Follow Reaction Wheel on Feedspot

Continue with Google
Continue with Facebook


I submitted the grades for my class, and suddenly found myself a bit idle. So it’s time for more things you already knew proven with math you long-ago forgot. This was supposed to be a mere bagatelle of a post, something to occupy an idle Friday afternoon. And it would have been if the distribution […]
Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 
Reaction Wheel by Jerry Neumann - 4M ago

I was responding to a pitch this week. Stellar team, big market, demonstrated customer need. I wasn’t going to invest, despite all that. I wrote: “not the kind of thing I invest in.” I’ve written this thousands of times in my life. But I knew that wasn’t really the reason, so I erased it.

I wrote that the idea wasn’t risky enough and the things that have made me money have all been very risky. But if they then assured me that it was far riskier than I thought I would like the company even less, so I erased that. Maybe it was that the ineluctable risk was too low? That felt like an invented reason. Perhaps it was that the idea just wasn’t crazy enough. But the founders had gone to the best schools, the best business schools, worked at the best companies. What would it mean telling them they weren’t crazy enough? Going after a huge market with formidable competitors was kind of crazy; I’m not really looking for crazy.

Gut decisions are bad practice. I try to make gut decisions not gut decisions. When I have a gut feeling about something I interrogate my gut: what is the reason behind this feeling? If I can dissect the feeling into underlying reasons, I can reason about them, I can see if they make sense or if they’re a statistical artifact of a low sample size. If the latter, I change my mind. I do this all the time. Gut decisions are unreliable.

This time all I could come up with was what I started with: this is not what I do. They were amazing businesspeople in a large market whose product addresses a demonstrated customer need. But I don’t look for amazing businesspeople in large markets whose products have a demonstrated customer need.

Venture capitalists will love this company: they will shower them with money so the company can spend years grinding through a recapitulation of what already exists so they can marginally displace a major incumbent and scrape a few hundred million dollars from the billions the incumbent sees. This will be enough, and the founders will be rich and the VCs will be successful and the world will spin on as before. This is venture capital at its best as practiced today. Read about the companies being funded every day by venture capitalists and tell me this isn’t true. I used to think that what I did was venture capital. But maybe my idea of venture capital was just wishful thinking, a vision of some golden age that never really existed. I don’t care about business acumen, I don’t care about proven markets, I don’t care about making a ton of money. I don’t. I don’t think I rise to the level of professional investor, someone who takes their fiduciary duty to maximize risk-adjusted returns seriously. I’m not sure I’m even really a venture capitalist at all.

So I told the company that I was no longer a venture capitalist, that I had retired. And I suppose this is a lie too, because you can’t quit something you never were. But when I wrote it I felt how a boiling frog who’s jumped out of the pot must feel. Relief to no longer have to think about myself as a venture capitalist, to try to shoehorn what I want to do into what I am supposed to do.

In the next couple of weeks I will be investing in a company I’m really excited about. A couple of post-docs with an elegant technology for a small market that might be made large by it. No customer has asked for this, because no customer knows it’s possible. This company is neither risky nor crazy. It probably will fail, and if it fails it will fail quietly. But if it works, the world will be different. This is the kind of thing I want to invest in. I once thought this was venture capital, now I’m not sure what to call it.

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 

My short post a few weeks ago on one of Schumpeter’s ideas lightly suggested you could derive most of modern business strategy from that one idea. I suppose this might technically be true if you are a genius with a lot of extra time, but why reinvent the wheel?

Last week I taught my class at Columbia’s engineering school about industry analysis. I used the Schumpeter idea as the starting point  for why entrepreneurs need to analyze the industry they want to enter but then quickly segued into Michael Porter’s work on business strategy for the rest of the why and much of the how. One of the readings was Porter’s classic article “The Five Competitive Forces That Shape Strategy” and you should read that before this. (You can see the rest of the syllabus here.)

Porter is business strategy from 30,000 feet. His tools give a great overview, but because they are so general they can be hard to apply to specific situations. Regardless, if applied with rigor–meaning with actual data and analysis, not anecdote and belief–you can glean a ton of insight reasonably quickly.

Below are the slides from the class. They are meant to:

  1. Introduce the idea of innovation as the source of excess profit;
  2. Show that innovation alone is not always a sustainable competitive advantage (“SCA”);
  3. Talk about how new value propositions supported by changes in value chains can create SCAs;
  4. Convince my students that to be able to find new value propositions and tell if the value chain to produce them is sufficiently differentiated so as to allow a SCA they must do some extensive industry research;
  5. Talk in more depth about Porter’s Five Forces as one of the tools to guide industry research.

Much of the material here is from Joan Magretta’s excellent overview of Porter’s work, Understanding Michael Porter. The class had read a case study I wrote about Google’s entry into the search engine market, so I used Yahoo! as the exemplar of the market prior to Google and Google’s new value proposition and value chain as examples. I’ve interspersed a few sentences explaining what I was going on about, but I hope otherwise it’s self-explanatory.

nb: One student objected that this analysis neglects the not-for-profit sector. It does. Porter’s methods can be applied to non-profits, but I’m not knowledgeable enough to do so.

Thank you to Benedict Evans and Justin Wohlstadter, who helped me find Mary Meeker’s late-’90s Internet Reports.

I know this example is toy, but it presents the issues of competition and innovation pretty starkly. It actually worked pretty well.

This below I talk about in the Schumpeter post if you want more explanation.

“Competing to be the best” is Porter’s way of talking about competition that involves no or easily copied innovation. All of the value created this way is garnered by their customers. This is good for the customers of course, but means that there is no return to the business on the time and money invested in this value creation. This is one of those places where I don’t know how to apply Porter’s ideas to not-for-profit enterprises.

One way to avoid competing to be the best is to segment the market, and we talk about that a bit. It is less interesting to me than creating an entirely new market, but it’s probably more common. In the second slide we talk about the Boston Beer Company (maker of Sam Adams) segmenting the beer market and in the third we use Andrew Parker’s classic slide to show how focused competitors picked off pieces of Craigslist as those markets grew big enough to be standalone.

I can’t imagine this diagram is actually as universal as business strategy classes would have you believe. Nevertheless, it’s a good starting point.

You don’t really start your analysis with the five forces, you should first gather the data. I did it first in these slides because I always run over time and I needed to make sure we talked about this and value chains in this class.

By “covering framework” I mean that the framework covers all the issues and that it’s not worth arguing about the fine points of which thing goes in which box. This comes into play later when I put the users of Google (ie. the searchers) in the “Buyers” box, even though they are technically more like suppliers than buyers.

The next five slides are direct from Magretta, an invaluable reference on the five forces.

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 

Columbia University
School of Engineering and Applied Science
Department of Industrial Engineering and Operations Research

IEOR 4998                                                                                                      Jerry Neumann
Spring 2019                                                                                                   TA: TBD

Managing Technological Innovation


Course Description
This course provides an introduction to and overview of technological innovation in small, fast-growing enterprises. We will explore both the theories of innovation as well as the strategic and tactical approaches and processes that these theories entail. We will discuss, both amongst ourselves and with speakers from startups and venture capital firms, the practical realities of implementing these approaches and processes in today’s entrepreneurial environment.

The textbooks for the class will be:

  1. Osterwalder, Alexander, and Yves Pigneur, Business Model Generation, John Wiley and Sons; 1st edition (July 13, 2010).
  2. Fitzpatrick, Rob, The mom test: how to talk to customers & learn if your business is a good idea when everyone is lying to you, CreateSpace Independent Publishing Platform; 1st edition (September 10, 2013).

There will be handouts and case studies distributed through Courseworks. There will also be several case studies you will need to buy through the Harvard Business School Publishing site, the link will be given on Courseworks. Other assigned readings will be available on the internet for free.

The readings do not generally duplicate what we talk about in class. Because we are learning in large part through case studies, you need to do the readings so you can participate in a meaningful way. The readings provide you with the tools to do the analyses we will work through in class.

The final grade for the course will be a weighted average of your grades on the group project papers, an individual paper, in-class participation, and homework.

Group paper 1             20%
Group paper 2             20%
Group paper 3             20%
Individual paper          20%
Participation & HW     20%

Group Papers
You will form a group of three or four people from the class. The group will choose an idea for a startup business and will use the strategies and processes we learn to explicate this business opportunity. There will be three group papers.

Individual Paper
You will critique the business plan of a startup company (list of companies to be provided towards the end of the semester) using the ideas, tools, and techniques taught in this class. This paper will be an individual effort.

