On Friday, Demand Media filed a form S-1 with the SEC in preparation for going public. During the last week, I’ve read through it to see how much the world’s most successful SEO company makes from SEO.

There are a few big surprises; they’re less profitable than I expected, even though their accounting is more aggressive than I’d expected. But Demand Media has two major positive features. By combining their long-tail content business with their boring domain registration business, they may have created a domain registrar that can afford to offer lower prices than anyone else. And their content business could conceivably fix an economic inefficiency even larger than the one eBay solved.

eNom, Demand Media’s Registrar—The Boring Business

eNom was the first company Demand Media purchased, in early 2006. Demand’s financial data prior to that just comes from eNom, so it’s possible to get an idea of how valuable the registration business is as an independent entity.

Even better, there’s an older pure-play registrar, register.com, which was bought out in 2005, providing a starting point for a comparison multiple.

In Register’s last four quarters of operation, it generated revenue of $100mm, and operating cash flow of approximately $11.2mm. Their final purchase price was around $90mm excluding cash on hand. (Since then, it appears that Register.com hasn’t grown; they were recently purchased by Web.com, which touted a “combined company has a non-GAAP revenue run rate of approximately $180 million”—compare that to their trailing revenue of about $100mm, and it appears that Register is now doing about 20% less business than it was at the time of the acquisition).

eNom was actually less profitable than Register.com during the brief period that their financial statements overlap, but it makes up for it with strong growth. From 2005 to early 2007, annual revenue was basically flat. But from March to the end of the year in 2007, it grew at a 45% annual rate; the next year, eNom showed 61% growth. From 2008 to 2009, it grew by just 6%, but for the first half of 2010 they’re showing (unaudited) 40% growth.

If slow growth is worth .9X sales, what is 40% growth worth? It wouldn’t be surprising to see them valued at 2X to 3X sales.

But that revenue number may be conservative. Domain registrars book their revenues over the life of a registration, even though they generally get the cash upfront. So when eNom shows $X in new revenue this year, that $X will show up during each of the next few years. They don’t seem to disclose the average length of time for which their domains are registered, but one year is the minimum. This can provide an additional cushion to investors concerned about short-term accounting losses; as Godaddy’s CEO explains, “ during periods of sharp growth it is very difficult, in fact almost impossible, for us to show a profit.”

At 3X sales, Demand Media’s registrar business is worth $366mm as of this quarter.

Demand Media Studios: The Long-Tail Content Business

Demand Media was the first company to transform “long-tail” from a concept into a business. Starting with Chris Anderson’s Wired article, people have understood that the low transaction costs of online businesses mean that most of the growth comes from obscure stuff; Barnes & Noble would have to make some serious changes to fit 100% more books into their stores, while Amazon could make less significant changes to accommodate 1000% more titles.

But if Amazon’s business can take advantage of the Long Tail, the average SEO-focused business can thrive with it. One thing the original long-tail theory didn’t grasp was that specificity, on average, leads to higher engagement. And the best measure of “Engagement” is sales.

For example, consider someone trying to file their taxes late. They might start by searching for “taxes,” and find the information too generic. They’d narrow it down (and move into Long-Tail territory) with a more specific query like “File late taxes.” If that didn’t get them accurate enough results, they’d take it one step further—”file late 2008 taxes”. If they did that, they’d probably end up on the site of one of my clients. That client has ceded higher-traffic terms like “taxes” and “late taxes” to the IRS and competing tax sites. But as it turns out, specific terms are more likely to lead to a sale.

One of the strategies I used was to create content on “Article Directories,” (of which the most popular is EzineArticles.com—see my Ezine Articles SEO guide for details).

Article directories are Demand Media’s primitive ancestors. Their business model works a little like this:

1. Let people create content on whatever topics they want.

2. Edit the content to make sure it’s in acceptable English and not plagiarized (i.e. that it fits Google’s standards for minimum quality.)

3. Let writers include a link or two to their own content, so it’s worth writing.

4. Plaster the articles in ads.

5. Use every possible SEO trick to make the articles rank well on Google.

This business has great margins. As long as the cost of editing and hosting articles (probably under $4/article) is less than the revenue from ads, the business will thrive. And improvements in SEO will increase revenue for all of the articles.

Demand Media tweaked the formula. Their typical article is not “whatever topic you want”—it’s a specific subject that they’ve picked out based on search traffic data. In exchange for determining the article topic, Demand Media pays writers, often $15 per article.

It doesn’t take much for this to be an effective strategy. As long as:

1. Demand Media has good data on what’s being searched;

2. They have a large operation—for any topic they come up with, they can find a writer;

3. They can maximize the number of clicks on ads, versus on links; and

4. Demand Media has superior SEO: if an article directory is going to rank for a term, it will be their article directory;

Then they’ll have a competitive advantage over everyone else in the industry. And at that point, the only thing that limits their profits is the number of commercially-viable long-tail searches.

