Research on startup teams and startup success

At the ACAC lunch today, Kathy Eisenhardt summarized her decades of research on tech startups. For an academic conference (the Atlanta Competitive Advantage Conference), Prof. Eisenhardt literally needs no introduction: as host Bill Bogner of Georgia State said, “Kathy Eisenhardt needs no introduction: if she does, you didn't pass comps.” (Academics would know her paper that has 20,000+ cites, while tech entrepreneurs might know her as co-director of Stanford Technology Ventures Program or her many STVP videos).

Eisenhardt's focus was on the importance of a startup’s management (“Top Management Team” in strategy jargon) in determining the success of a small or young firm in a highly uncertain environment. She identified three factors that explained that success

1. Optimal management team
2. Optimal strategic decision process
3. Matching strategy and structure (at “the edge of chaos”)

1. Management Team

We know that successful teams need to be larger, diverse and have prior work experience together (and thus trust). However, there is an interaction effect between the team and the sort of opportunities they pursue. As scholars who study tech startups will tell you, firms tend to be veterans of an industry who start firms in that same industry that they know.

Her 1990 paper with Kaye Schoonhoven showed that the best firm growth came where a top team caught a great opportunity. A great opportunity was a market that’s at the takeoff phase of a growth market: at least $20 million of industry revenue and 20+% annual growth. In California-speak, Eisenhardt said this is a great surfer catching a great wave.

In specific domains, she cited the recent research of Anne Fuller and Frank Rothaermel on star faculty entrepreneurs as well as various papers by Sonali Shah on user entrepreneurs. As she noted, Chuck Eesley of Stanford (an MIT alum) who estimated when experience is more valuable than talent, based on a survey of entrepreneurs from among the 100,000+ MIT alumni.

2. Strategic Decision-Making

Her old studies on TMT decision making showed that for fast choices in highly uncertain environments, managers need more information and more alternatives, as well as a decision process that is midway from managerial fiat and (a hopeless search for) total consensus. She also noted later work of researchers who examined improvisation and bricolage.

Her former student, Sam Garg, has studied how CEOs manage their boards. The best CEOs constrain the interactions with the board and don’t give up their power over leading the company, using it to make decisions (not generate ideas), by using a divide and conquer strategy.

3. Strategy and Structure

Summarizing her research with Chris Bingham of UNC, she noted that young firms were most successful when they could use their experience to generate heuristics. Experience was valuable when it created “simple rules” that firms could apply over and over again: such rules were both quicker and often better in solving problems in conditions of high heterogeneity and high uncertainty.

Eisenhardt noted that firms (like parents “raising your teenager”) face a dilemma between too much and too little structure. In a simulation with Bingham and Jason Davis, they found that in a highly unpredictable or turbulent market, too little structure is more dangerous than too much. (This also sounds like raising a teenager).

Finally, in cases of high ambiguity (e.g. nascent markets), success is more determined by luck than skill. Therefore, skillful managers want to reshape the market to fit their skills — rather than leave the outcome to dumb luck.

Seeking a judicious decisiveness

Coming to the end of his second journal editorship, economist (and Yahoo researcher) Preston McAfee reflected on the thousands of decisions he had to make — rejected 90+% of the submissions.

His views on decision-making seem directly applicable to the core problem of an entrepreneur: making decisions quickly based on incomplete information:

When Paul Milgrom recommended me to replace him as a co-editor of the American Economic Review, a post I held over nine years [1993-2002], one of the attributes he gave as a justification for the recommendation was that I am opinionated. At the time, I considered “opinionated” to mean ‘holding opinions without regard to the facts,’ and indeed dictionary definitions suggest ‘stubborn adherence to preconceived notions.’

But there is another side to being opinionated, which means having a view. It is a management truism that having a vision based on false hypotheses is better than a lack of vision, and like all truisms it is probably false some of the time, but the same feature holds true in editing: the editor’s main job is to decide what is published, and what is not. Having some basis for deciding definitely dominates the absence of a basis. Even if I don’t like to think of myself as “obstinate, stubborn or bigoted,” it is valuable to have an opinion about everything.

In his resume, there’s little evidence of any entrepreneurial bent, and in fact he was the economist helping the Federal Trade Commission attack the creative (if controversial) Rambus business model.
I do see one problem in applying his model to a startup. The editor of an elite journal has hundreds of very bright minds to draw on. Yes, half of them may say “no” when asked, and some have personal agendas. But still, this is a tremendous pool of knowledge that can correct egregious errors by the leader.

No such pool of knowledge is available to the tech startup, which raises the risks of overconfidence by its leaders. Certainly scientists (and to some degree engineers) tend to have the view there is one “right” answer. When it comes to the merits of an idea, some overconfident leaders tend to assert “this idea is wrong” when really “this idea is wrong for us.”

Henry Chesbrough famously noted that in innovation, firms often control for false positives (Type I errors) and predictably end up creating too many false negatives (Type II errors). This is one of the reasons he came up with “open innovation” paradigm — both to remind firms to avoid Type II errors, and to suggest specific mechanisms for profiting from good ideas that don’t fit.

