Tim Oren's Due Diligence

Dissecting the Media: Trust and Transactions

The legacy media in general are threatened by audience loss to the Internet and citizens' media, and Rathergate is merely the latest example of a credibility apparently sinking by the day,

There's been ample analysis from the perspectives of professional media, journalism, and politics. But from an investor's perspective, there's the possibility that one of the major value chains in modern society - media and advertising - will be rearranged, at least in part. That makes an economic analysis of the issue rather interesting.

I'll start from the perspective of transaction cost economics as originally invented by Ronald Coase. If you haven't encountered it, go read a bit. Trust me, this is one of the master ideas for understanding the evolution of the Internet.

In the venture capital trade, transactions costs usually turn up in a pitch from a company proposing a better scheme of microtransactions or other technology so that readers can pay for bits of content on the net, a few pennies at a time. While this notion has been kicking around almost as long as the commercial Internet, none of these schemes have been successful, and no investors have cashed out, though they keep trying.

Why not? There's more to transaction costs than the mechanistic elements of the payment mechanism. There are also what are called search costs, the time and effort expended in finding the right thing to buy. And there's that little frisson, wondering if what you've bought will be worth the expense - direct and indirect. This element turns out to be significant. In fact, American slang has an expression for it: "Being nickeled and dimed to death." This is not a positive sentiment, and this visceral embodiment of an economic theory has been sufficient to keep microtransactions a dead letter. (If there are equivalent expressions in other cultures, I'd be interested in hearing about them.)

You may visit Andrew Odlyzko (PDF file) or Clay Shirky for a further dissection of micropayments, but I will pass on to strategies for overcoming these problems. A common way around the nickels and dimes is 'bundling', which is simply selling a number of related items together. Here a real world example may help:

Suppose you'd like a few dozen packets of instant oatmeal in various flavors to feed your child. You don't have shelf space to keep full boxes of all of the flavors, so more than likely you buy two or three cartons of assorted flavors. Each of those cartons is a bundle. You give up the flexibility of picking each packet separately, but also lose the time and minor effort involved in making the selection. The manufacturer and retailer gain the benefits of scale economies in manufacturing and simplicity in stocking. This strategy is so stable that single packets of oatmeal are almost unobtainable.

Note there's a modicum of trust involved in creating this bundle. You may be willing to take the chance that a few packets will be a flavor that will draw a 'yech' from the kid, and you'll end up eating them. But if you ended with an 'assorted' carton that had been half-filled with a dud, you probably wouldn't buy that brand again. Note the brand as a source of trust, and as damaged if the trust is removed by experience.

(Let me acknowledge right here that I'm doing a lot of simplifying to get to the meat of the argument. Bundling can be beneficial to both producer and consumer, but it can also be used strategically as an anti-competitive weapon. I will acknowledge the legal and regulatory side of the issue and move onward.)

Getting on to the media, we can note compact discs, and magazines and newspapers bought off the rack as bundles of content in the real world. Many such bundles have been driven by the physicality of the medium, and the economics of creation and distribution. In days past, I might cheerfully pay a quarter for a copy of the Chron, only to read the sports, one column of classified ads, a few local stories and throw the remainder away unread. All of that wastage was acceptable to both buyer and seller, given the scale economies of printing. Note again a modest element of trust: If tomorrow I buy the same paper, but my interests have changed from sports to movie reviews, I should believe that information will be there and be useful and credible.

These bundling strategies are not stable of themselves, they exist only within the context of technology, distribution and transaction costs surrounding them. When these change, bundles may collapse. The CD is the obvious example. With individual digitized songs now easier to duplicate and distribute than the physical bundle, the albums raison d'etre has disappeared. As the simple playlist replaces the album's remaining value of simplifying choice, CDs commence a slow glide to oblivion, moderated only by the installed base of equipment and consumer habit.

