Tim Oren's Due Diligence

VC Disruption? Part One.

I've finally decided to weigh in on the 'VC Disruption' thread, in part because one of my earlier posts seems to be heavily cited in the discussion. A couple of caveats before I proceed.

First, I'm not going to run straight at the problem. In this post, I'm going to define the issue more precisely in order to cut it down to size. Along the way, I'm going to reference and only partially define some terminology. You will need it to follow the argument, so if it's not familiar, chase the links.

This post may optimistically refine the discussion a bit, so those in the conversation can have a better defined playground should they want it. I'll then suggest some hypotheses re where we're headed in a second post.

Second, the one term I'm not going to use or define is 'disruption'. Any VC has seen that word sufficiently abused to cynically interpret it as 'bu****it'. If it's not backed up with some analysis of the mechanism of change, and who will benefit and how, its use can be discrediting. Let's try something else.

Let's Talk Transaction Costs

If you've read this blog before, you've likely noticed I'm a fan of Coase and Williamson's transaction cost economics. Simplified, transaction costs are all the overhead elements of doing a deal, from legal fees to insurance to search time to bargaining and on and on. (Go read the article).

Transactions are to this extent like any other good or service: If they cost more, there will be less demand for them. To take a relevant example, if the costs of the rather elaborate transaction called an IPO go up, due to the addition of SarbOx compliance overhead, then the number of IPOs will decrease. (And the average size and power of public companies will increase, due to reduced inflow of small companies, and the need to amortize SarbOx costs over larger revenue streams. Probably not what some of SarbOx's supporters had in mind...)

Before moving on, let me dispose of the notion that a reduction in IPOs spells doom for venture capital in general. All of you software guys spend a few moments investigating the pharma marketplace. Plenty of venture capital running around, but not very many IPOs. The root causes are the costs of FDA trials and approvals, and the need for a distribution network for the drugs, including expensive compliance obligations and insurance integration. Venture typically funds much of the technology risk of early drug candidate evaluation, and then exits (hopefully at an average profit) by selling out to an existing pharma company that can mitigate the other risks.

By existence proof, shortfall of IPOs won't stop VC overall, it just changes the parameters. We must dig deeper to find a VC killer. Dig where?

Transactions and Economic Organization

Another result of this school of economics is that transaction costs determine the organization of production. Look at it this way: If transactions are simple and cheap, there will be a lot of them. What's that called? A market. Consider construction as a category for a moment. Suppose I want an unskilled ditch digger. The amount and quality of work is self-evident, the pay is cash, the day laborers are more or less interchangeable. That market exists on a lot of California street corners.

Crank up the complexity. Now I want the bathroom remodeled. I don't want someone off the street, I want references, I want a contractor's license number, I want workman's comp insurance and detailed specifications, and I may need to have a city inspection. He wants a mechanic's lien, progress payments, has to pay taxes, may need to finance materials. The transaction costs have gone up. We are now in the realm of contract. The deliverables are spelled out, so are schedules, payment terms, liabilities for missed performance and accident.

Take up the complexity again. Now I want someone to manage the specification, leasing and/or construction and the install and ongoing operations of a major corporation's network infrastructure, both home office and distributed. When you're on this scale, with myriads of decision points, curve balls thrown at you by the users and the market, and an ongoing commitment, it usually results in an employment relationship. The organizational literature calls this hierarchy, but the decision boundary is reached when you hear: "We're paying too much to the lawyers to draw up contracts. Just hire someone and we'll make this up as we go." It's better to eat the upfront costs of the employment transaction, and amortize it out over a lot of projects in the future.

OK, now forget the construction example and keep the concepts. What about innovation? How does it play against the market/contract/hierarchy continuum? And where does venture capital fit?

Organizing Innovation

Let's start out by acknowledging something about innovation: Its ineffable quality naturally makes for a higher transaction cost. It's worse than the "I'll know it when I see it" qualifier. Innovation is more like "I'll know it when the market buys into it". Answering that market question is inevitably fraught, but without it we don't know if we've created value or a curiosity. Our model of innovator isn't Einstein, it's Thomas Edison.

The 'hierarchy' part of innovation is pretty obvious - that's the corporate R&D world. The innovators are employees, both technologists and product marketers. If their output pays off, they may get an inventor's bonus, a promotion, etc., but they aren't independent actors. Their downside (and their employer's) is the old technology transfer problem: before facing the market test, an innovation has to make it out of the lab or proposal stage and into the product side of the shop.

The corporate world of innovation exists in a sort of rough equilibrium with the VC world. We get innovators frustrated with transfer barriers and itching to prove their point. They get people with a lower risk profile or wanting a bit more stability in life. We each exploit the people networks created on the other side. And a lot more ideas wander back and forth in people's heads than the IP hard-liners would like to admit.