In-class participation
You are expected to read regularly the assigned material before each lecture and participate in the general discussion during the lecture. Participation in case-study discussions and in-class group assignments is critical to the learning process.

Because of the importance of participation and the interactive nature of most of the class, don’t use laptops and phones during class. Class slides will be posted after every lecture, so taking notes may not be necessary (but is up to you.)

Written assignments to be handed in for about half the classes. These are meant to familiarize you with the problems to be discussed in the next class. As such, they will be graded with a check, a check minus, or ‘not handed in.’

Everything you hand in should be through Courseworks. Please put your uni in the filename of homeworks and the individual paper and make sure that everything you submit has both your uni and your name in the body of the submission.

Attendance is mandatory. Our teaching method is interactive, if you don’t come to class, you don’t learn. If you can’t make it and have a good reason, email me and the TA. I understand if you need to miss a class or two.

On the days we have a speaker, if you can’t make it into your seat by the scheduled class time, please wait until the speaker leaves (after about an hour). Coming and going while an outside speaker is speaking is rude.

This class is unlike most engineering classes: we are learning a way of engaging with society across a broad range of poorly-specified problems. As such, communicating is important, both in writing and verbally. Real-world data, analysis of extant data, and expert third-party opinions are convincing, your personal opinion less so, unless you happen to be an acknowledged expert on the issue being discussed.

For clarity and conciseness, I prefer a writing style that is heavier on bullet points, charts and other graphical demonstrations, and analytical argument, not the first-person essay-style writing generally taught in high-school.

Schedule of Meetings, Readings, and Assignments

Section I: Problems and Solutions


January 25, February 1, February 8, February 15


Introduction to MTI

What this class is and what you will have to do
What is technology, what is innovation
Sources of innovation


Problems and Solutions

Google Case
What do startups do?
Customer decision making
The competitive environment
Founder-market fit

Readings for this class:
– Google case, on Courseworks
– Altman, “Startup Playbook

Put together a pithy (less than one page, PDF) introduction to yourself with a recent picture of you that will help me recognize you in class. Include your UNI, your name, what you’re called, your school and major.

For my information to assess the level of knowledge of the class in aggregate, I’d appreciate if you could also note any entrepreneurial/startup experience you might have, and any entrepreneurial/startup/business classes you have taken.


Industry Analysis

Basis of competition
Competitive positioning and market segmentation
New markets
Market creation vs. discovery

Readings for this class:
– Porter, Michael, “The Five Competitive Forces That Shape Strategy
– Alvarez and Barney, “Discovery and Creation: Alternative Theories of Entrepreneurial Action
– Anderson and Tushman, “Technological Discontinuities and Dominant Designs

Do a Five Forces analysis of Google Search at the time of its introduction.


How to Think

Guest teacher: Justin Singer

– Sterman, “All Models Are Wrong
– Heuer, Psychology of Intelligence Analysis (Part I only)
– Costa, “Describing the Habits of Mind
– Menand, “Everybody’s an Expert


Patterns in Technological Innovation

iTunes Case

Readings for this class:
iTunes Case
iTunes Case, Technological Innovation
Christensen, “The Evolution of Innovation
Aldrich and Fiol, “Fools Rush In? The Institutional Context of Industry Creation

Answer one of the case questions. Be prepared to discuss the others in class.


Group Paper 1 due March 1


Section II: Experimentation and Feasibility


February 22, March 1, March 8, March 15, March 29
(March 22 is Spring Break)


Business Models

Business Models
Business model canvas (and customer archetypes)
Financial models

Readings for this class:
Osterwalder, Business Model Generation, pp. 14-119, and pp. 200-239.

Prepare a business model canvas for a product of your choice (don’t use one that’s in the book.)



MVP & Iteration
Key Performance Indicators
Measuring progress

Readings for this class:
Ries, “The Promise of the Lean Startup
Fichtner, “Lean Startup 101 for Developers
Thompson, “Building a Minimum Viable Product? You’re Probably Doing it Wrong


Financial Analysis on the Fly

Market size
Unit economics
Lifetime Value
Cost of Customer Acquisition

Readings for this class:
– “Unit Economics & Lifetime Value”, in the Files section
– Blank, “Market Size Hypothesis
– Mauboussin and Callahan, “Total Addressable Market

Other resources (not required, but if you want more info on market sizing):
– Gentschev, “The Secrets of Market Sizing
– Gentschev, “A Full Market Sizing Example
– Preuss, “Modeling Total Addressable Market

Do a market size analysis of one of the following:
– NYC pay-per-ride scooters
– Quantum computers
– Martian colonization (just number of “customers”, not potential revenue)


Circle Lending Case

Readings for this class:
The Circle Lending case (on the HBSP site)

Prepare one of the following for Circle Lending:
– A market sizing, both top-down and bottom-up
– A business model canvas, with a one-page discussion of value propositions and associated customer segments


Group Paper 2 due April 5


Section III: Building a Product and Company


April 5, April 12, April 19, April 26, May 3


Defining the Product

Customer interviews
Product management
User stories

Readings for this class:
Fitzpatrick, Rob, The Mom Test

Prepare a customer interview script as if you were researching the idea for Circle Lending or another startup of your choosing.


Medium Term Planning and Analysis

Fishbone Analysis
Financial models
Analyzing results and changing course

Readings for this class:
 “Building a Financial Model”, in the Files section
Yoskovitz & Kroll, “Lean Analytics” presentation


Zipcar Case

Readings for this class:
The Zipcar case


Do one of these analyses and hand it in online. Bring a copy to class so we can discuss.
– Fishbone analysis
– Model the Zipcar LTV and COCA with appropriate assumptions (make sure you note your assumptions in detail and provide either support from the case or some other rationale.)


Financing the business

Corporate structure
Venture capital
Employee stock options
Other means of financing

Readings for this class:
“Explanation of corporate structure and equity”, in the Files section
Zider, “How Venture Capital Works
Dotzler, “What do venture capitalists really do, and where do they learn to do it?
NVCA 2018 Yearbook (will be in the Files section once released)


Group Paper 3 due May 3


Individual Paper due May 10

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 
Reaction Wheel by Jerry Neumann - 5M ago

It’s Friday and I’m procrastinating, so here you go.

Let’s talk about Joseph Schumpeter. Good old John Joseph Jingleheimer Schumpeter, as he wasn’t called. Schumpeter once wrote in his diary that he aspired to be the greatest economist, horseman, and lover in the world1. I can imagine the women and horses edging away nervously. Luckily he had it going on with the economics.

The mainstream of economics, then as now, pretty much tries to describe the economy as if it shouldn’t change. If it is changing, it’s changing towards an equilibrium, where it won’t have to change any more. Schumpeter noticed that this is not how it works2. Both the economy as a whole and individual businesses change constantly. His model of the latter, in his Theory of Economic Development3, explains how some entrepreneurs make an unusually large amount of money.

I would quote the book itself, but the argument is spread out over the course of the chapter. Schumpeter wasn’t a bad writer and the chapter is worth reading, but he never really summarizes his main points. So this is my recapitulation of it.

There are three main parts.

First, almost all entrepreneurs don’t make an abnormal amount of money, even of the successful ones. They make the same amount as if they were doing the same job for someone else. This is not what our entrepreneurial mythology tells us, so some explanation.

In a market economy, at equilibrium, Schumpeter says profit gets competed away. By profit he means “surplus” profit: the money a company makes if its inputs are priced correctly. Crucially this includes the cost of money adjusted for the risk the investor is taking. That is, you can’t increase risk and say “look, now there’s a profit.” That profit is the cost of the money used in the business.

The reason this is true is that if a company produces something using the same inputs as its competitors, and has the same outputs as its competitors, then the costs of its inputs and the price of its outputs are the same. If there were profit above the cost of capital, “surplus” profit, then existing businesses would lower prices to get more customers, or new businesses would start to take some of the surplus profit, until there was none left to take.

These new businesses would be startups, and their founders entrepreneurs. But these entrepreneurs would earn no more than they would if they did the same job as employees for someone else. This is because even the founder, as manager of the company, is an input, just like the other employees. The founder makes the same amount of money for their job as they would working the same job in any other business..there is no other money to make, there is no surplus for the founder. There is no “entrepreneurial profit”, as Schumpeter called it.

Is this true in the real world? Below is a chart from Scott Shane’s The Illusions of Entrepreneurship.4 It shows employee income compared to entrepreneur income, by decile. Note that for the middle eight deciles, they are the same. 80% of entrepreneurs make the same amount of money they would if they were employed.

The vast majority of entrepreneurs are people creating their own job so they can work for themselves.