Do they have this advantage?

That’s a tougher question. But they’ve made some intelligent moves. As mentioned before, their domain registration business gives them a unique source of data and traffic. As long as they’re already in the business of monetizing pageviews based on traffic data, that’s going to provide them with extra profits.

And they’ve purchased some great domain names. Google has a policy of giving extra weight to established sites; if you’ve been online for a decade, you’re not going to throw away your reputation, so Google can afford to rank you better. eHow was registered in 1998. Cracked was registered in 1997. Trails.com was registered in 1999. You can only beat that by paying for it, and it’s tough to find cheap domain names that were registered that long ago. (It wasn’t so expensive a few years ago, when Demand Media was buying.)

They’ve also picked up some domains that will naturally earn links, like Livestrong.com, which was brilliant. They took goodwill for Lance Armstrong, as expressed by links, and turned it into a high-ranking health website.

But Demand Media’s real advantage may be their article quality. They’ve set prices at a point that makes it uneconomic to write a lengthy, well-researched article; better to write something quick that covers the topic without ultimately answering the question. And that is the perfect way to create an article that is less compelling than the ads. Right now, I can see the Demand Media is offering $16.00 to write “How to Build an Acorn Skiff”. I could write 500 words on the subject, but if someone read what I had to say, and saw ads for an “Acorn Skiff Kit,” or “Acorn Skiff Instructions,” they’d probably opt for the ad.

Demand has revealed their return on new articles, and the numbers are stunning. According to their own metrics:

We base our capital allocation decisions primarily on our analysis of a predicted internal rate of return and have generally observed favorable historical returns on content. For example, our article content published on eHow in the third quarter of 2008, or Q308 cohort, generated a 58% internal rate of return. This internal rate of return measure does not account for any revenue after June 30, 2010, although we anticipate that our Q308 cohort will continue to generate revenue for the foreseeable future and therefore achieve a higher internal rate of return. For example, article content produced in the Q308 cohort achieved 62% revenue growth in the second quarter of 2010 as compared to the second quarter of 2009.

Oddly enough, Demand is almost keeping two sets of books. While their IRR is calculated as:

the discount rate that, when applied to the advertising revenue, less certain direct ongoing costs, generated from the cohort over a period of time, produces an amount equal to the initial investment in that cohort.

But:

[W]e have paid substantially all of our freelance content creators upon the creation of text articles and videos, rather than on a revenue share basis, and we capitalize these payments.

In other words, the reported results for their content strategy bear only a tangential relationship to how profitable it is. Their return on new content is ridiculously high, but they report the upfront cost over several years.

And it’s growing. As of their S-1 filing, they generated “5,700 text articles and videos,” per day. Compare that to “the 4,000 videos and articles that Demand Media publishe[d] every day” as of last October. Later in their report, they reveal the “RPM” (revenue per thousand pageviews) for their content business over the last few years. In 2008, that number averaged $10.56; for the first six months of 2010, it was $11.81. Compared to the first half of 2009, total page views are up 23% (to 3.9 billion). Meanwhile, they have 2.2mm pieces of content (videos and articles). If that number is growing at 5,800 per day, it’s growing at approximately 24% per quarter.

This could be a sign that their ability to find profitable long-tail content to create is diminishing. (At some point, we’ll run out—or at least get to the point where the amount of content available expands based on new searches, not existing searches that are underexploited.)

Assuming the 24% quarterly growth number is correct, their annual growth in total content is around 136%, compared to their year over year pageview growth of 23%.

For their content business to generate continued growth, they need to do two things: they need continued RPM growth to counter the fact that they can’t buy as many pageviews as they once could. And they need to exploit higher-quality content.

With annualized revenue of $130mm+, annual growth of 41%, and an IRR of 58% (and declining, but perhaps gracefully), Demand Media’s current content business is probably worth at least 5X current sales, or $650mm.

Demand Media and eBay: Is an Attic Full of Collectibles like a Stay-at-Home Mom with a Masters Degree?

Demand Media is a puzzle, but you can put the pieces into place by examining what economic inefficiencies they exploit. Their name emphasizes the first one—they create content based on demand, which they can determine algorithmically. But an equally big story may be their supply.

Demand Media’s content business has solved a huge labor market inefficiency: there are millions of people who can write reasonably well, but who can’t efficiently get a full-time job. Think of students, housewives, retirees—and ignore anyone below the 90th percentile. It’s a gigantic population of people who are unlikely to get full-time jobs, but who can’t effectively spend their time doing piecework or freelance writing.