The Economic Virtues of VC Goobledegook

The existence of this equilibrium should give some clue to where we fit. VCs are part of the world of contracts. But with a peculiar twist: We write contracts to build companies. That is inherent in the VC's need to exit investments. The corporation is the container this society provides to capture equity value. As some would have it, that's our quantum size.

Now it's a peculiar thing that the contracts associated with a venture financing usually contain little to nothing about the actual business model or technology being backed. Perhaps a few items in the exceptions schedule and some milestones for unlocking tranches of an investment may give a clue to the real business at hand. The contracts themselves contain dense legal verbiage that implement term sheet items such as "2x participating liquidation preference". "broad based weighted average", "four year vesting with one year cliff", "pay to play" or "founders' stock repurchase rights". Arcane to the outsider, these are our equivalents of the boilerplate language in your bathroom remodeling contract. They have evolved over the years as semi-standardized ways to govern the process and mitigate financial risks of turning raw innovation into equity. In short, they reduce transaction costs. If a VC says "There's too much hair on the deal" you can hear it as "It doesn't fit the standard patterns; there will be too much legal and management overhead for this to be profitable."

So if we've gone that far in standardizing the container, then what about the contents, the actual market or technology innovation? That's the unavoidable hard part, which actually requires knowing something about the target market and technology, or being able to run a process that's sufficiently informative to proceed in the face of doubt. "Domain expertise" sorts out the investors by target markets, stage of investment, and risk preferences (do you like 'technology plays' or 'consumer plays').

When Models Change

OK, let's recap. Existing companies and the VC process are part of the overall innovation landscape. (I'll bring markets into it in the next post.) The boundary between them is related broadly to the transaction costs associated with bringing innovations to market. In the VC world, you can sort out the transaction costs into deal structure and domain specific costs and risks.

It you're with me so far, you should see the obvious corollary: If transaction costs of innovation somehow change systematically, the relative territory of VCs (contract) and corporations (hierarchy) are going to change as well. Costs up - corporates gain; costs down - VCs gain.

Case in point: The new IRS regulation called 409a has introduced uncertainty and apparent risk into the world of startups. Until and unless this risk is understood, systematized and mitigated, the introduction of this regulation will reduce the number of VC deals, moving some of them back to the corporate side (or to limbo).

I Know That's Not What You Asked

The ongoing discussion doesn't center around term sheet arcana or IRS regs. It's about a potential change in the investing landscape due to a shift in software and services architecture, that suffers under the indefinite name "Web 2.0". To be specific, we are talking about a conflation of:

1. A uniform TCP/IP transport layer for connectivity among services and users


2. An open source software (OSS) process that has created cheap, consensus platforms such as the LAMP stack


3. Lightweight data standards including RSS and other XML derived specs for 'web services'


4. RESTian architecture and other measures to control the complexity of distributed systems

I'm sure I missed a bit, but that's the gestalt. But is this related to transaction costs?

Yes, big time. Asset specificity is one of the elements of transaction costs. Specificity relates to the ease (or lack thereof) of mixing and matching a particular asset among potential transactions, and vice versa. Want generic white tile for the bathroom job? You've got a dozen local sources to bid for your business. Want that just perfect off-white bisque from a small town in Italy? Gonna cost you.

So too in software. Back in the bad old days, there was no such thing as an external market for (say) a source code maintenance utility written in Burroughs' dialect of Algol. That's a highly specific asset. Each layer of technical standards, accreted over time, has served to reduce the specificity of software, making it valuable in multiple settings. Voila, CVS!

The economic value of standards, including software standards, is the reduction of transaction costs by reducing asset specificity. There's the linkage. It's in the 'domain' part of the venture process. We're talking about whether the architectural shifts listed above are a sufficient systematic change to transaction costs in the software domain to alter the efficient boundaries among hierarchy, contract, and market in organizing software innovation. And perhaps, in wider services markets as well.

That's a lot of words to ask the question, but some precision and the requisite vocabulary may prove useful in addressing it. I'll get back to that Real Soon Now.

Be Vewy, Vewy Quiet - We ah hunting... MSM Sales Forces

Via John Battelle, a pointer to a Google pilot to auction off surplus space in print magazines. This follows right on the heels of Google's acquisition of a radio-ad insertion platform, which in turn barely follows the pilot phase of a newspaper ad purchase program announced late last year.

What tha? Why is Google haring off after the legacy media, in preference to pushing the net edges? Is this the company's typical 'energetic motion in all directions' style, or something more serious. The latter, I think. There's a real schwerpunkt to this movement:

As the MSM slowly circle the drain of irrelevance, one remaining asset is their ad sales forces. If you're a small business, either local or just too small to warrant the attentions of an agency, the advertising sales calls you get are from the local newspaper, the radio stations, the phone directory, and perhaps print mags that serve your topical market. For a long time, that's been your only way to get the noise out, unless you want to post on kiosks and litter parking lots with flyers.