Shane notes that the median revenue of an owner-managed firm is $90,000 and that 81% of founders have no desire to grow their business. This is because most founders are “just trying to make a living, not trying to be a high-growth business.” And they “start firms in industries where there are a lot of firms already in operation” and “report they have no competitive advantage.”5 Why do these people go to the trouble then of starting their own company rather than just taking a job? “The real reason most people start businesses…has nothing to do with wanting to make money, to become famous, to better their own communities, to seek adventure, or even to improve the world. Most people start businesses simply because they don’t like working for someone else.”6

Obviously, some entrepreneurs do make a lot of money. This is the second part of Schumpeter’s argument. Those that make money, an entrepreneurial profit, do so by breaking the status quo. They innovate. They either get their inputs for less or they sell their outputs for more.Their innovation is either an efficiency innovation that allows them to create the same output with less input, or a value innovation that allows them to create a better or different output (that the can therefore sell for more) with the same input. Or they have a little bit of both. This allows them to create an entrepreneurial profit.

Third part of the argument: this entrepreneurial profit goes away over time. Competitors figure out that there is this extra money and they imitate the innovator. When this happens, the surplus or excess profit is worn away as imitators enter the market and compete with the innovator.

Fyi: this diagram is not to scale, and the decay rate is just notional.

The total excess value created by the entrepreneur here is the area under this curve. This value is somehow split between the various participants in the startup–the founder, the employees, the financiers, etc.

So this is a kind of cool, and perhaps somewhat obvious in retrospect, model. Innovation leads to excess value that is then distributed mainly to founders and VCs. Using it leads to a couple interesting lines of thought.

It explains most of mainstream business strategy

As the second diagram above shows, there are two ways to create excess profit through innovation: by using innovation to lower cost or by using innovation to create a product you can charge more for. Doing one of these or both of them is integral to breaking out of the pack of competitors. This is one reason businesses focus so heavily on innovation. The bigger the innovation, the greater the excess profit, but even small innovations that result in a slightly lower price or slightly improved performance over the competition is important.

Interestingly, these are the only strategies to create excess profit that Schumpeter’s model allows. You might notice the resemblance to Michael Porter’s “generic strategies” (from his book, Competitive Strategy, considered one of the most important books on business strategy of all time.) Porter identifies three “generic” business strategies in the book. He calls them “cost leadership”, “differentiation”, and “focus”. The first two are the strategies Schumpeter called out. The third, “focus”, counsels a company to focus on a sector so they can get to some combination of the first two strategies. An innovation that allows one of these three strategies Porter calls a “competitive advantage”.

The third diagram expands on this. Since the total value is the area under the excess profit curve, you increase total value realized from an innovation by slowing the decline of the excess profit and extending the period until it is competed away. Doing this creates a “sustainable competitive advantage” (also commonly called a “moat”.)

One way to do this is through hampering competition with regulation: intellectual property or licensing standards or whatnot7. Another is to continuously improve the innovation to keep ahead of the competition. If you constantly tinker to improve, “learning by doing” then you benefit from the “experience curve“.

Another way is to slow competition is to divvy up the market between existing competitors so that none of you are really directly competing. This is generally illegal if it involves colluding, but a legal way to do it is to segment the market into slightly different products: sports cars for one company, pickup trucks for another. This is called “positioning”. Positioning can even be done into positions that didn’t previously exist: “blue oceans”. Just enumerate the various attributes your product can meaningfully differ on and find an unexploited combination that your company can excel at. Voila, zero competition…for a little while.

At the corporate level–that is, above the product level–you want to make sure that when one source of excess profit is petering out, you have another ready to go. You do this by investing in innovation. Some of this investment is in existing products, some of it is in new ideas. Rationally, you should invest the most in products that can be developed into the largest profits. The Boston Consulting Group’s growth-share matrix (“cash cows”, “stars”, “question marks”, and “dogs”) gets at this.

Etcetera. It’s too bad Schumpeter didn’t write business how-to books. He would have made a killing.

There are two fundamentally different types of startups

We call all companies started by entrepreneurs “startups”, but the kind that is started as a job replacement and the kind that is started to create entrepreneurial profit are different animals altogether.

As an aside, before I launch into this, don’t think this is a dis. I have started the first sort of company, the job replacement type. It happens to be in the venture capital space, so sometimes I confuse myself with the kinds of entrepreneurs I back–who start the second kind of company–but I would probably take home the same amount of money as an employee at a fund, though I would then have to work for someone else, which would suck.

Anyway, back to the two kinds of startups. The first kind is incredibly important for the people who start them. Starting your own company is, in the words of the Kauffman Foundation, “a self-actualizing and a self-transcending activity that—through responsiveness to the market—integrates the self, the entrepreneur, with society.”8 Who doesn’t want to self-actualize and integrate their self with society? I mean, poetry it’s  not, but underneath the weird phrasing these are actually excellent goals: doing what you love and having an interesting time while you do.

But the second kind of startup is responsible for most new jobs, brings fundamentally new products to market, and improves the overall quality of life. These are the sorts of goals that policymakers aim for when they talk about startups.

There is no accepted nomenclature that differentiates the two kinds of startups so people conflate the two. This leads to all sorts of confusion among governments, academics, financiers, journalists, etc.9 When newspaper articles talk about the decline in entrepreneurship, do they mean the first kind of startup or the second? Or both? They don’t know because they don’t even know there’s a real, fundamental distinction. This lack of understanding leads to bad conclusions.

If we could get people to understand that the kind of entrepreneurship that lets people work for themselves is primarily valuable to the individual while the kind that creates new products, lowers costs, and creates new jobs is valuable to society as a whole, we could direct attention and resources better.

People will always want to work for themselves, we don’t need to encourage them, we just need to let them. Less regulation, universal healthcare, a better small business loan infrastructure…these would all increase the number of this type of startup.

If we want more world-leading companies we need more funding for basic research, easier and cheaper access to higher education (and not just STEM) so people aren’t as burdened with student debt, and a better understanding of what makes these companies succeed. Different kinds of startups, different policies.

None of this is new–Schumpeter wrote the first edition of The Theory of Economic Development in 1911 (though my copy is the 1934 edition). So I’m not really expecting people to suddenly understand this difference. But at least now you do. It’s worth thinking about when you’re procrastinating on a Friday.

  1. McCraw, Thomas. (2007). Prophet of Innovation, p. 4. 

  2. As you do, if you have spent any time outside the ivory tower. Schumpeter did, and not just chasing women and horses: he was Austria’s finance minister and later a banker, where he reportedly made and then lost a fortune. Nothing convinces you change is real like losing all your money, in my experience. 

  3. You can read the chapter on “Entrepreneurial Profit” at Google Books, here

  4. Shane, Scott. (2008). The Illusions of Entrepreneurship

  5. Shane; 2008, pp. 36, 41, 65-67. 

  6. Shane; 2008, p. 43 

  7. I would argue in this context that a brand is a product differentiator protected by intellectual property laws. 

  8. Kauffman Foundation. (2008). Entrepreneurship in American Higher Education

  9. Here’s a whole paper on this: https://www.kauffman.org/what-we-do/research/2013/05/a-tale-of-two-entrepreneurs-understanding-differences-in-the-types-of-entrepreneurship-in-the-economy 

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 

Sorry this is so short. It’s an interesting topic that I don’t have time right now to do justice.\[\]

Last week I updated my “am I going to run out of money before I die” spreadsheet, as I’ve been doing every January for ten years. I need to do this because, aside from the paltry salary paid me as an adjunct professor, all of my income for quite a long time now has been capital gains from my venture portfolio. And I have no idea if any of my current or future investments will pay out a single slim dime. I believe they will, but belief is a thin shield between me and living in a cardboard box when I’m 90. There’s zero predictability, zero ability to plan.

The VC business model sucks (especially when you don’t get management fees, like me, because you’re investing your own money). Not just for founders, but for VCs. Why not do something different? Why not take less risk and accept less reward, as finance theory has it? Why swing for the fences and usually strike out, why not just hit singles or doubles consistently? This is a great question, and I’ve been thinking about it for 22 years.

It seems straightforward: invest in companies that won’t grow as fast or get as big, but fail less frequently. Even if it’s hard to sell/exit a small company with moderate growth, you could just share in the earnings or revenue. The present value of your payments should reflect the value of your share of the business and the expected value could be similar.

But is it so easy? There seem to be only two ways that small businesses have ever been funded by investors: bank loans and venture capital (I consider people like JP Morgan funding people like Edison venture capital, though it wasn’t called that then). If there is some third way that is better, why hasn’t anyone been doing it? That’s not a rhetorical question (though I’m not accepting answers along the lines of “this time it’s different.”)