Thanks to Demand Media, they can do incremental writing work. If they can’t work full-time and it’s hard to judge them part-time, their labor is massively undervalued; in fact, if you look at the rise of blogging, tweeting, Wiki-ing, Q&A sites, etc., you could assume that the median price of their time is zero. By paying more than zero, at a huge scale, Demand Media may be able to buy and judge more hours of decent writing than any other company.

Treat those underused hours like assets, and you can see Demand Media as an opportunity like eBay. What propelled eBay’s growth was that a huge number of households owned knicknacks and antiques that they couldn’t cost-effectively sell. eBay created an efficient market in those products, and in just over a decade they cleaned out a couple million attics and garages, at a profit. Now that the big opportunity is gone, their growth has slowed down, but they’ve still established a wildly profitable business.

Demand Media has a similar opportunity—at a scale at least an order of magnitude larger. What’s the book value of all the antiques in the US? Something in the tens of billions sounds reasonable; something in the low hundreds of billions might be the upper limit. eBay earns commissions on those sales, with a maximum of fifteen percent.

Now, compare that to the value of the potential labor everyone in the 90th percentile of writing or movie-making skill, who doesn’t have or doesn’t want a full-time job, but would like some extra income. Figure $30/hr times two hours per week times population times (1 – labor force participation rate) times 10%, and you get a total value of $29 billion per year. That’s the amount of money left on the table if the top 10% of the 30% of people in the US who are not members of the labor force could have each written four articles for eHow.

But that’s just the price. If Demand Media earns even a small positive return on those otherwise unused hours of labor, the value could go up even more.

It gets better. Demand can afford to target the top 10%, but what about the top 1% or the top .1%? Once they have a system for ranking everyone’s quality, they can also start selling off premium talent at a premium price.

And that’s exactly what they’re doing. Many of Demand Media’s newer deals aim for a higher price point: premium content on SFGate.com and NFL.com, and $80 homepage HowTos for eHow.

As more of their writers are making (and earning!) $80 for a good article instead of $15 for a mediocre one, they’ll be able to attract better writers. At their current growth rate, there’s a good chance that they will be able to pay a more accurate market price than anyone for writing talent. We could see an online world in which people are divided into a) brand-name authors, most of whom make basically nothing, and a few of whom have great name recognition, and b) Demand Media authors, whose skill and value are efficiently quantified, and who make exactly what they’re worth.

That’s the big opportunity for Demand Media. Every day, talented people are wasting their time. And valuable content that could be written doesn’t get written. Matching these two parties is a many billion-dollar opportunity, and (at least in the short term), Demand can capture the majority of the value from arbitraging it.

That’s the upside. A $30 billion market, coming out of nowhere, where Demand Media has a de facto monopoly.

If there’s a 5% chance of creating this market, and Demand Media ends up with a 50% share of a $30bn market, with a 15% profit margin (low for a dominant content company; Google gets close to double that), they’d have a 1% chance at creating a business with $2.25bn in annual profits. At 15X earnings, that’s worth $33.75bn. Multiply by the 5% odds, and you get $1.69bn. Assume it takes ten years, and discount that to the present at 8%, and this opportunity is worth $782mm.

What is Demand Media Worth?

That’s a tough one. According to the Financial Times:

People familiar with the offering estimated it would value the company at about $1.5bn and would be priced by November.

Based on some tricky estimates, vague comparisons, and wild guesses:

• The registrar is worth $366mm.

• The current content business is wroth $650mm.

• The small chance of a hugely successful content business is worth $782mm.

• And they have cash on hand of $33mm.

At those prices, the total value of the company is $1.83bn.

Ten Questions for Demand Media’s Management

If you’re thinking of investing in Demand Media, or you get a chance to participate in their IPO roadshow, here are some unanswered questions that could firm up their valuation.

1. Why equity? If investing in new long-tail content produces such a high IRR, with such low marginal cash requirements, why not issue bonds instead?

2. Why is capitalized content amortized over five years, and not based on lifetime expected traffic?

3. How fast is the total amount of long-tail traffic growing? When will Demand Media saturate the immediate market?

4. What were the circumstances behind buying The Daily Plate from Demand Media employees—and are other employees working on similar projects?