Still, a judicious decisiveness is essential for any entrepreneur or entrepreneurial management team. From my own experience, it’s clear that postponing decisions often makes the decision for you. The key is that if it’s important, the startup can’t afford to “watch and wait” but instead must aggressively investigate to obtain the missing information.

References
R. Preston McAfee, “Edifying Editing,” American Economist, 55, 1 (Spring): 1-8.

Hat tip: pointer to McAfee essay via blog of Greg Mankiw.

VCs and green entrepreneurs

At the Academy of Management (#AOM2010) conference this week in Montréal, one of the more lively sessions was a “caucus” on Monday entitled “ Venture Capital Investments in Cleantech: An Act of Passion or Another Bubble In-The-Making?”

As the title suggested, the audience of about 30 academics argued about the rationality of VC investments, how they might be studied by academics, and of course the underlying industry dynamics.

It was an interesting albeit disjointed discussion. Imagine 15 smart (but perhaps undisciplined) people having a series of a stream of consciousness monologues (while the other 15 sat and watched). Still, this was probably the most concentrated grouping of management academics interested in cleantech entrepreneurship anywhere at the conference. Since most of the VC (at least by $) has gone to solar, this was also a RE/EE discussion.

Not all of the discussion was about VC. (Since I don’t personally study VCs, I only think about them in the context of the availability of funds for startups.) Some of the discussion was about more basic questions of studying cleantech startups.

My own question was about how SIC/NAICS doesn’t capture cleantech industries — the so-called “cleantech sector” — nor will it until the next revision of NAICS. (Cleantech seems no more a sector than nanotech.) The answer seemed to be that various “experts” have made databases by hand that they claim capture the entire industry. (Desiree Pacheco of Portland State seemed to be the most knowledgeable here, and had clearly been down this road before.)

The session was organized by Eva Yao (Colorado), Antoaneta Petkova (SF State), Sanjay Jain (Santa Clara) and Anu Wadhwa (EPFL). They have an (in progress) study that shows that high status VCs invested first in cleantech, and the room offered a bunch of possible explanations (such as the bigger VCs have more scale, more status and more slack).

In addition to Toni and Sanjay, the Bay Area was also represented by Geoff Desa of SFSU, Gregory Theyel of GreenVisions and CSU East Bay and of course yours truly. Colorado is a hopping place for renewable energy entrepreneurship, but my hunch is that overall California has the rest of the country beat.

Training entrepreneurial engineers

As an engineer who went to business training almost the minute I took my first job, I appreciate both the value and limitation of a business education for engineering professionals. Of course, I appreciate it even more now that I’ve started a copy, made tons of mistakes, got a formal business education and now teach at a business school.

Five elite US engineering schools got really nice press Tuesday in the Financial Times for their Master’s in Engineering Management program. With a little digging, it was clear they have a website and a “consortium.”

The FT article puts the best case forward:

Joseph Helble, dean of the Thayer School, says Mem programmes are “for engineering grads who know they don’t want to spend their entire careers in design or in a lab. They want to do broader, systems-based engineering, by identifying promising new product lines. They want to create a vision for the technology in the broadest business sense.”

As someone who once thought he would teach in engineering management, I’m all in favor of having more such programs.

Many engineering students have taken my technology strategy class, but without a business core, they get glimpses of insights, not the way to build upon a firm footing in accounting, finance, marketing and organizations. So an integrated master’s degree combining technical and business topics is a great way to address this.

However, when looking at the MEM consortium, I’m not quite clear what makes these five schools different. Do they have explicit standards? Is it the shared “MEM” brand? Or are they a members-only club? For example, the list excludes two of the top engineering schools in the country: Berkeley has a top-flight interdisciplinary Management of Technology program, while MIT has its System Design and Management program, jointly run by Engineering and the Sloan School.

And then there’s the question of how much any school prepares an engineer to be an entrepreneur. Plenty of engineers became entrepreneurs without formal preparation and did quite well. (Hewlett & Packard, Bose, Jacobs & Viterbi, and the list goes on...)

I don’t think a class or a business plan competition is going to help a 23-year-old engineering student create the next HP. Interning with a young company (as I did) would help, as would working for a top-flight company in the same industry. But nothing beats going to work in Silicon Valley, Boston or one of the other regions where you are immersed entrepreneurship, venture capital, startup infrastructure and a startup culture.

Teaching mobile programming

1st of 4 parts.

I’ve spent time the last two weeks visiting various universities to see how they do research and teaching on mobile phone programming. I’m posted on the research elsewhere, but wanted to summarize here what I learned about teaching at MIT, Georgia Tech and UCLA.

Based on what I’ve seen, the model for a mobile phone programming class would be a project-based class that combines needs analysis, use cases and prototype system development. Add in some sort of revenue model analysis, and you have a course that introduces budding software engineers to the possibility of entrepreneurship. (Absent revenue model analysis you have the typical engineering “make something cool and let’s see if we can sell it.”)

The class assumes students already have basic programming down pat, and also have at least one project course under their belt. This would be either a master’s level class or an upper division class that follows (say) software engineering.