My hypothesis is that there is no third way. And my model is this:

Assume that startups, all sorts of startups not just the VC kind, end up with values chosen from a power law distribution (a “PLD”; if you really want to follow this, you can read my previous post on power laws here). A PLD is parameterized by its alpha: the lower the alpha, the fatter the tail. Different startups draw their value from PLDs with alphas that reflect their riskiness. Riskier startups have lower alpha, less risky have higher alpha.1

What we want to know: if the distribution of startups you are investing in are characterized by a PLD with a specific alpha, does investing in those startups have a high enough risk-adjusted expected return to be worth investing in?

1. Modern Portfolio Theory

We’ll call the expected return of an investment \(ER_i\) and to risk adjust it we’ll divide this by \(\beta\), the riskiness of the investment. Beta is defined as \(\beta = \rho_{i,m}\frac{\sigma_i}{\sigma_m}\), where \(\rho_{i,m}\) is the correlation in returns between the investment and the market, \({\sigma}\) is the standard deviation in returns, \({\sigma_i}\) for the investment, \({\sigma_m}\) for the market. The important point here is that \(\beta \propto \sigma_i\).

For an investment to make sense, its risk-adjusted expected return must be at least the expected return of the market as a whole: \(\frac{ER_i}{\beta} \geq ER_m\).2

2. Power Law Distribution Returns and Risk

The expected return of an investment drawn from a PLD is the mean of the PLD, \(ER_i = \frac{\alpha – 1}{\alpha -2}x_{min}\), which is infinite when \(\alpha < 2\).

The variance is \(\frac{\alpha – 1}{\alpha -3}x_{min}^2\), which is infinite when \(\alpha < 3\). The standard deviation is the square root of the variance, so \(\beta \propto \sqrt{\frac{\alpha – 1}{\alpha -3}x_{min}^2}\).

3. Risk-adjusted expected return at various alphas

a. When alpha is greater than 3, the risk-adjusted expected return is

\(\frac{ER_i}{\beta} = \left. \frac{\alpha – 1}{\alpha -2}x_{min} \middle/  \sqrt{\frac{\alpha – 1}{\alpha -3}x_{min}^2} \right.= \frac{\sqrt{(\alpha -1)(\alpha – 3)}}{\alpha -2 }\)

If the market expected return is around 8%, then \(\frac{ER_i}{\beta} > ER_m\) when alpha is slightly above 3. These are good, low-risk investments, the kind that banks traditionally made to small businesses.

b. When alpha goes below 3, the standard deviation becomes infinite, so the beta does too. Now \(\frac{ER_i}{\beta} = 0 \). This is less than \(ER_m\), so these are bad investments.

c. When alpha goes below 2, the expected return also becomes infinite, so \(\frac{ER_i}{\beta} = \frac{\infty}{\infty}\). This is an undefined quantity and can’t be compared apples to apples to \(ER_m\). Are these good investments? What we know is that early stage venture funds seem to have alphas of about 1.95, just below 2 (see the post I referenced above). This is the area where venture capitalists operate.

The gap between alpha > ~3 investing and alpha < 2 investing means that using VC style investing with slightly lower-risk companies won’t work, just as using small-business lending investing but with slightly higher-risk companies won’t work. When alpha is between 2 and ~3, the risk adjusted expected return does not warrant investing at all.

Why don’t VCs take a little less risk and settle for a little less return? Because they’ve already crowded up against the risk limit: if they increase their alphas by even a little bit they end up in no-go territory. Their only option is to skip all the way to alphas > 3, the much lower risk, much lower return sector.

This is a theory and a model. There are a lot of assumptions. What should you take away from this? That there may, in fact, not be a spectrum of riskiness that investors can profitably invest in. That VCs may only invest in very high-growth-potential companies in very large markets not because they are bloody-minded but because that is the best they can do given the constraints of PLDs. That the allure of medium-risk investing is perhaps an illusion.

  1. I’m not going to go into why I think that if VC-backed startup values follow a PLD, then all startups must, but if you disagree then you have to posit some bright line qualitative difference between the two, and I don’t think you can. 

  2. I’ve assumed the risk free rate is zero. 

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 

Ralph Waldo Emerson reportedly once said “If a man can…make a better mouse trap than his neighbors, though he builds his house in the woods, the world will make a beaten path to his door.”1 Every engineer fervently wishes this were true. It is not.

The success of any interesting product is not just a function of its efficacy, but of a complex social and economic process generated by the interaction of the technology with the market and the sociopolitical establishment. I’m going to break this down using the MP3 player market as an example. You should read my last post, iTunes and the Basis of Competition in the MP3 Player Market, first. If you already have, here is a timeline to refresh your memory.
This post is about innovation from a technology point of view. It talks very little about the human side of it or company strategy, those are the next two posts.

Technological Innovation

We tell stories about technological innovations like this:

  • a genius scientist discovers some fundamental insight into how the world works,
  • one or a small group of against-all-odds inventors use that newly discovered phenomenon to create a new technology,
  • a forward-looking company turns that new technology into a product and commercializes it.

These stories make the best narratives, and the narrative I have chosen is no different.

The fundamental insight was perceptual coding, the discovery that the human auditory system does not “hear” every sound and so some information can be removed from an audio stream without changing what the listener hears. The invention was turning this into a workable audio compression algorithm. The product was embedding this algorithm into a portable media player2, which was then commercialized. Along the way more mundane innovations changed the constituent technologies to make them more useful.

Some thoughts:

This is easier said than done. The process of turning a new scientific insight into a technology—something that can be used—is not always straightforward. It took Karlheinz Brandenburg’s team and its predecessors years to turn the insight of perceptual coding into a usable algorithm. It reportedly took DuPont decades after the discovery of polytetrafluoroethylene to figure out how to get it to stick to frying pans and call it Teflon. I was taught in college that gallium arsenide was to imminently supplant silicon in integrated circuits: this has still not happened a few decades later, mainly because no one has figured out how to do it economically. One of the core uncertainties in technological development is whether you can and how long it will take to turn a discovered phenomenon into a usable technology and then into a commercializable product. Often this is simply unknowable.

Invention is not always science-driven. There is an idea (especially prevalent among scientists) that technology is nothing but applied science. Sometimes it is, and sometimes it’s not. Sometimes phenomena are analyzed by scientists and that analysis allows engineers or others to create technologies. But sometimes the opposite happens: “Through most of history science has arisen from problems posed for intellectual solution by the technician’s more intimate experience of the behavior of matter and mechanisms.”3 Thomas Newcomen, the inventor of the first steam engine, was not formally educated and developed his engine through trial and error. Edison knew almost nothing of electromagnetic theory. The Wright brothers were not scientists. In these cases the explanations for the phenomena followed the development of the technologies rather than preceding them.4

Of course science plays a part. Scientists experimented with electricity before Edison. Scientists had discovered the vacuum before Newcomen. But phenomena are discovered through experimentation, and the inventor’s trial-and-error can be as powerful a process of discovery as the scientist’s formal experimentation. Sometimes a phenomena is discovered, then explained by scientists, and then harnessed in an invention. Sometimes the invention comes first, with the scientific explanation following.

With MP3 there was a bit of both. There was preexisting science to inform the algorithm, but there was also a long period of practical experimentation–tinkering–to make it work. Brandenburg listened to Tom’s Diner hundreds of times, testing tweak after tweak.

Innovation is more than invention. Invention is a matter of degrees. We can say that Brandenburg (with his colleagues) invented the MP3 algorithm. But did SaeHan “invent” the MP3 decoder chip? Was it an “invention”? We usually use the word invention to mean the creation of a technology, not the implementation of one. But technologies seem less invention-y to us the further they are removed from the fundamental phenomenon they instantiate: the MP3 algorithm was an invention, instantiating perceptual coding. SaeHan’s MP3 chip instantiated the MP3 algorithm, but the algorithm wasn’t a fundamental phenomenon so it seems a little less like an invention, and so on. Who deserves credit for “inventing” the MP3 PMP? Who invented the automobile, the personal computer, the pizza?

While we should care about discovery and invention because they are necessary precursors, real impact comes from innovation, not invention. An innovation is an invention that is useful and is put to use. Inventors and discovers get most of the glory, but innovators cause change and make money. Brandenburg is widely credited with the invention of MP3, Fraunhofer made millions from their patents, but it was Apple–who invented neither the algorithm, the device, nor anything else–that changed the industry and made billions.