5. Does the RPM for content vary based on how it’s found? Is there more growth in low-RPM or high-RPM traffic sources?

6. What is the IRR for Demand Media’s different content types—long articles, short answers, and video content?

7. Does Google view eHow, in its current form, as a problem?

8. What aspects of Demand Media’s niche-targeting strategy are not replicable?

9. What IRR do competitors get on content they create?

10. Does Demand Media expect to compete by paying more for content, earning less on ads, or both?

Demand Media is the most exciting IPO since Google. It’s a transformative business—for anyone who works online, sells online, or could do valuable work with a more efficient labor market. For many people, it’s disruptive; I expect Demand Media to reduce the demand for the work I do, and to outbid me for the people I’d like to hire. But overall, the effect is positive. Ultimately, Demand Media is living up to their name—they’re matching supply and demand, and they could make a fortune from it.

Full Disclosure: This is complicated. I have written content for Demand Media, and gotten paid for it, but not enough to affect my lifestyle. I also compete with Demand Media, in the sense that we’re both often trying to create content that ranks for similar terms. And since I’ve used ads on Google’s Content Network, I’ve also contributed to Demand Media’s revenue.

The Bubble and the Blowout Preventer

But an increase in talent acquisitions will directly lead to an increase in the amount of money angel investors can afford to invest. And by offering a more efficient market for top startup talent, this increase in available capital will be absorbed by qualified new founders. There may be a risk of a bubble, but untapped talent is a blowout preventer.

Talent acquisitions lead angels to invest more because illiquidity is one of the big reasons not to invest more. If the average startup takes ten years to show any return, then a prudent angel investor should, on average, invest 10% of his or her Angel Bankroll in any given year. There’s no sense in being fully invested by year five, for example, and passing on investments (or allocating more money to an illiquid asset class) for half a decade.

But that ten-year average can’t last long in the face of a few year-one and year-two acquisitions. The dollar amounts are low, of course, but that’s compared to low valuations. Talent acquisitions almost always happen before the bigger VC rounds, when the founders own lots of stock and the number of potentially disgruntled investors is minimal.

And if the turnaround time drops from 10 years to 8 years, the optimal amount to invest each year rises from 10% of bankroll to 12.5%—raising the total amount of money available for angel investments by 25%!

Are Talent Acquisitions Getting More Popular?

It’s hard to say. They’re less likely to be reported than big acquisitions. And it’s easier to say “It was a talent acquisition,” than “We got basically nothing, for a business worth less than zero.” But there are signs:

• Facebook’s CTO got there through Friendfeed

• Today, LinkedIn bought a company for their “Recommendation Technology, Team,” which is at least halfway there.

• There are even questions about whether Google paid $182 million for Slide in order to get Max Levchin and his team.

• Facebook’s Hot Potato acquisition has also been branded a talent acquisition.

That’s enough for a trend story, but not enough to declare a trend. But there are good reasons for talent acquisitions to pick up. With more API- and platform-based companies, it’s easier to see another firm’s work up close. And the startup scene is simply better-networked than it was in the past. Meetups, retweets, LinkedIn recommendations—it’s hard not to hear about interesting new companies when their biggest asset is still their teams.

The Result?

Here’s the most optimistic scenario: more angel investing could lead directly and proportionately to more wealth creation. According to Paul Graham:

Good investors are rare, even in Silicon Valley. There probably aren’t more than a couple hundred serious angels in the whole Valley, and yet they’re probably the single most important ingredient in making the Valley what it is. Angels are the limiting reagent in startup formation.

If those good investors can do a little more investing, that’s positive.

Even better, this trend will push talented people to start more companies. Joshua Schachter doesn’t have to stick around at Google when he can do his own thing. Worst-case scenario, he’ll be back at Google (or Yahoo!, or Facebook, or Twitter) in a couple years, with a couple million dollars for his (and his investors’) trouble.

Actually, I know designers who will get basically 0%, a few who will get 20-50%, and a few who will get 100%. The 0%ers are the ones who work for someone else’s company; the 20%+ crowd started their own. In this industry, that looks like the simplest way to move along the continuum: if you’re at an established company, you’ll be lucky to capture $5,000 in extra cash from your million-dollar ideas. If you start your own company, the extra profits translate directly into higher value for your equity.

This used to create a troubling cycle among top talent: people would join a good company as an audition, but the more talented and ambitious they were, the better off they’d be starting a competitor (or at least joining one in the early stages). Hiring talented people was a great way to subsidize your next competitor.

That’s a nasty bug, but there’s a surprising solution. As more companies create APIs, platforms, and other ways to build something on top of an existing business, they create a middle ground between doing your own thing and working for someone else. If Twitter is missing a feature, you don’t have to build a better Twitter; you can make Twitter for stock traders—on Twitter!—or Facebook for casual games—on Facebook!

Stocktwits and Zynga both left money on the table by doing it that way. A Zynga as big as today’s Zynga, without the Facebook platform, might be earning hundreds of millions of dollars more in revenue over the next few years. Of course, Zynga without Facebook is inconceivable.