We all know that programs like AdWords have opened up an alternative path, perhaps best suited to topically focused businesses, but increasingly used for local context as well. So a substantial number of smaller businesses are now buying space online, as well as in legacy. What could be more convenient than to have one stop shopping for their advert needs, perhaps adding in some net-like features like auction pricing and topically relevant placement?

As the mass media cash cows are milked, they've slowly whittled down the newsrooms to ill-disguised opinion writers. But the ads sales forces have been more defended - that's where most of revenue originates. Here comes Google with a nice proposition: Fill up that extra, marginal space with new buyers, without undermining your existing business. Incremental revenue. Give it a try (maybe you don't need all those guys on the phone.) Have this needle. Go to sleep. You won't feel a thing.

The War and Technology Discontinuity

Four years from our entry into World War II, the battle was won, and the landscape of technology had been transformed. Radar, jet aircraft, atomic weapons and many others came to fruition in those brief, terrible years. War replaces normal investment concepts of ROI with the calculus of death and hazard. Investment flows towards developments that would be premature in peacetime, and is often siphoned away from areas with no application to the conflict.

The ill-named War on Terrorism (really the war with Islamofascism) now has a longer duration than our involvement in WWII. Where are the attendant technology transformations? To be sure, since this is a long term conflict with an economic aspect, we've sought to minimize the displacement of commercial investment. But having started in many ways ill-prepared for an asymmetric conflict, should we not see an acceleration in technologies that address those shortcomings? This year's In-Q-Tel CEO summit gave me a chance to reflect on the question.

In-Q-Tel is the intelligence community's venture investment arm, which reaches out to early stage ventures and other private interests to acquire technologies needed by that community. The CEO summit, held at an undisclosed location in Northern California, brought together several hundred investors, venture company officers, In-Q-Tel staff, and agency and DoD personnel ranging from the very highest level to senior analysts and even operators. Last year, I marked the event with my own ad hoc best of show awards. Since this year's roster is largely the same, that'd be rather pointless. Instead, I watched for the emergence of war driven technology shifts, since intelligence is central to success in an asymmetric war. If In-Q-Tel and the community have been doing their jobs, after four years of war we should be seeing some maturity in both their portfolio and the systems built from the acquired technology.

Did I see such technology shifts? Yes, in multiple cases, both real and prospective. Here are a few.

The venerable BBN consultancy is a contractor to the agencies, including recognition and retrieval of audio. They have put up an end-user version called Podzinger, which allows a plain text search over the contents of podcasts. The coverage is regrettably narrow, but you can get a notion of the capabilities of what is likely not the best they have to offer. (And also a demonstration of the shortcomings of unscripted conversation when rendered into text.) There is obvious dual-use here, as the relationship between search and advertising revenues is now evident to all.

Venture funded Language Weaver showed a taped demo of a bidirectional Arabic - English speech translation system, running on a laptop and used to interview informants and suspects in the field. It's not possible to assess its speaker independence and range of domains from a tape of course, but it's a reasonable first draft of something I've wanted since I first read about translator discs in Larry Niven's Ringworld, over thirty years back. Based on the pace of AI to date, I'd more or less given up on seeing it.

Neither of these, and few of the other unstructured information technologies, are based on clever symbolic AI or knowledge representation. Their roots lie in corpus based linguistics and statistical modeling, using lots of sample data and processor time. At the root, these problems are being beat to death by the combination of cheap silicon and lots of money. The resemblance to the Manhattan project, taking just-adequate technology and running it very hard, is not accidental. If you're one of the bad guys, or someone who gets bent out of shape when the agencies try to monitor those who might be talking to the bad guys, or even concerned about behavior-based marketing, you should be worrying.

You can see other pieces of the future emerging as well. In-Q-Tel has a variety of investments in sensor related technology, including network architectures, next generation protein and gene chips, and fundamental nanotechnology fabrication techniques, with wide applicability to chemical and biological threats. We didn't see any complete systems, but I don't think we're on the need-to-know list either. Much of this is also dual use. Signatures like gene activation that diagnose a biological attack also have applications in disease screening. Combined with work ongoing in the healthcare domain, it's increasingly likely that we'll be able to diagnose the preconditions and onsets of disease long before having specific treatments in hand. The technologies that hopefully forestall any Al Qaeda bio threats will also be smashing the notion of pooled-risk health insurance.

Focusing on the intelligence community misses some of the technology discontinuities being generated on the 'kinetic' side of the house, such as robotics and other semi-autonomous systems. But it's enough to show that war-borne technology discontinuities are not a thing of the past. In unstructured information and sensing, there are big changes coming that align with existing drivers in civilian markets, and they are quite evident to the armchair open source analyst.