How much is an idea worth? I do not talk much about intellectual property in my class. Patents and such-like are important to startups because they defend the company from others having interfering patents in the future. But predicating a strategic advantage on being able to defend a patent works for only a specific type of company. Patents can cost millions of dollars to defend so they are only a viable way to prevent competition if you can spend those millions of dollars once your patents are infringed. If you need time to build your market and company value after your product is released, better resourced companies can copy your idea or work around your patent while you are still too small to defend yourself.

Patents work when, by the time you release your product your company is valuable enough to spend the millions to defend them. For instance, if you develop a groundbreaking pharmaceutical launching into a new market, your company becomes very valuable simply because you have released the drug; you don’t have to wait until a market is created for it, the market preexisted your launch. On the other hand, Fraunhofer could not prevent others from stealing its intellectual property because there was not enough money at stake when it was stolen for it to be worth them suing everyone. (There was enough money later on, when the market was established, and there were patent suits aplenty. But by that time Fraunhofer had lost control of the market.)

Patents are valuable. But they will not necessarily win you the market.

What is a technology? The inventions we have been talking about are all technologies. Technologies do something useful. They are discovered or invented and then put to use in a process of innovation. Later they might spread through the economy in a process of diffusion.

Technologies are not always “things”. A technology does not have to be a physical object. The MP3 algorithm is a technology, even though it is intangible. The idea of a limited liability corporation is a technology, even more intangible.

Some new technologies are the result of turning a fundamental phenomenon into something useful, but most technologies are created by combining already existing technologies. The MP3 chip was the combination of the MP3 compression technology with integrated circuit technology. The MP3 PMP was the combination of the MP3 chip with digital storage and display technologies. The advent of a new technology makes many other new technologies possible because it is a new thing to combine. Some new technologies are so useful and can be combined in so many ways with already existing technologies that they create a wave of innovation. The internet is an example.

Some innovative technologies don’t depend on anything new at all: Apple’s entry into the PMP market contained very little new technology, and the new tech it had was not the primary cause of the iPod’s success. What Apple did was connect the component technologies in a new way, a way that was much more useful to its customers. This type of innovation is called architectural innovation.

This process of combining technologies is accelerated through modularity. People are ingenious at coming up with combinations of technologies to solve problems. Sometimes this is the result of deep understanding of the component technologies and what they will do when combined, and sometimes it is the result of making do with what the engineer has at hand, a type of “bricolage.” The usual difficulty in either case is getting the different component technologies to work together well. Much of the “tinkering” that goes on with new technologies is just this: the gradual improvement of components and linkages to make the combination of technologies function better.

One common way to make this easier is to modularize components. By having a clear delineation of what a component technology’s function is and by laying out how it works with other technologies—an interface—components can more easily be combined with other technologies. When SaeHan built the MP3 chip, they included a description of how it could be combined with other circuitry: SaeHan intended it to be used by others and hoped for it to be adopted widely. Technologies that are not meant to be used outside of the ambit of their manufacturer often are not modularized in this way, and this can slow their diffusion if unanticipated uses are later found.

Modularization is powerful because the interface is an abstraction of what the component does: the user does not need to take the messy details of the component’s internal functioning into account. This allows the modularized component to be improved or replaced entirely without disturbing the linked technologies, so long as the interface remains substantially the same (or has backwards compatibility).

The standards process is a way to encourage modularity, by guaranteeing functionality and an interface. The MPEG standard-setting process that resulted in the MP3 standard, among others, is an example. The standards group knew that by having a few, well-understood, standards, progress in technologies that used these standards would be faster. Fraunhofer worked hard to have its algorithm standardized: they knew having a standard blessed by a recognized standards group would reduce a couple of sources of uncertainty for adopters: it was a type of legitimation, a mark of approval that the technology worked as described; it was a guarantee of modularity, that the interface was stable and would not be implemented in incompatible ways between different products, allowing interoperability. More about legitimation in the next post.

Modularity encourages experimentation. The MP3 PMP itself was several technologies connected together. The MP3 PMP as technology consists of these component technologies as well as the architecture of their connection, forming a system.

First Generation iPod Component Technologies

Because the components are linked, the form and function of each component depends on the components it is linked to. Innovators improve a complex system either by adding new technologies to it (creating a new technology), improving the technologies and linkages that make up the system (sustaining or incremental innovation), or by changing the linkages between technologies (architectural innovation).

Modularization of the components encourages experimentation: new components can be added more easily or components can easily be swapped out for components that perform the same function in different ways, especially if the substitute has the same or similar interface.  When Apple developed the iPod, they worked from a reference design they bought from another company. But their substitution of several new components—the touch wheel, the 2.5” disk drive, etc.—while still using the same reference design substantially improved the resulting MP3 PMP while avoiding the expense of re-architecting the entire thing.

Experimentation is a primary mode of technological improvement. When there is uncertainty about what exactly the market wants from a technology, or about how changes to a technology will affect its overall functioning (as often is the case when the technology is ahead of the science) then experimentation is the best way to improve the technology. This kind of experimentation often takes the form of tinkering by people with deep tacit knowledge of the tech gleaned through long periods of learning by doing.

Most technological change is improvement. We notice and talk about innovations that make a big difference, that are either a noticeable step-change in the functionality of a technology or are a technology that does something fundamentally new and interesting: radical innovations. But usually innovations are of a quieter sort, they take an existing technology and improve it. While each of these incremental innovations only adds a bit, they are so much more common that they can create much of the eventual value from a technology.

Data from Anderson, P, and ML Tushman. “Technological Discontinuities and Dominant Designs: A Cyclical Model of Technological Change.” Administrative Science Quarterly 35.4 (1990): 604–633.

Compare the first MP3 PMPs and a late-stage iPod. They are fundamentally the same technology and serve the same use, but the PMP–in itself, as a standalone portable media player–was incrementally improved over time to become more and more valuable.

Incremental change mostly occurs along a very few metrics. When I say the iPod was an improvement I mean that, given the choice, people would prefer it to the Eiger MPMan. You could even suppose that if both were available at the same time that people would pay more for the iPod, and this would be one way to measure how improved it was. (This is difficult to do in the real world because technologies often become cheaper as they become more valuable. This is not because price and value are disconnected, but because several types of incremental innovation are usually ongoing at the same time resulting in simultaneous improvement in functionality and decline in manufacturing cost.)

To improve performance, innovators must first figure out what “higher performance” means. A higher performance disk drive is one that holds more data or reads/writes data faster, all else being equal. The performance metric always seems obvious in retrospect, but finding the answer to the question “What is Quality?” is one of the few true defining moments in a technology’s life-cycle.

competitive basis is more than just a quality metric, it is a metric that becomes a primary differentiator between competing products in that market. You might prefer a black disk drive to a gray disk drive, all else being equal, but making black disk drives is not a basis for competition. It is not that important compared to other attributes of the disk drive and it is not how products can meaningfully differentiate themselves from the competition. In some cases there might be many metrics customers care about. If you are buying a car you might compare fuel efficiency, or comfort, or acceleration. If each of these metrics appeals primarily to a different customer group, the market becomes several markets. If you care most about acceleration you don’t shop for a car, you shop for a sports car. Others shop for a fuel efficient car or a luxury car.

The MP3 algorithm itself was developed with two quality metrics in mind: the heard fidelity of music, and the computational power needed to encode/decode. The MPEG authorized several standard digital music algorithms at the same time (there was a MP1, MP2, etc.) and MP3 was superior to them in terms of lower computational power, but inferior in sound quality. This was intentional: it was meant for bandwidth-constrained channels. This compound metric–better in bandwidth-constrained channels–was, at the time of introduction, a solution in search of a problem. It wasn’t until the consumer internet started to grow that this metric came to the fore as the competitive basis. MP3 outcompeted MP2 in the internet sharing of music market because it was better suited. It also outcompeted because it was made free. While the free software was illegal, it created enough users and MP3-encoded music that it became a de facto part of the interface that linked the various pieces of the digital music ecosystem together.

Though this quality metric guided the design of MP3, it was not one that was especially compelling while the algorithm was being developed. Fraunhofer worked for quality along this metric because that was the unique strength of perceptual coding, the technology that Brandenburg happened to know. Researchers often create the technology they know how to create without regard to what the market wants. If they are lucky, as Fraunhofer was, the market appears, but this is in no way a given.

In other cases, innovators have to search for the right metric to compete on. They can do this by talking to customers and trying to determine which problem they are solving, but customers often do not know. So the search is very often done through experimentation: many product designs are tried until one is found that resonates with customers.