The newspaper bundle, with over 350 years of market experience behind it, is a tougher nut, but it is also starting to crack. Your average newspaper takes over 2/3 of its revenue in the form of advertising, and it's there the damage is deepest, particularly in the classified section. Between eBay, Autotrader, friendster, match.com, monster.com and their competitors, the competition is already large and continues to grow. And it's also obviously growing on the content aggregation side, with services like My Yahoo and Google News, and all the myriad of blogs and other citizens' media. If these choices were only the equal of those available in the fish wrapper, the bundling argument - "we simplify things" - might still hold. But each of those services has advantages in scope and personalization that the newspaper cannot match. I start to become less happy about the quarter for the Chron when I know I can scratch today's information itch online faster, better and often for free.

But wait, it's actually worse than this. Newspaper and other periodicals would prefer to have a subscription relationship with their readers, rather than rely on newsstand sales. It makes the circulation more predictable for the advertisers, and smoothes out the reader payments, which would otherwise peak around major events and fall off dramatically during slow times like the proverbial August 'silly season'.

A subscription is a bundle of a different sort. It combines multiple transactions into one by collapsing them in time. It therefore adds a futures element to the transaction. More so than the spot transaction of buying a single item, trust becomes an issue. The purchaser is betting that the supplier will be reliable in the future and, in the case of a media periodical, continue to deliver a collection of content of value. Working out the mathematics that express this choice in a given situation can be pretty darn complex. But it's certainly possible to make some obvious qualitative deductions:

If the subscriber observes that the value of the content delivered is decaying over time, either absolutely or compared to competitive sources, then a renewal becomes less likely - the futures bargain no longer works. If this is widely true of a subscriber base, then churn (lost subscribers) will be increasing over time. To keep the revenue line stable, the subscribers must be replaced, incurring subscriber acquisition costs. Since every new subscriber's choice is made in the context of competitive options, including spot purchase as an alternative, the per-subscriber acquisition costs may be rising at the same time. At the point where discounted future value of a subscriber, given churn, becomes less than acquisition costs, the business model that looked like a reliable cash spinner is suddenly upside down. All that is left is to milk the existing subscriber base as it decays.

Up to now I've tacitly assumed that content may be more or less interesting or relevant, but has no valency. Now let's drop that assumption. Suppose the reader discovers through independent means that at least part of the content being purchased is inaccurate, either in fact or through unrepresentative sampling of the reality. And further, let's assume the reader is convinced that the basis for this is not incompetence - which would simply lower the value - but is intent. In short, the buyer has been sold propaganda. If the buyer is in search of objectivity, the value of the bundle decays. In the case where the ideology of the bias is inimical there will now be a negative value assigned to that part of the bundle. Only in the case where the buyer is aligned and not searching for ground truth will an increase in value result. This will cause a differential churn and subscriber acquisition cost depending on the beliefs of the buyers, and discovering this bias in one element of the bundle is likely to tarnish the reputation and brand of the whole in the eyes of those not in concert. We have just reinvented the party organ newspaper. We might also have invented a way to salvage a decaying legacy business by guiding it into a stable, but ideologically defined niche. I leave the search for examples as an exercise for the reader.

I've simplified this by focusing on the readers' decisions and on newspapers as a physical bundle. Of course, the actual situation is more complex. Newspapers and networks are also ways in which viewers are bundled for sale to advertisers, who hope for increased sales of their own goods and services. A churning subscriber base will decrease the certainty of this bundle, and send more advertisers to spot purchase vs. long term buys. Introducing ideological bias has the possibility of creating negative splash back on the advertisers - boycotts do work. With the advent of channel surfing, television networks have become bundlers less to consumers, and more to advertisers. The effects on advertising placed on network branded entertainment shows ,of perceived bias in news programming from the same source, is right now an unknown, but the experiment is underway.

The increasing competition from Internet information sources, and the exposure of incompetence and bias via citizens' media, will have a fragmenting effect on existing information bundles. The attempts of many legacy media sites to forbid 'deep linking' and/or set up registration walls is testimony that they are - whether analytically or viscerally - attempting to retain the illusion of bundle, rather than the reality of cherry picking by the audience.

But what of my original premise, that on the average the reader would prefer to avoid the granular choice or purchase of content? If the winds of change have blown apart the legacy media bundles, can the value of transaction cost reduction be recreated in another fashion, and revenue extracted for it? Now that's a question to get a venture capitalist's attention!