This creates a surprising new set of incentives for successful web companies. They can afford to care more about career upside; their most aggressive, ambitious employees might still quit, but they’re more likely to quit in order to build something using the APIs of the company they left (see Quora, founded by Facebook alumni—which used Facebook accounts to authenticate identity and market itself). Mark Zuckerberg can’t be happy that Adam D’Angelo and Charlie Cheever quit Facebook to start Quora, but at least the path of least resistance was to leave Facebook in order to build a site that made Facebook a little more useful.

“Not only that, but it’s an exceptionally long-lived asset. Once you own it, you’ll be getting dividends for your entire working life.

“While it’s not as cheap as it used to be—in fact, it’s going up in price at about twice the rate of inflation—it’s never been easier to get government-subsidized loans to purchase it. In fact, third parties may pay for some or all of it for you!

“The asset is, of course, a college education. Now, wouldn’t you like to review the prospectus?”

If you’re an average American, an undergraduate degree is the second biggest purchase you’ll make, after a home. It’s well-known that a college degree is an investment with a positive return. Most successful people are college graduates, and most desirable career tracks require at least a college degree.

But I contend that college is not a good investment. It’s a bubble. Why?

1. The price of college is continuing to rise faster than the rate of inflation.

2. There is empirical evidence that the dollar value of a college degree is flat—or possibly declining.

3. If college loses its signaling power, this will make past degrees retrospectively worth even less.

What is the Return on Investment for a College Degree?

I’d like to preface this by noting that I don’t think all degrees are a bad investment. A technical degree at a decent school is almost always worthwhile. It’s very hard to do a decent job at a top-tier school and not end up with a degree worth having. That said, most of the growth in the college population comes from:

1. Lower-tier schools expanding their classes. (Arizona State has grown from 26,000 to 68,000 students since 1970. Yale’s undergrad class grew from about 5,000 to 5,300 in that period.)

2. New schools. The University of Phoenix was not exactly a major force a few decades ago. Meanwhile, schools rarely shut down.

Actual return data are hard to come by, but one source is to look at the difference in earnings between college graduates and non-graduates.

This data has an interesting feature: it’s based on average earnings over a ten-year period. That has two effects: first, to the extent that college degrees are for signalling, rather than for skills, it will exaggerate their effect; if you’re a smart engineer with no degree, it might take you a couple years to get the job you would have gotten out of school, but you’ll get it eventually. The more pernicious effect is that if college degrees are suddenly worth less, it will take a while for this to be reflected in the average. In a worst-case scenario, where the implied value of a college degree spikes and then plummets (the way technology stocks did in 1999-2002, or housing did in 2000-2009), this effect won’t show up in median earnings.

The government data table is missing two columns: first, the average difference between college grads and high school grads. And second, the growth rate of that difference. Add them, and the result is striking:

. Year Median Wage HS Wage Grad wage Difference Annual growth rate . 1980 $46700 $44200 $52300 $8100 0 . 1985 44000 40000 54800 14800 12.81% . 1990 41200 36300 52300 16000 1.57% . 1995 38900 33900 52700 18800 3.28% . 2000 42500 36300 57500 21200 2.43% . 2005 38600 33100 55100 22000 0.74% . 2008 40000 32000 55000 23000 1.49%

College was a great investment in the early 80's (when, incidentally, parents who are paying for college now probably formed their picture of the value of a degree). But now the cost of a degree is rising at twice the rate of inflation—while the return on that investment is rising by half the rate of inflation.

What is a "Degree," Anyway?

If tough degrees from good schools aren't numerous enough to match the demand for college education, that demand will be expressed elsewhere. As mentioned before, this can show up in the enrollment growth at lower-tier schools. It can also show up:

1. In the growth of lower-income, less demanding degrees. According to the National Center for Education Statistics, the number of Engineering degrees granted grew by 53% from 1970 to 2008. The growth in psychology degrees: 142%. (Math and statistics, by the way, has declined 38% during this period, though it used to be a proxy for computer science, too.)

2. In the growth of for-profit schools.

All of these factors change the definition of a "degree," and not in a positive way. According to the same NCES data, Engineers made around $58,300 a year in 2001; psychology majors clocked in at $35,100. 87% of engineering grads were employed full-time; 76% of psychology grads were. For-profit schools don't release numbers on their students' incomes, but from the résumés I've seen as a recruiter, the numbers are not likely to be impressive.

In other words, the growth in the number of people who have a degree has coincided with a decline in the average value of those degrees. The class of 1970 was comparatively more likely to go to an Ivy-league school, major in math or engineering, and get a job. The class of 2011, God help them, is more likely to go to a state or for-profit school, graduate with a degree that has a lower expected salary, and have trouble finding employment. And all else being equal, more subsidies for higher education will imply more of all this—it is vastly cheaper to get another sociology major into Capella University than to get another Astrophysics major into Princeton.