At the beginning of the MP3 PMP market companies competed on many quality metrics, trying many variations of the product. Some players held more songs, some were smaller, some had better user interfaces, some were better for exercising with, some integrated with the user’s PC better, etc. The profusion of companies and PMPs between 1998 and 2003 was a result of this search for the basis of competition.

It is hard to build a market with many different bases of competition. It is confusing to the customer and the market fragmentation means the resources available for incremental innovation is spread across many technologies so each improves less. But what tends to happen, especially when the technology in question is a compound technology—one that is comprised of many components in a system—is exactly what happened in the MP3 PMP market: there is a profusion of varieties of the technology, each emphasizing a different combination of strengths. This profusion of products is usually accompanied by a profusion of companies. At some point a product hits upon the..

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 

CC-BY. PDF here.

iTunes and the Basis of Competition in the MP3 Player Market

“It’s official: the only thing more popular than MP3 is sex.”

So said Rolling Stone magazine in 19991:

Virtually unheard of a year ago, MP3—short for MPEG 1 Layer 3—is an audio-coding technology that allows digital music files from CDs or other sources to be compressed into a size practical for Internet transmission and PC storage. Whereas a typical digital recording of a song might take up 40 megabytes of space, an MP3 version will take up roughly 3.5 MB. Those same files can be downloaded in ten minutes instead of two hours (and even faster with the T-1 connections many universities and businesses have). As a result, music is a major Internet attraction.

Karlzheinz Brandenburg had begun work on an audio compression systems years before, while a doctoral student in the early 1980s. The work moved slowly until 1986, when computers began to catch up to the algorithmic complexities of audio compression.2

To make an MP3, a program called an encoder takes a .wav file (or some other audio format) and compares it to a mathematical model of the gaps in human hearing. Based on a number of factors—some chosen by the user, some set in the code—it discards the parts of the audio signal that are unlikely to be audible. It then reorganizes repetitive and redundant data in the recording, and produces a much smaller file—often as small as 12 percent of the original file size. The technique of removing redundant data in a file is called compression. The technique of using a model of a listener to remove additional data is a special kind of “lossy” compression called perceptual coding.3

Brandenburg and others worked to make the compressed and then decompressed audio relatively faithful to the original. The researchers used Suzanne Vega’s a capella version of “Tom’s Diner” to listen for imperfections. Brandenburg claims he listened to the song between 500 and 1000 times during the development of the MP3 algorithm.

The Moving Pictures Expert Group (MPEG) began a standard-setting process for digital compression of music in 1988 under the auspices of the International Standards Organization. MPEG’s purpose was to develop standards to digitally encode moving pictures, including their soundtrack. The first meeting of MPEG’s audio subgroup ended with a call for digital audio compression proposals that would work for “various applications like CD-ROM for audio and video, DAT recorder, [and] Digital Audio Broadcasting.”4 In 1989, the ISO received proposals for audio coding from 14 companies (later merged into four groups). The proposal including Fraunhofer’s work was adopted as a standard in 1992.5

This standard, named MPEG-Layer 3, involved technology developed by Fraunhofer, Thomson, and AT&T. At the same time, MPEG-Layer 1 and MPEG-Layer 2 were approved. Each involved different tradeoffs: Layer 3 had the highest quality audio and smallest file-sizes. Layer 1 had lower quality audio but was a less complex algorithm. Layer 2 was in the middle.

Layer 2 received most of the initial attention: it was chosen as the audio standard for video compact discs and satellite radio, among others. Even the first internet site to offer MPEG-encoded files, the Internet Underground Music Archive, launched in 1993, initially used MPEG-Layer 2. Brandenburg said, “In 1992–1994, the main focus was to find companies who would really use this MPEG audio layer 3 and with the exception of some professional applications in the first year, layer 3 was out of luck. Everybody else decided to go with layer 2.”6 The Fraunhofer team began to believe that Layer 2 was winning because it was backed by established music industry companies Philips and Panasonic, that it was a political rather than a technical decision. There were a few early adopters: Microsoft bought a license in 1995 to use the spec, although the Microsoft applications at the time were not much used. Fraunhofer began to lose hope.

Creating a Category

Using MP3 for portable music was not an obvious use case: before 1995 the internet was not in widespread use, and where it was bandwidth was low. In the physical world, where digital music was delivered through the medium of a compact disc, compression was not needed. But Fraunhofer was willing to try anything, and portable music was one thing they tried.

In 1995 Fraunhofer created software to play MP3s on a PC. It was called WinPlay3 and was distributed as freeware. It was ugly, hard to use, had no playlists or other simple features, and could only play 20 songs before the user had to mail in a license fee to Fraunhofer and wait for an activation code to be mailed back. It was little used. (It did, however, need a file extension, and .mp3 was settled on, the first use of that name.) At a 1995 trade conference a Philips executive said to a Fraunhofer representative “There will never be a commercial MP3 player.”7

The problem was that no one wanted a player if there was no music, and no one wanted to make music in the format if there was no way to play it.8 It seemed the only way to get a critical mass of music in MP3 was to have the music industry embrace the standard. Without the cooperation of the music industry there seemed no way out of this chicken and egg problem, and the music industry—fat on CD reissue revenues and ever-wary of the potential for piracy—had no reason to cooperate.

But the growing consumer use of the Internet created a demand for music files delivered across a slow network connection, and MP3 was far better for this than Layer 2. When the corporate world did not address this problem an anonymous hacker took it into his own hands. As Brandenburg tells it:

In, I think it was ’97, some Australian student bought professional grade—from our point of view—encoding software for MP3 from a small company in Germany. He paid with a stolen credit card number from Taiwan. He looked at the software, found that we had used some Microsoft internal application programming interface … racked everything up into an archive and wired some Swedish side [sic], [and] put that to a U.S. university FTP site together with a read-me file saying, ‘This is freeware thanks to Fraunhofer.’

He gave away our business model. We were completely not amused. We tried to hunt him down. We told everybody, ‘This is stolen software so don’t distribute it,’ but still the business model to have expensive encoders and cheap decoders [was] done. From that time on, we reduced the cost for encoders. There was a company, Music Match, which allowed people early on to take a CD’s music, read it into the computer and then have their own music jukebox on that. And they were legal, they paid for the patent fees so that was fine.9,10

Fraunhofer had planned to license its patents to create encoding and decoding software. A license for the encoding software would be expensive and Fraunhofer assumed only large companies would license it. A license for the decoding software would be cheap and Fraunhofer hoped many products would be built around it and they would collect a small fee for each product sold. This vision—a few companies making recordings available to a large audience—had been the recording industry model since its inception, the typical mass-media model. The widespread illegal distribution of their algorithm made this business model moot, but the result was that the entrenched music industry players were no longer the gatekeepers to widespread use.

The explosive growth in use of MP3 started right after the Australian release. WinAmp, the first PC software music player for MP3s since WinPlay3, was released shortly afterward. It was downloaded more than 3 million times in its first twelve months.11 MP3 had become synonymous with digital music, as well as with music piracy. By 1998 the category was established, with one mainstream music magazine writing “Making MP3 copies of favorite songs—then zapping those songs electronically to anyone with an Internet connection—has become not just an underground craze but an international epidemic.”12

Creating a Market

In 1997 Micronas, a German chipmaker, released an MP3 decoder chip.13 A Korean company, SaeHan Information Systems (now TAK Information Systems) started buying the chips to build the first portable MP3 player, the SaeHan MPMan. The MPMan was introduced in Asia in early 1998. It cost 39,800 Yen (about $400 at the time) for the 32MB version, which could hold about six songs, and 59,800 Yen (~$600) for the 64MB version.14 The design was licensed to Eiger in mid-1998, who distributed it in North America, selling the 32MB version for $250. The Diamond Rio, also with 32MB, quickly followed in September 1998, selling for $20015, and then a slew of others.

The first MP3 players used flash memory. This made the players portable but limited their capacity. Some later players used hard drives to hold the music. These had much higher capacity, but were larger, heavier, and more expensive. The HanGo Personal Jukebox, for instance, cost $799 when it was introduced but had almost 5GB of storage. At 6x3x1 inches, it was unwieldy for a handheld. Solutions that were both high-capacity and small, like the I2Go eGo, cost about $2000. Other innovations included expandable memory, longer battery life, ability to record voice (all three of these were included in 1999’s RaveMP), radio reception (1999’s Nomad), better sounds (2000’s Nomad), better user interface (Apple’s iPod), video support (2002’s Archos Jukebox Multimedia), etc.