If the answer were obvious, we'd already be there. And if I had a lock on it, I'd darn sure get a few investments placed before blabbing around the blogosphere. But neither is true, so let me speculate a bit....

• Google's business model is provocative in partially reassembling the bundle from the advertisers' point of view. Through search related ads, bundling around declared interests rather than demographics can be achieved. Adsense goes further in attempting juxtaposition of ads with actual content on the same basis. I'm awaiting with interest the form that advertising will finally take on Google News. Google is leveraging cheap cycles and a lot of algorithms research against the bundling needs of advertisers, but largely leaving the readers to fend for themselves. But, it has the advantage of a clear business proposition.

• RSS aggregating software and services are a provocative attempt to let the readers build their own bundles. This is impossible in the legacy media, and creates a sharp differentiation from the old style of bundling. The juxtaposition of citizens' media (blog posts) with legacy media content ripped from its home site goes one step further in exploding the apparent value of the old bundle. Reader side aggregation can thus destroy old value, but hasn't so far shown an ability to extract serious revenue from readers.

• Technorati is another cut. It's not a bundling solution at all. Instead it seeks to reduce the 'search costs' associated with following threads of interesting discussion across the Web. If the transaction costs of retrieving individual information bits is reduced, the need and attraction of bundling is reduced. But, there's also the problem of a lacking business case. Perhaps that can be found from the advertisers' side. If promotion to demographic or general interest bundles is giving way to selling by influence, then tracking the conversation becomes of value. Technorati appears to be a radical unbundling hypothesis on both the reader and advertiser sides.
  

Some part of the media value chain is becoming collateral damage of the Internet, further accelerated by war and politics. Motivated by business opportunity, ideology, and just plain fun, the insurgents are gunning for the legacy media. The game is afoot!

Update: Jeff Jarvis comments and summarizes:

What's so fascinating about Tim's post is that he takes a social issue -- news and trust -- and measures it through a business perspective. I have always said that in the news business, our only asset is credibility. Tim is now measuring the declining value of that asset in the midst of scandal and in the face of new, trusted competition.

Yes, thank you. That's exactly what I was trying to say. One of the best things about this medium is you can get other bloggers to write your precis for you!

A Eulogy for HyperCard

To the surprise of few, Apple's Hypercard passed away quietly this week, after life support was finally withdrawn by the company. It had a run of over 16 years - though the last were in circumstances of at best benign neglect. Not a bad duration for a software product, but it still hurts to see it go, since I had some part in its gestation.

HyperCard was Bill Atkinson's brainchild. Though influenced by a number of others, most notably Ted Kaehler, Dan Winkler, and Chris Espinosa, it was
Bill's vision, tenacity and willingness to go to the mat with John Sculley and even the board of Apple that got it out the door, over the resistance of Jean-Louis Gass�e and others who saw it as 'competing with our developers'. It was launched with a typical Apple PR blitz at Boston Macworld in 1987, and was also a centerpiece of the fall 1988 launch of Apple's first CD-ROM drive and its push towards mulltimedia. (The latter was my show; my role was producing large scale database and multimedia pieces with HyperCard to show off the CD-ROM and multimedia strategy, and get some developer momentum going.)

HyperCard always had a marketing problem of not being clearly about any one thing. Since it was initially packaged with every Mac shipped, it's likely the majority of buyers used it as a quicky Rolodex, if anything. But HyperCard's biggest win was a very low entry threshold for those who wanted to build their own 'stacks' - combinations of user interface, code, and persistent data. There were plenty of examples to suggest ideas, and all the code was open for tweaking. This did enable a burst of creativity by users, many of them educators and artists with no training in programming or database.