Popping the Higher Ed Bubble

The bubble in higher education is fueled by culture and subsidies. Culturally, college fits in about where housing did until very recently: it is simply something you do. If you've succeeded in life, you've gone to college and paid for your kids to do the same. It's hard to change that perception; if Bill Gates didn't, Mark Zuckerberg won't.

But what we can change are the policies that subsidize college. Student loans should be drastically curtailed, especially for the "at-risk" groups (which are also the fastest-growing). Loans to students at for-profit schools have an appalling 40% default rate. While I can't find data for degrees, I suspect that hard science graduates default more rarely than social science graduates (except for education, where the compensation is nothing if not reliable).

This is a bubble like any other bad bubble (i.e. a credit bubble). It ultimately rests on the twin mistakes of extrapolating based on bad data, and assuming that investments will behave the same way when they're made en masse and by default, rather than individually and with great care.

The solution is to pay attention. There shouldn't be a "higher education" bubble, because higher education shouldn't be treated as a single aggregate entity. As long as it is, we'll have a surplus of degrees from the University of Phoenix—just as the last bubble gave us a surplus of suburbs outside of, well, Phoenix.

Full Disclosure: Depending on how you calculate this, I own a little more than half of a degree in economics or math. Basically all of my friends have degrees, including postgraduate degrees. Everyone I've ever hired or been hired by has a degree. I haven't found a way to short higher education as a whole. If you can think of one, let me know.

This works pretty well for some companies. Facebook, for example, hit 100 million users in mid 2008, with “more than 600″ employees. Now they have 500 million users, and 1400+ employees. 400% growth in users; 130% growth in employees. That’s exactly the kind of math VCs like to see, and it’s why they were willing to fund Facebook generously in the early stages. Facebook will run into scalability barriers after a while. At some point, the limiting factor is not engineers (a scalable resource) but customer service reps and servers.

For many of the current crop of hot companies, the end of scalability is coming much faster.

Long-Tail = Labor-Intensive

Google wants to organize the world’s information, period. Demand Media wants to create every single piece of content for which there’s demand. Enumerated.

Google’s technique involves aggressive automation: spidering, creating indices, weeding out link farms and spam, and automatically analyzing content. When humans interact with that process, it’s generally at the margin: tweaking an algorithm here, fiddling with backend infrastructure there.

Demand Media requires consistent human interaction at every level. They do have an algorithm, but it spits out data that need digesting: title editors turn search query snippets into headlines, then content writers create them, then editors review them. If Demand Media wants to grow (and of course they do!), they need to crank up their inputs.

It’s the same for any other business that wants to hand-craft content for a niche audience. It takes continuous input to get continuous growth. And even increasing the growth rate is labor-intensive; Demand Media screens every applicant, and you can find plenty of stories of journalism majors getting rejected.

It might be a fantastic business, but it doesn’t scale the way the last generation did. No wonder they’ve raised $355 million. The real question is: why are they raising equity? They have immediate, predictable cash flow; they might as well replace their operating leverage with the old-fashioned kind.

Local Marketing = Telemarketing

If you’re going to talk about massive money raising, you’ll have to talk about Groupon. $173 million, including a big check from Digital Sky Technologies. They’re also in an established space, with lots of competition, including companies that have recently gone public. And why did Reach Local go public? So they could hire more sales reps!

A surprising fraction of my friends work for companies that do group buying or location-based marketing. And there’s a reason for that: once you have your platform and your userbase, the big obstacle is signing up new small business customers. And so far, nobody has found a better way to do that than hiring lots of salespeople and getting them on the phone.

Even Foursquare is growing into that situation. They’ve done some big, showy deals, but I bet the main value they’ll see from them is that they can call up the local coffee shop and say “Wouldn’t you like some of those customers who order the $5 Venti at Starbucks?”

The VC Overhang: Cause or Effect?

It can’t be a coincidence that VCs have raised too much money since the late 90′s, and there are now startup-looking businesses that can (and must) deploy huge amounts of capital.

It’s hard to say which is cause and which is effect, though. Groupon’s founders stumbled into their business—no cynical ploys there. I wouldn’t put it past Demand Media’s CEO, Richard Rosenblatt, to have planned this out (he knows how to sell high).

But my best guess is that it’s more complicated, because you can reframe the question: what these companies have in common is that the limiting factor for their growth isn’t technology, or brand recognition, or network effects. It’s capital. And for once, at least in part of the economy, capital is overabundant.

