stronger when people think in different ways: it is much easier to work with people who are similar to you, think in similar ways, or agree with your views. However, an effective team will bring together complementary skills and strengths and perspectives.

Building up trust and camaraderie also builds resilience, fortitude, and the will to succeed. Bonding as friends brings strength to the team when superhuman efforts are needed.

Inghelbrecht: Conventional wisdom will tell you that business and friendship don’t go together. Shazam has proven this wrong, and as a matter of fact, business and friendship mixed together can reinforce each other. Our friendships are stronger today because they survived some nasty business fights. As a business, Shazam has made better decisions along the way because we were able to support, listen, and compromise as real friends would do to each other.

Newnham: Mobile has skyrocketed in recent years. What advice would you give to someone with a groundbreaking idea?—?someone who wants to create a business that they believe will change the mobile landscape?
Mukherjee: My advice is not to look at the market today but to aim for what the market will look like in the future. When Apple launched the App Store, they changed the game. Launching a new app is not a breakthrough, but if the idea has the potential to be a game-changer, then it is worth going after.

I would also pay very close attention to cash: where is it coming from; how is it being spent; how long is it going to last? A solvent business can make mistakes, course-correct, and may eventually succeed. An insolvent business is most often the end of the dream, which is no fun at all.

Don’t assume that the natural laws of startups won’t apply to you. Most things take longer and cost more than you expect. Build that into your planning and thinking, and don’t let it surprise you, or worse, bankrupt you.

“Think big and shoot for the moon; plan for every person to use your product at some point in time. It’s much better to fall short on a lofty goal than to exceed a reasonable projection.” Philip Inghelbrecht

Inghelbrecht: Like Dhiraj, my advice would be to look way ahead of your time, at the risk of launching a company with a slightly wacky idea or incomplete technology stack. While things always take longer to materialize than initially expected, they do come together quickly when it does happen. Being there at that moment will pay off handsomely e.g. Shazam was built before the iTunes store, not in reaction to it.

Secondly, think big and shoot for the moon; plan for every person to use your product at some point in time. It’s much better to fall short on a lofty goal than to exceed a reasonable projection. This is what sets people like Larry Page or Elon Musk apart from other entrepreneurs.

Barton: My biggest piece of advice is to be careful about the assumption of “If I build it, they will come.” In the mobile application ecosystem, 99% of mobile applications never get significant user adoption. Some of these mobile applications probably have great potential if they are able to sustain and iterate, but they end up giving up because they can’t get user traction.

Building the next Instagram, Pandora, or Shazam that somehow organically achieves tremendous popularity is like winning the lottery. Even a delightful, beautiful, simple, and compelling mobile application will most likely get lost in the huge, long-tail desert of undiscovered mobile applications.

A mobile entrepreneur should make sure that part of their innovation includes a path to acquiring users and driving usage. Getting these users through an innovative channel that is free will be the key to such a successful business.

---

This excerpt (edited for brevity) is taken from Mad Men of Mobile, available on Amazon, and was originally posted on Medium.

My second book, a collection of one-on-one interviews with female founders and innovators in tech, will be released Spring 2016.