The proliferation of ideas created its own confusion. What was this thing? Programming and user interface design tool? Lightweight database and hypertext document management system? Multimedia authoring environment? Apple never answered that question. Probably the answer was 'yes, all of the above', and HyperCard could have been forked into several related products, each tailored to a specific market. But instead the forces against internal software projects won out, and HyperCard was shunted off to Apple's Claris spinoff, where it lost in the battle for attention with Filemaker and Apple/ClarisWorks. Several improved versions came out, but the code was never even completely ungraded to handle color displays, killing off interest from the multimedia and UI markets. Hard core supporters, particularly from the educational community, kept it alive when Apple reabsorbed Claris, but only on sufferance.

From a software architecture point of view, HyperCard had a number of interesting ideas which might bear reexamination. At a time when persistent object stores were still novel, HyperCard was built around one. It's not going too far to say that its user interface was simply a reification of the object database. HyperCard's programming model was object-like, but didn't fall neatly into either the class/instance or delegation styles. Individual visible cards in a stack were created as instances of prototypic backgrounds and could be pre-populated with text fields and action buttons. Default message passing was an odd hybrid of visual containment and fixed object hierarchy. These features, plus a very texty scripting language, seem to have made for a very approachable tool for the nonprofessional coder or database creator. (To my knowledge there was never a study of programming usage and usability of HyperCard. A real gap.)

HyperCard also had its share of problems, particularly as a programming environment. Almost uniquely, every data element in HyperCard was a string. (Anyone else remember SNOBOL?). If you wanted another data structure, you built it out of strings, or expanded the programming language by adding in new compiled code primitive 'XCMDs'. The binding of code to individual cards and objects within them made it easy to create an unmanageable project with code snips splattered all through the stack, and many of the neophytes fell into this trap. Ditto the lightweight database had very weak identity and abstraction capability, another trap for the budding multimedia author. The tight binding of interface and data store also created weaknesses only obvious in retrospect: There was no cut line at all between client and server, and creating one was probably impossible in the original code base. HyperCard implemented everything in script code, even links. It was about as far from RESTian as you can get. If HyperCard had in some way mutated into an alternative to the Web, we'd be living in an even worse malcode Hell than we've got now.

So, adieu, HyperCard. I had a heck of a lot of fun with you (and the gang that birthed you), and I know others did as well. Your passing will leave a gap, particularly for the educators who still have no clear (and cheap) alternative. I hope some of the lessons you taught are passed on to new projects that allow just plain folks to try out coding and authoring.

No Exit: When Venture Capital Isn't Right

(Ephemeral update: Welcome, eTechers. Please check out this post for more recent thoughts, particularly if your ideas are OSS based. There's other stuff of potential relevance in the VC category on the sidebar.)

Everyone in venture capital gets plans that might be great businesses, but don't fit the venture model. And one also occasionally sees ideas which might fit the model, but only by stretching the original notion in a way that distorts it, perhaps to the detriment of the founders and the business. This is a first take at laying out the parameters of the model, and of these two no-fit situations.

OK, so what's this 'venture model'? We have to start with how a venture fund is organized. The garden variety venture capital fund is a limited partnership (LP) with a fixed period. That means the investors (limited partners or 'limiteds') commit to put up a defined amount of capital, which is drawn and invested over a period of a few years. The partnership agreement typically concludes seven to ten years after start, with options for extension for a few years depending on market conditions. During the partnership period, the capital is invested and controlled by a management company which is usually organized as a limited liability corporation (LLC), and is the general partner of the LP. It can and typical does manage multiple venture funds diversified in start year ('vintage'), investment focus, or stage. For most people, the management company is the continuing, publicly visible presence. When you say 'August Capital' or 'Kleiner, Perkins', you are talking about a management company and its people.

Actual investment dollars, though, come out of one or more of the limited partnerships, which are finite in duration. The limiteds expect their money back, with a healthy ROI, within the term of the partnership. Each of the 15-25 venture investments made within a fund is usually in a private company which has no open market for its stock - it can only be resold to qualified investors, who have to go through lengthy diligence processes to value it. Even if it has increased in value, the limiteds of the fund don't want to receive their pro rata share of this stock as their eventual return. They want saleable shares, or preferably cash. So each of the fund's investments should 'exit' or 'get liquid' before the end of the LP term. The most likely ways to exit are by initial public offering (IPO), where you end up with shares that can be openly sold, or by having the company acquired by another that can write a check that won't bounce, or that is already public and has liquid shares to swap.