Google is the Big Default. If you want to find something, but you’re not precisely sure what, Google is where you start. For about eight years, that’s where I’ve started, too. But recently, two sites have started to replace Google. And what’s especially dangerous about them is that they’re both encroaching on Google, starting at opposite ends of the spectrum of services that Google Search provides.

DuckDuckGo: The Best New Search Engine Since Google

The first Google Threat I found is the DuckDuckGo search engine. It’s a one-person company, so it’s entitled to some quirks. But it’s also entitled to strong opinions on privacy, and that low overhead means they can be hostile to ads that are hostile to users.

DuckDuckGo’s founder has some complex ideas about what Google can’t copy, but the summary is simple: Google is at, or close to, a local maximum. Any change they make will, on average, make the typical user’s experience worse. If there’s a better way to search than Google, there are many worse ways to search between that and Google.

Which makes DuckDuckGo a great business: if they Get Search Wrong and only have a small core of loyal users, they can at least sell ads against those pageviews. But if they get it right, Google can’t afford not to buy them—why not marry Google’s Adwords team (and all their search data) to whatever search interface happens to work best?

What I’ve found is that for a specific set of queries, DuckDuckGo wins. Specifically, if you are sure that what you’re looking for exists, DDG will feature it more prominently. They hide ads, let you downgrade commercial sites, and highlight official sources. Basically, they’re a better user interface for a better Wikipedia.

Quora: The Longest Possible Long Tail

On the other end of the spectrum, there are some things you would not expect to exist online. For those, your best bet is Quora, a smart Q&A site. They were invite-only until literally the day after I sent a bunch of people invites, so their userbase consists of people who are a few degrees of separation away from the founders. And since Quora’s founders came straight out of Facebook, that’s a smart group.

Quora is full of people who can answer the questions Google can’t. Questions like:

• What are some behaviors that are accepted now but might be considered immoral in the future?

• How many times a year does an air conditioner fall out of a window in New York City and kill someone?

• For what purposes are fictitious social networking profiles created?

I wouldn’t bother asking Google. Quora has answers. Quora is targeting the longest possible long tail: pieces of content that at least one person can create, and that at least one person cares to read.

Quora doesn’t, however, have a business model (they have enough funding not to care). DuckDuckGo may or may not be profitable, but probably isn’t a huge business.

But they’ve both replaced Google as a default. And they did it for less than it would cost Google to take that default status back.

At 21X earnings, Google is not a short. There’s a lot they can do to diversify, and a lot they can do to get closer to DuckDuckGo and Quora. But all that will cost them something, so buying Google now is a tricky bet—that they’ll be able to change direction in time, before they end up being another Yahoo.

Take a look!

There’s a good reason most people choose to “Reply All”: all of the recipients of an email have to assume that, until they hear something about it, whatever the email says must be done still must be done. If you’ve ever replied directly to the sender of an email that was sent to ten people, you’ve gotten one of two responses: either ten minutes later you’re “Reply All”‘d on another email that makes yours redundant. At one minute per email times ten recipients, it’s easy to see how a simple task can take an hour or more total—and that’s ignoring the cost of disruptions.

I have a simple solution: “Reply All” should not allow you to compose an email reply; it should send a default answer like “It’s being taken care of.” To recipients who need to know more, you can elaborate; to everyone else, well, it’s being taken care of.

(In the meantime, you can start replying-all with that line. Hopefully it will catch on.)

• In an equity bubble, investors have limitless optimism about the future. They expect many of the companies they invest in to fail, but believe that the 95th- or 99th-percentile performers will more than make up for this.

• In a credit bubble, investors have limitless faith in the status quo. They expect volatility to decrease, and they believe they can estimate returns with increasing accuracy. If they want higher returns, they know they can use leverage—but for the most part, investors celebrate the middle of the bell curve, and expect the tails to cancel each other out.

The only thing these bubbles have in common is that they involve collectively bad decisions with money. The kinds of decisions are vastly different, and the consequences are, too.

Equity Bubbles Considered Not-So-Harmful

In an equity bubble, resources get distributed towards whatever investment is the most levered to optimistic outcomes. These don’t have to be the same outcomes—AOL’s walled-garden got a high valuation at the same time that Yahoo’s anything-goes approach did, for example.

Because of this, they attract capital (and talented people) to anything sufficiently unusual that fits into the bubble narrative. This entails lots of investments in bad ideas (Pets.com) and lots of investments in necessary ideas (online brokerages, which automated the most automatable part of investing).

It’s a huge subsidy for creative destruction, with an emphasis on the creativity. Unlike the usual economic process of local optimization (making a factory marginally more efficient, reducing inventories a little bit at a time, finding a new customers), it represents a massive change in the process (creating a completely new sales channel, revamping the entire inventory management process, creating a financing model that’s based on venture funding plus a negative cash cycle).