MIT

Building Mobile Applications. The most famous course on this topic probably that by Hal Abelson, who is now the dean of teaching programmers at MIT. 24 years ago he shifted the EECS introduction to programming to Scheme (MIT’s local dialect of Lisp) with his text, Structure and Interpretation of Computer Programs.

We met with Abelson for nearly an hour to discuss his Spring and Fall 2008 classes offered to MIT CS undergraduates. The spring course (Building Mobile Applications with Android) got great writeups, perhaps because one of his teams won $300K in prize money from the Android Developer Challenge, Google’s Java programming contest.
The student application, Locale, was one of the first made available on the Android Market, and the team is evaluating commercial possibilities. Not a bad return for a 13-week class.

This semester, the class has 10 teams and about 40 students. For Abelson, a founding director of the Free Software Foundation, the iPhone was ruled out due to Apple’s NDA requirements for iPhone developers (abandoned too late for this semester). The projects are instead spread across three platforms.

Instead of being entirely about Android, the Fall course has more balance. Four projects are using Android in Java. Three are using Microsoft Visual Studio and the Windows Mobile SDK, supported by Microsoft Research New England. The remaining three are supported by the Nokia Research Center in Cambridge — two Java applications and one using Python (PyS60).

The class is extremely labor-intensive. Abelson credited Andrew Yu, (manager of mobile services for MIT) with much of the work. Each team has an industry veteran mentor — essentially a voluntary TA. Paul Oka of MSR said he spends 3 hours/week in the class and team meeting, and as much as another 5 hours early in the semester when the students are getting started.

Pervasive Computing. Abelson’s was not the first mobile phone programming course at MIT. Larry Rudolph taught a series of pervasive computing courses, first with the iPaq and then with Nokia phones using PyS60. From this effort, he wrote a boo to provide a Bluetooth programming tutorial and a 2003 pedagogy article in IEEE Pervasive Computing.

NextLab. Abelson’s course is not even the only mobile phone course at MIT this semester. If Rudolph’s course was more technology-oriented than Abelson’s, then the Nextlab course at the Media Lab is more project and need oriented, with a distinct social entrepreneurship spin. The “NextLab” course is part of MIT’s “Next Billion Network,” referring to the next 1 billion cell phone users expected to be added over the next four years — mostly in less developed countries.

We met one of the NextLab instructors (Luis Sarmenta) and visited a class session run by the other (Jhonatan Rotberg). We heard presentations by three projects, servicing Mexican farmers, rural Indian mobile commerce, and Boston low income preschool parents.

UCLA

In Spring 2008, Deborah Estrin taught “Current Topics in Computer System Modeling Analysis.” The class held 25 students — typical for a lab class — mostly master’s students, and was taught uses PyS60 with the Nokia N95.

The assignments are consistent with Estrin’s large research project on mobile sensing, with the students assigned to gather location and sound data and plot the data using Google map APIs. The 12 projects tended (not surprisingly) towards mobile social media.

Georgia Tech

At the College of Computing, there was other interest in mobile computing among researchers. I found two classes.

Mobile Computing. Thad Starner is a longtime (and prolific) researcher on pervasive and ubiquitous computing. Thus, it’s not surprising he’s taught several courses on “Mobile & Ubiquitous Computing.” I couldn’t find the website, but Starner said this semester he’s teaching about 45 students using the OpenMoko handset. Openness is a big deal to Starner, who is well known around Tech for not doing any business with Microsoft.

Augmented Reality Games. Conversely, Blair MacIntyre is teaching augmented reality (the intersection of virtual reality and reality) game programming using Gizmondo. This discontinued Windows CE-based handheld gaming console has limited communications capabilities.

Other Schools

My list is of necessity incomplete: I haven’t been able to visit all of the top C.S. programs in the country. Each school visit took a minimum of 2.5 hours, and then there’s the matter of the plane tickets. Here are a few that I found on the web.

Stanford. At least one faculty that I met this week mentioned Stanford’s course, “iPhone Application Programming,” being offered this quarter (Fall 2008).

Carnegie Mellon is offering a course this semester entitled “Mobile and Pervasive Computing.”

Latest research on high-tech startups

Since 2003, an informal consortium of four (now five) West Coast research universities have been hosting an annual research conference on technology entrepreneurship. The 6th West Coast Research Symposium on Technology Entrepreneurship was held earlier this month at Stanford (engineering not business). The other sponsoring organizations are USC, Oregon, Washingto and UC Irvine.

The conference is no longer just a West Coast affair, but this year drew an national (if not international) contingent of tech entrepreneurship researchers, with presenters from MIT, Wharton, Illinois, Northwestern and INSEAD (among others). Because it’s a small conference — consisting mainly of the paper authors and discussants — there’s not enough room for all the people who want to attend.

In an interesting innovation, Stanford has put MP3 files of all the presentations up on the Stanford Technology Ventures Program website. (Independent of WCRS, the STVP also has an impressive list of talks posted to its own iTunes U site as well as its own website).

Next year’s conference will be back at UW Seattle. Let’s hope that this innovation becomes a regular feature of the program.