The need to exit is the core of the venture model. You might have a wonderful business idea, that will make its customers, employees, and managers all happy, but if it is unlikely to exit, and with a sufficient return on the capital put in, we don't want it. It could in fact generate wonderful cash flow forever, but if we can't monetize that in a liquid form by the end of the fund, it won't make it through the partners' meeting. It is inherent in the nature of the beast.

What makes us believe you can exit? The best (and sometimes circular) argument is if you are in an industry segment where exits happen. Where the public markets have a past appetite for IPOs. Where there are large companies that are known to grow by acquisition. Preferably both. Got something that Cisco or Broadcom might want to buy, and that could get public if things go really well? Ding! Go on to the next test. No? Better have some reasoning as to why you're like something that's worked before, or a really convincing argument of some buyer waiting to scoop you up when you win, even if it takes five years.

What should the exit look like? Everyone has their own numbers, so let's just play with the proverbial 10x the invested capital. A good return, though not an ohmigawd-it's-another-Netscape return. In fact, it's about 39% compounded interest if the holding period is 7 years. Even if it takes 10 years it's still 26%. A bit better than a CD, eh? Maybe everything the lefties say about rapacious VCs is true? I wish. Problem is that's the upside picture. That return also has to make up for doggy deals that have lesser return, or that lose the invested capital entirely, and still let us make a decent total return for the venture category (which is in the mid-teens, with a scary variance). The nasty bit from the entrepreneur's standpoint is that your wonderful business plan, sure to succeed, is essentially being taxed via that horrendous discount rate, all because of those other wretched loser deals we're going to do. If only we knew which was which.

Now we can do a few numbers. For simplicity let's say there was one round of investment, of $15m, with a generous pre-money valuation of $9m. We now need a convincing case that a decently winning outcome will produce a total enterprise value of about $240m in five or six years - which we suspect will take a year or two longer anyway. Time to look for comparables: what were the price/sales or price/earnings (you should be so lucky!) ratios for recent IPOs or purchases in your category? Beyond the direct financial implications, these ratios have baked in a lot of market wisdom about margins, entry barriers, pricing power, and so forth. Higher is better. Let's just pick a whopping 3.0 price to sales, you lucky dog.

So you need to book $80m in sales five to six years out. Now your Wonderful Widget takes 12 months to cook, even with prototype in hand. Year one revenue: $0. You sell the Wonderful Widget to an industry that turns over its products every 9 months, and has a three month typical sales cycle to its customers. Year two revenue: $250k for some samples and a bit of NRE. You closed two OEM customers, representing 15% of the total industry volume, but they only put it in their high end product. Year three revenue: $5m. Imagine these data points on a graph, along with year five/six: The 'hockey stick curve' is born. OK, we don't believe the specfic numbers on the chart any more than we do the Gartner numbers, but there had better be a good scale-up argument. Why an initial penetration can turn into a rapid breakout, and why your technology, sales channels, management team, and suppliers can all sustain that growth.

That's a simplified picture, but it should show why some business types usually don't make it in normal times. Professional services? Low multiple. Market share game. Scales by hiring people, not running the fab line or CD press or download site. Next!

That's one end of the game. Now the less common case. Suppose you actually could fit the model, but you only need (say) $1.5m of capital to create the Wonder Webservice. You'll get some chunks of the functionality on open source, plug into published APIs, get the protectable algorithm on a cheap license from your old university, and are just going to sell it through the net anyway, using a promotional model that can largely self-fund if it wins. If those stingy VCs only give you a premoney valuation of $2m, then if you're valued at $35m five years out, they ought to be happy! Well, maybe. Who ever heard of a $35m IPO anymore? Even the investment bankers aren't that hungry yet. For that matter, how many funds want to deal with that small a raise? That doesn't move the needle if you're managing $200m. And there's always the sneaking suspicion that anything that can be built that cheap doesn't actually have enough competitive barrier to stop the big boys from just reverse engineering it, instead of ponying up for the acquisition.