The MP3 standard was not the technically best compression algorithm. The AAC format—created by Fraunhofer and introduced as an MPEG standard in 1997—was better (and later used by Apple), as was the open-source Ogg Vorbis format, released in 2000 and free to use (it is now used by Spotify, among others). Microsoft introduced their own format in 1999 (WMA), perhaps to avoid paying Fraunhofer licensing revenues. But players that bucked the increasingly popular MP3 standard had a hard time gaining any traction. Sony joined the digital audio player market in 1999 with their Vaio MP-P10 Music Clip. Despite its innovative design and small size, it did not natively support MP3 (it used Sony’s proprietary ATRAC3 format and included MP3 to ATRAC3 conversion software) and was a failure. Sony tried again in 2003 with the NW-MS70D, but it also did not natively support MP3 and failed. It was not until 2005’s Sony NW-HD5 that Sony released a player that supported MP3.16 The same problem that had dogged MP3 in the beginning—no one wanted to introduce a player that had no music and no one wanted to encode music in a format that no one could play—now worked in MP3’s favor: if all of your music was already in the MP3 format, why would you buy a player that used a different one?

Table 1: Some Digital Audio Players Supporting the MP3 Format


Model Date Released Capacity
SaeHan MPMan Spring 1998 32 MB
Eiger MPMan May 1998 32 MB
Diamond Multimedia Rio September 1998 32 MB
Sensory Science RaveMP 2100 Mid 1999 64 MB
Creative Labs NOMAD June 1999 32 MB
HanGo/Compaq Personal Jukebox 1999 4.8 GB
I2Go eGo 2000 2 GB
Creative NOMAD Jukebox 2000 6 GB
Cowon iAudio CW100 October 2000
Archos Jukebox 6000 December 2000 6 GB
Intel Pocket Concert 2001 128 MB
Bang & Olufsen BeoSound2 2001 128 MB
Apple iPod October 2001 5 or 10 GB
Archos Jukebox Multimedia 2001 10 or 20 GB
Creative Labs Muvo 2002 64 or 128 MB
Apple iPod 2nd Generation 2002 10 or 20 GB
Creative Labs Nomad Jukebox Zen 2002 20 GB
Creative Labs Nomad Jukebox Zen NX/Xtra 2003 60 GB
Creative Labs MuVo NX 2003 128 or 256 MB
Apple iPod 3rd Generation 2003 40 GB
Diamond Rio Karma 2003 20 GB
Microsoft Zen Portable Media Center 2004
Apple iPod Mini 2004 4 GB

The Recording Industry Backdrop
“Home Taping is killing music.”

– Logo of the British Phonographic Industry’s 1981 anti-piracy campaign

Home taping did not end up killing music, but the recording industry has never not been paranoid about piracy, the illegal copying of music for distribution. Music piracy was not invented by MP3 aficionados, but the combination of perfect digital copies and worldwide, instant internet distribution made the problem several times more acute.

Cassette tapes meant for audio storage were introduced in the 1960s. Early sound quality was poor, but advances in technology meant that by the early 1970s, cassette tapes rivaled eight-track tapes in sound quality. Not coincidentally, the International Federation of the Phonographic Industry pushed through the first anti-piracy measure, the Geneva Phonograms Convention. During the recession of the late ’70s and early ’80s, the IFPI tied the decline in record industry revenues to the increased sale of cassette tapes and recorders.

But the possible impact of tape recording on record sales was fundamentally limited by the technology: a tape had lower quality than its vinyl source, a tape of a tape lower still. The distance between the listener and the original purchaser could not be high. Regardless, the IFPI estimated piracy at 11% of the total market in the US and Canada in 1982 and higher elsewhere. So when consumer digital audio recorders started to become available in 1987 with the release of Sony’s Digital Audio Tape (DAT) standard, creating the potential for flawless duplication, the recording industry pushed back: the Recording Industry Association of America (RIAA) began lobbying for legal protection. The result, in the US, was the Audio Home Recording Act of 1992 (AHRA) that, among other things, mandated a Serial Copy Management System to be built into all digital audio recorders. This SCMS chip prevented second-generation recordings. AHRA also levied fees on DAT recorders and blank media to be distributed to the existing recording industry. This crippled the market for DAT and it never caught on among consumers outside of Japan. The personal computer industry successfully lobbied to have personal computers exempted from AHRA, not a huge concession at the time given how small the PC market was.

Five years later, the cheap availability of both encoding and decoding software and the exemption of PCs from AHRA laid the groundwork for the growth of MP3s on PCs. Early digital audio enthusiasts were limited because without widespread access to CD burners and with no MP3 players yet on the market, songs had to be listened to on a desktop computer, they could not be played in the car or while running, or anywhere else that people listened to music. And the rarity of high-bandwidth consumer connections to the internet in the 1990s meant that even a 3 MB MP3 file was difficult and time-consuming to share. CDs were still a better way to consume music.

But when the Diamond Rio MP3 player came on the market in the Fall of 1998, this all changed. The RIAA decided to nip the phenomenon in the bud. The Rio had no SCMS chip, and the RIAA contended that it thus violated the Audio Home Recording Act. But by June of 1999 the courts ruled that since the Rio was only a storage device, not a recording device, it did not fall under the Act. Between the exemption for PCs and the failure of the RIAA to prevail in the Diamond lawsuit, the MP3 player market was entirely legal.17 In fact, the publicity generated by the lawsuit seemed to spur MP3 player sales.

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 
Reaction Wheel by Jerry Neumann - 9M ago

I saw Bill Janeway speak at Columbia the other night. One of the things he said was “pessimists don’t make good VCs.” A truism.

This Summer, talking to a younger colleague about the pace and direction of technological innovation I said there was less potentially society-changing technology today than at any other point in the last fifty years. And that this drought was likely to last a while. That we are stuck in an era of small changes, of gradual improvement, of fine-tuning processes, of reacting to change, of head-to-head competition. He replied, tartly, “maybe you’re in the wrong business.”

This week, I was having coffee with a European investor I see once a year or so. He asked what I was looking for. I don’t know, I replied. “Haven’t you been investing?” he asked.

I’ve made five new investments in the last twelve months. I am more excited by these companies than by almost any other company in my career. They have the potential to radically change their industries. NeMedIO could dramatically reduce the upfront cost of developing new medical robots and devices, causing a proliferation of high-quality new tools for doctors. Sila could provide the underlying infrastructure that unbundles the banking industry. Edmit‘s higher-ed pricing transparency could cause the industry and its customers to refocus on education.

And these companies are using new technologies. Sila is using the blockchain. Unsupervised is using topological data analysis, a new form of unsupervised machine learning, Lately is using AI. These are valuable and groundbreaking innovations.

But they are not the same as the new technologies of twenty years ago or thirty years ago or forty years ago. In the tree of information and communications technology where semiconductor logic and TCP/IP are near the root, these technologies are closer to the leaf. They are not less technically valuable, but their value lies in their application itself, not in their ability to enable further technologies. The market application is to the end-user, the person solving existing business problems; not as components of further technologies. This means that the commercialization focus has to move from the technology itself to the user-problem the technology solves. The market for each company is thus relatively smaller and not subject to the same kind of decade or longer evolution that provides the runway for scores of startups.

Blockchain, for instance, is a sweet technology. As an engineer I consider it beautiful. But in Sila I didn’t invest in a blockchain company, I invested in a company using the blockchain as a chisel to cut the current banking oligopoly into hundreds of innovative companies.

This may seem like semantics, but it’s not. I no longer believe in investing in technologies. I don’t think it’s any longer sustainable to be a blockchain investor, or an AI investor, or a ____ investor. If you ask me about exciting new technology trends I will sound like a pessimist. But I am optimistic about companies. Even while saying there is nothing interesting I run across insanely interesting companies every month.

This makes it hard to do my job. It would be easier to say I invest in autonomous drones in the home automation space. You would then know exactly which companies I would be interested in seeing. I would also see all the companies competing in that space, so I would be better at picking. This is why VCs choose specific theses. If I had a thesis I regarded as a viable strategy, I would be happy. Instead I have arrived, finally, where Michael Moritz did long ago: “The business is like bird spotting. I don’t try to pick out the flock. Each one is different and I try to find an interestingly complected bird…”1

The project now is to operationalize this as a strategy. How do you pick companies that fit no previous pattern?

  1. https://web.archive.org/web/20070715123315/http://vcratings.thedealblogs.com:80/2007/07/mike_moritz_of_sequoia_capital.php 

Read Full Article
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 
Reaction Wheel by Jerry Neumann - 1y ago

Thank you to Tom Eisenmann of HBS, who gave valuable feedback on a couple of iterations of this.