But all this can be fixed! Retool the plan to create a bigger barrier, and burn more cash. Maybe it needs to be an appliance, not pure software. Or a blade or a chip, that would be even better! Probably we need to build a complete distribution based sales channel! OK, you can see where I'm going. Sometimes that's piling up risk needlessly, and turning cash consumption into a bogus security blanket. And taking on a need for an exit where that might not fit the inherent business situation.

It's possible that given some of the enabling forces I mentioned for the Wonder Webservice, that there will be more business plans with the option to run capital light. If your grand idea fits into this category, consider the big capital option, but give just as much time to figuring out how you could cut the costs even further: turn beta users into discounted paying customers, play guerilla marketer, cut a revenue sharing sales deal with someone who cares about cash, not exits. The VCs won't be offended, there are a lot of fish in the sea for us to chase.

Ruminations on venture capital, trust networks, and information theory

Update 11/3/03: I've noticed most of the inbound traffic to this article comes from a site called 'Anti-Venture Capital'. If your concern in visiting here is to figure out whether it's worth your time looking for VC, I suggest you also look at this post which specifically talks about which businesses don't fit the venture model. Here's a list of all of the VC oriented posts that have appeared in this blog. Might I also recommend VentureBlog, run by the August Capital folks. Read it all in as a cold a blood as possible, considering it's just your livelihood on the line, remember the guy who sent you here has something to sell as well - not that there's anything wrong with that - and make your own decision. And may I wish you the best of success in whatever venture you have in mind?

Since the last few posts have been serious geeking, I thought I'd try something more related to the day job. The following is edited from a post I wrote for an old college friends' mailing list last November, so it should be 'suitable for general audiences.' I've updated where necessary. I should say I'm relatively junior in this biz, at it for three years, so I'm sure a lot of veterans will look at this and think 'Well, duh!' But then one of the few advantages of being a newbie is the possibility of reflecting on your work before it becomes completely internalized. Anyway, here's a bit of analysis of the VC scene from the points of view of 1) transaction cost economics, 2) internal efficiencies of VC firms and 3) the current post-bubble funding environment.

1) Venture capital activity in general, and early stage private firms in particular, are what economists call very 'specific' goods. There are valid reasons that the Congress and SEC have made it impossible for most investors to give their broker a CUSIP number and get a few shares of SmallPrivateAmazingWidget Co. or even Joe Beet's VC Fund. The majority of these firms are talking about filling needs that aren't yet proven, in markets that may not exist, with products and services they haven't built yet, with financial plans that are often - ummm - sketchy. Getting enough information to put a fair price on that kind of deal, even if it's real and not a fantasy, is a full time job and very easy to get wrong even then. Same goes for VC funds themselves, figuring out who can pick and manage early stage deals is itself a bit of a chore.

In economic terms, the transaction costs of this kind of deal are quite high. Finding the deal, assessing its value, negotiating it, and then making sure it goes vaguely according to promise constitute big overhead. Economic theory suggests that in these circumstances, transactions don't happen in open markets (like the NYSE) instead they proceed through one-off contracts (or are funded internally in existing firms - that's called R&D). Sure enough, almost all VC deals are defined by one-off contracts called 'term sheets', though these are constructed from more-or-less standard sets of terms and conditions Another thing suggested by economic theory is that in the face of uncertainty, contracting parties will try to mitigate risk by trust. So here's one open secret of the trade: venture capital is not at all an open market, it is a trust network. Get this straight and you'll know more about VC than most of the financial writers out there.

In the face of all the uncertainty elements I gave above, deals are often made or broken based on direct or indirect personal connections among the parties involved, because reputation is the best assurance of both diligent performance, and good faith in sustaining the deal if things don't work out as planned (a 99% probability). One of the best sources of trust is "He made me money before". (By the way, there's even a private VC database called VentureSource where you can find out who was involved in most any deal and how it fared. Need to check up on someone? Find out what deals they've done, and you'll almost surely known someone who was involved directly or at one remove. )

As far as I can tell, though, being from the right sorority or fraternity gets you bupkas, at any rate no one has tried to give me a secret handshake yet. The networks that count in Valley VC, in rough order are:

A) associations among VC firms and partners and seasoned entrepreneurs

B) who you worked with before, and what they thought of you

C) ethnic networks

D) the ol' school tie, sometimes

E) non-business networks like soccer parents, public affairs clubs and religious affiliations

(A) is fairly obvious. If you do deals with other firms you find out what kind of thing they like, what and who they know, where their expertise lies. Got a deal that's maybe interesting but you can't evaluate efficiently? Toss it to a friend who knows the area. If it works out he'll offer to cut you in on the deal, even if not he'll toss one your way someday. This is part of our economic function - we're a message passing utility that tries to line up people, money and opportunities in the presence of huge amounts of noise and uncertainty. Once you've been an successful entrepreneur at the CEO or exec VP level, you're also in this network. (Being an officer in a public company does not necessarily make you part of it.)

(B) is the most powerful in terms of quantity. Work with someone in a tough job and you'll know what he or she is made of - it's the business equivalent of 'foxhole loyalty.' If I worked with and trusted good old Joe, and he in turn tells me someone works like a dog and knows his stuff, then he'll get a look. There are large companies that have notoriously strong cultures - and high turnover - and thus become a network nexus of referred trust. I'm a member of the Apple Mafia, having served 8 years there. I can probably check up on any former Valley employee of Apple in no more than two hops. One of my partners is Intel Alumni, that's even better. Big Valley firms create a secondary public good (externality) by putting diverse people into pressure cookers and turning up the heat, forming trust networks.

(C) is part of what keeps this from degenerating into an ol' boy club. According to my friend AnneLee Saxenian at Berkeley, over 25% of Valley startups now have CEOs of Indian or Chinese descent. (Lame local joke: "The Valley is built on ICs. Yeah, Indians and Chinese!") One of my partners is Indian, so he's tapped into the so-called TIE crowd. (TIE = The Indus Entrepreneur, both Indians and Paks). These guys are awesome networkers. Being Israeli or Korean or Brit also helps, but not quite so much.

(D) the school thing, is variably useful. Being from a smallish, select, tight school setting does help, say Stanford business school. Being from a large, diffuse, not-so-select school doesn't usually help unless there's something like a work relationship as well.

(E) Trust and introductions spill over from social settings, but usually have to be corroborated through business networks. Can get you in the door, though. We've looked at deals picked up from a partner's cabinetmaker, and via a friend's daughter's fiance. I you want to leave your work at the office, this is not the place for you.

(I should also mention geography as a negative selector. Since early stage investing is inherently hands-on, many funds won't do early deals very far from their home offices. Our rule is now 'Pacific time zone only.' Distance also makes the networking more difficult, even in the Internet age.)

If you're an entrepreneur trying to get a listen, you really need someone on the team that connects into the VC community as far up that list as possible, with as strong a recommend as possible. Remember that person is putting their reputation on the line when they connect you, so treat them well. And if you repeatedly can't find that connection, go look in the mirror, the world may be telling you something.

2) Second big topic, and another open secret of VC: One part of operational VC success is saying 'no' efficiently. Consider the following: We fund about half of 1% of the deals we see at some level. This number is typical or even a little high. When I pick up an unknown plan, my prior probability that I'm going to dump it is 99+%. My cost is the time I may take reading it or checking up on the idea, the market, the people. A 'false positive' is the penalty case, I spent that cost with no reward. You can see why I'm a priori biased to drop it if anything looks amiss. This is one reason that VCs get the deserved reputation for being opinionated: If I decide I hate a market or technology area, that becomes a reason to reject deals, and makes me more efficient. Maybe I was wrong? It won't be obvious for 3 or more years, and investors seldom see what deals a fund drops. There are NO penalty functions for false negatives at this stage, unless you never do any deals at all! Unlike many VCs, we list some of the things we dislike right on our website, hoping that people won't bother us with them - not that it seems to stop anyone.