I teach entrepreneurship at Columbia University. We devote the second-to-last class to a Harvard Business School case study about the early years of Zipcar. It’s a good case and we can review most of what we learned through the lens of the problems Robin Chase faced just before raising her second round of financing.

In the last class we talk about fundraising and venture capital. For continuity, I decided to use Zipcar as an example of the mechanics and norms of VC funding: pre- and post-money, preferred stock, rounds, dilution, etc. Since Zipcar had been a public company, I figured public documents would contain most of the numbers I would need to piece together the story.

This is that story, as far as I could really piece it together. It was stranger than I had assumed it would be, and that was good. My general philosophy about taking VC money is as Rilke’s to writing: if you don’t feel you absolutely must, then don’t. The story of Zipcar’s financing illustrates one of the many ways a company can be successful while its founders are not. It also shows some major mistakes by early investors.

This post isn’t a primer on venture capital, it’s a real-world example of how things can play out. I know that I’m throwing you in the deep end, but Fred Wilson and Brad Feld have explained the mechanics well: if you don’t recognize a term, visit one of their blogs and search for it.

Most of the information in this analysis is drawn from the 424(b)(4) form that Zipcar filed with the SEC in April 2011, soon after their IPO. The 424 is called a Prospectus. The SEC requires a form 424 after an IPO and it contains much of the business and financial information about a company that a public market investor needs to make a rational investment decision. The SEC only requires most of the information required for the 424 to reflect the three year prior to the filing, so many details about the early years of the company can only be inferred from agreements and ownership that survived until the three years covered in the filing. I have had to make some guesses to fill in the gaps.

One thing to note in this analysis is that I refer to the number of shares on an “as-converted” basis. This is important because the company did a 1:2 reverse split before the IPO–probably to make sure the trading price at IPO would be above $10/share. Because of this, all of the number of preferred shares in the 424 need to halved to know how many shares the holder will own after the IPO. That is, if a shareholder owned 1,000 Series B shares, those preferred shares will convert into 500 Common shares at the IPO. It is easier to compare the value of different types of shares over time using their Common equivalents, so that is what I did. The base data and calculations I made are in this Google Sheet, with references back to the Prospectus. I don’t think you all need to refer back to the 424 but included the references so if you want to have a go at reading the SEC document and tying it back to the actual business, you can. I have found the ability to parse SEC documents extremely useful in my finance career so if finance is the road you’re going down, reading through and understanding a prospectus is a good thing to be able to do. Just sayin’.

The second one thing to note is that some of the breadcrumbs around ownership in the early years come from interviews in the media with Chase or others. Chase, in particular, seems to be an unreliable narrator. For instance, in one interview she says “When I finished raising a $7 million round of financing in 2003”1, but the financials show that only $4,000,000 was raised in the round that closed that year. And that round was announced as a $2 million round by Crunchbase2. The information on Crunchbase seems to reflect the first close in a larger round and Chase may be confused about timing, with $4.7 million closed in December of 2002, $4 million closed in November of 2003, and a $2 million bridge note somewhere in that time-frame as well. The point is: what people say to the media may not be accurate, for whatever reason. I relied on SEC documents for the facts, and considered other information interesting but not definitive.


The company was formed in 2000 by Antje Danielson and Robin Chase. They split the equity 50/50. This analysis assumes they each received 570,000 shares of the new company.

This number, how many shares the founders started with, is one of the most speculative assumptions in the analysis. There is no way of really knowing from public information. In fact, I am pretty sure it isn’t exactly right: who starts a company and decides each founder gets 1,140,000 shares (pre-reverse split)? Why not exactly 1,000,000 each or 2,500,000 to split? That said, it’s my best guess. I arrived at it from three directions:

  1. The number of Common shares at IPO that weren’t otherwise accounted for was 1,081,610, and some of these seem to have been issued when the pre-Series A convertible note converted (the second most speculative assumption here);
  2. An article about the company, with an interview of Danielson, said “Danielson remained a Zipcar shareholder until Avis bought the company in early 2013. ‘I started off with 50 percent of the company,’ she says. But after multiple rounds of funding, she ended up with 1.3 percent.”3 To make the number work, I have to assume this is 1.3% of the company before the IPO and not counting dilution from unexercised options and warrants;
  3. A different article about the company said “People close to the company said that shortly after Zipcar was launched, Chase controlled more than half of the shares. An early round of capital raising brought her share down to around 20 percent. But the real share-killer was a round of financing the company did in 2002, when the capital markets were reeling after the dot-com bust and 9/11. That investment round brought her shares down to around under 10 percent. By the time the company went public in 2001, Chase owned around 3 percent of the company…”4 Chase could not have owned 3% of the company just before the IPO: there just weren’t enough Common shares outstanding. Perhaps she owned a huge chunk of the options, but that seems unlikely–both because of what happened and because she would have had to exercise them when she left the company, and there weren’t that many exercised options. But it seems likely that Chase did own more than Danielson; I am told by someone who has spoken to Chase that Danielson voluntarily walked away from some shares when she left the company. Maybe this is true.

There is also the possibility that Chase or Danielson or both sold shares back to the company at some point (probably not in a secondary because then they would still be outstanding Common shares). If this happened it was probably after the Series C because the company wouldn’t have wanted to part with the cash before then. There is no way to know from the information I have.

Assuming this, at founding the capitalization table–the enumeration of who owns what shares–would have looked like this (shares and dollars in the cap tables are in thousands):

Owner Date Common Stock Preferred Stock Total Stock % Investment  
Chase  1/2000 570 570 50.0%
Danielson  1/2000 570 570 50.0%
Total 1,140 0 1,140 $0

Series A

The Zipcar case says “[Chase and Danielson] had incorporated in January 2000 and raised their first $50,000 from one angel investor…By October, the fledgling company had 19 vehicles, nearly 250 members, and the founders had raised—and spent—an additional $325,000 to fund the early stages of operations…Beginning in early 2000, Chase had made a series of presentations to potential investors in which she sought $1 million in capital.”5

The Prospectus tells us how many shares of each series of preferred stock are outstanding as of the IPO and what their preference is. Since venture capital preferred stock always has a purchase price equal to the preference, from this we can tell for each series how much was invested and how many shares were bought. And indeed, it shows that in October 2000 the company raised $1,035,606 by selling shares of Series A stock with a preference of $3.80 per share (again, on an as-converted basis): 272,528 shares.

In an interview years later, Chase said “The decision to expand to other cities came after we closed $1.3 million in Series A financing.”6 Similarly, the MIT version of the Zipcar case study says “With just three weeks until the company closed on its first round of funding worth $1.3 million…”7 I’m going to assume that the difference between the $1.3 million reported as raised and the $1.0 million actually raised is the amount of convertible notes raised before the Series A. If we assume the HBS case misspoke when it said $50,000 plus an additional $325,000 and meant simply $325,000 total, then the $1.3 million makes sense. (If it were indeed $375,000 on top of the $1,035,606 then they would have almost certainly rounded to $1.4 million. In general, everyone involves likes reporting larger numbers).

But then this means the Convertible Notes converted into Common stock, not Series A Preferred. That isn’t unusual, although it’s a really bad idea if you’re the investor, as you’ll see in a minute. I assume the Notes converted at 85% of the Series A price (in my experience, 15% was a more usual discount in 2000, while 20% is more usual now). This resulted in an additional 100,619 Common shares.

I also put in an option pool equal to 20% of the post-round company. This is fairly usual, although the size of the pool can vary. But I was also trying to get the founders closer to the “around 20 percent” mentioned in the quote above. This doesn’t get them there, but I can’t square the 20% after the first round with the 10% after the second round anyway. I think the quote given to CNBC is probably inaccurate.

Cap Table After the Series A

Owner Date Common Stock Preferred Stock Total Stock % Investment Value
Chase 1/2000 570 570 30.1% $2,182
Danielson 1/2000 570 570 30.1% $2,182
Convertible notes ?/2000 101 101 5.3% $325 $382
Series A investors 10/2000 273 273 14.4% $1,036 $1,036
Options + pool 378 378 20.0%
Total 1,619 273 1,892 $1,361
Read Full Article

Read for later

Articles marked as Favorite are saved for later viewing.
  • Show original
  • .
  • Share
  • .
  • Favorite
  • .
  • Email
  • .
  • Add Tags 

Separate tags by commas
To access this feature, please upgrade your account.
Start your free month
Free Preview