Suppose I do like the first sniff of the deal, and have them in for a meeting, and then back for a meeting with the rest of the partners, and then do a little diligence on the people and competition. Just like a pharma company working up a drug for market, my costs go up at each step. I will spend lots more of my time, hire a diligence consultant to check up technology details or a foreign market, burn up favors at other firms asking for their opinions or referrals. And suppose I do get ready to call for a vote on the investment from my partners. What is my penalty case then? Why it's once again a false positive - put your investor's money into a turkey and the penalties are big. False negative, once again little cost. The only way you'll even know is if someone else funds them and the deal wins - then you'll laugh about it over drinks five years in the future.

You can see how this works with (1) above to make it a brutal proposition. Don't have an 'in network' connection? Unless there's something eyecatching about the plan, and no obvious problems, it won't make it through the first stage of the process above. Though most every VC firm has a website and other ways of sending in cold call plans, it's axiomatic that plans from the 'slush pile' are crap, because we all know the yield is far less than personally referred plans. Since we're early stage, we may look at a larger fraction of the slush pile than the bigger firms, but I've had only one plan survive first reading in over two years of operation.

Second moral of the story: Every VC firm and every partner has different trivial reject rules, and you don't know them. They may have decided they hate a particular market or technology area. They may be ready to make an investment in a competitor. They may have decided they are overweighted in the market or technology you represent. Who knows? Remember, the prior probability of getting 'no' is 99+%, so hearing it doesn't give you much information unless the fund supplies some details (we try to be good about this, when it's something fixable by the company, but no is still no). Your only hedge on this is to keep trying, because everyone's tastes differ. If you get similar 'why' replies after four or five tries, then get a clue. Again, a network connection will get you more diagnostic data than a cold call every day.

3) And some comments on the current state of play: The VC world is still somewhat in shock. I thought I had seen the bottom in sentiment last spring, but it seems like everyone came back from summer vacation 2002 in an even lousier mood. A friend who is partner at a larger firm explains it this way, paraphrased: 'Last fall (2001) everyone understood that the market was overinvested and that we were in for some nasty triage. We merged or killed off the hopeless cases, patted the ones doing well on the back, and got to work on the ones where TLC might make a difference. This year we came back again and realized that 80% of the those cases are going to die, and we have the whole miserable thing to do over again.' I'll add that though the noise made by the 'dot-com' implosion was large and noticable in the popular press, since many companies were publicly visible, there is more capital being written off in busted switch, router, and optical component companies, because end customer demand in that whole area has imploded. The net of all this is a VC business that is emotionally biased toward the downside. And we honestly don't know where the real capital demands of information technology innovation are going to stabilize. We're now back to the 1995 investment run rate, and no one really knows if that's the baseline, if we have undershot, or have further to fall. (My guess is we are somewhat undershot, but I have no way to prove it.)

Just to compound the angst for early stage guys, there's also the spectre of the 'down round' or 'pay to play.' Glossing over some details, this is where a later stage investor in a company offers a much lower valuation for the company than in previous rounds of investment and/or demands that earlier investors put in more money or see their stock be converted to 'common' as well as being heavily diluted. There are a lot of such deals around now, since valuations have fallen overall, and the economic conditions have made it hard for startups to make sales even if everything else is working right. This is of course terrifying to early stage funds like ours, and even more to individual 'angel' investors. Unless we reserve enough money to keep playing in later rounds, we can lose all the value of the cash and effort we put into the earlier stages. The reaction of the smaller funds is to circle the wagons by forming funding syndicates for each company, that have enough capital to carry through to profitability even if no larger funds offer favorable terms for later funding rounds. This of course takes more time to assemble a syndicate, more chances of trivial rejects, etc., etc. If this continues indefinitely, it means the whole idea of stage specialized VCs was bogus and a figment of the bubble. I'd prefer to think not, being obviously biased, but only time will tell.

This last set of conditions feeds directly back into the two big areas above. Apparent risk level in all deals has been raised not only by the negative market for technology in general, but also by internal factors in the VC industry that are causing a lot of funds to be running more aggressive 'trivial reject' filters that reflect the kind of syndicates they believe they can assemble. With apparent risk up, the person to person trust network element of this kind of investing is an even stronger factor than usual right now.