There’s a lot of malinvestment during these bubbles, but they also lead to lasting companies. They also create new infrastructure (bad telecom investments in the 90′s made Myspace and Youtube possible in the 00′s). And they help illustrate the ineffectiveness of older companies, though it’s often not until the bust that these companies go under.

Of all of the ways that the market can overshoot itself, equity bubbles appear to be the most beneficial. In the short term, they subsidize creativity; in the long term, they subsidize creativity tempered by sanity. And if “creativity tempered by sanity” doesn’t describe the best parts of the world economy since the industrial revolution, I don’t know what does.

Credit Bubbles: Doubling Down on the Status Quo

Credit bubbles are toxic. They rely on ignoring possibilities, rather than embracing them. They allocate human capital away from analyzing uncertainty, and towards creating artificial certainty. Credit bubbles are more self-perpetuating than other kinds of bubbles, and they leave behind an inflexible economy.

A credit bubble is a bet on the continuation of the status quo. Since lenders have limited upside and lots of downside, they don’t have any reason to care about best-case scenarios; instead, they look at average scenarios. That requires historical data, which can best be obtained from older and more static industries.

The most pernicious effect of credit bubbles is that they tend to be self-perpetuating. In the 00′s, the dominant investment thesis might be summed up as follows:

• Our economy is stable, financial markets are liquid, and information is freely available, so volatility should gradually decrease.
• This liquidity and information also means we won’t have huge economic dislocations; future change will be gradual.
• We can use historical data to model prices, and our models keep improving.
• Thus, the way to make money is to buy what you can most effectively model, treat it as an asset class rather than a claim on specific cash flows, and lever as much as your volatility permits—after all, you have historical data on exactly how markets will behave under any circumstances.

This results in smooth returns with low (and declining) volatility for people who buy into the thesis. And if they’re competing with other investors for funds, that means more money gets allocated to investors who buy into the bubble thesis—pushing prices further in the same direction.

All else being equal, skeptics get priced out. During the mortgage bubble, the skeptics who profited the most were the ones for whom all else was not equal—value investors who got into real estate as a sideline (like Michael Burry and John Paulson), or unusually talented traders who subsidized their negative bets with their market-making profits (like Greg Lippmann).

Meanwhile, the bubble believers see their ideas constantly confirmed. Everyone’s returns are in line with predictions—unlike equity bubbles, there aren’t any sudden gains. In a credit bubble, it’s possible to do what Bear Stearns’ internal hedge funds did, and show consistent returns of about 1% per month with no down months.

All this means more resources allocated towards precision, rather than accuracy: if prices converge on what the model says they should be, a more refined model will create more incremental profits than a different understanding of what’s being modeled. Over time, this simplification leaves room for the underlying assets to change: the capital for mortgages from 2006 was allocated based on the returns from years beforehand, which made that capital less sensitive to declining credit standards.

In the terminal phases of a credit bubble, investors are precisely wrong—whatever they’re most bullish on and most certain about is bound to be the asset declining in quality the fastest, and the one whose future pricing is thus the least predictable.

What to Do About a (Bad) Bubble

Investors in the 90′s were overconfident about the power of technology to change their lives in the near term. They were also overconfident about the power of financial models to predict sovereign debt prices and equity volatility—in the latter case, the collapse of LTCM scared them straight, while the Fed’s subsequent liquidity injections kept that collapse from flattening the broader economy.

This illustrates the fundamental problem: given enough liqudity, good and bad bubbles can last indefinitely. And identifying them can be tricky: a large component of the “tech” bubble was the “IT bubble”&madsh;a huge increase in the demand for people with otherwise obsolete skills, who could update systems ahead of Y2K.

You can ask yourself a simple question: “Is what these people are throwing money at, fundamentally, boring?” If it is, buy puts: you know they’ll be cheap.

If you’re a money manager, you’re out of luck: the more effectively you short the bubble, the worse your returns will look. And since credit bubbles appear quite conservative (they are, after all, based entirely on historical precedent), you’ll look bad and imprudent. Worst of all, if you bet on the bubble as a manager and short it (or even deride it) as a private investor, you’ll probably end up in court.

So your best bet is to quit your job and start a blog or something. (Then, buy puts.)

In the case of the US, India, or China, that tax rate is determined by all sorts of lobbying, voting, personalities, and compromises. In Facebook’s case, they can collect taxes by forcing companies to use Facebook Credits. It’s just a question of maximizing profits. And that means taxing companies that profit from Facebook at the peak of a slightly modified Laffer Curve.

The Laffer Curve normally describes government revenues compared to marginal tax rates. It’s a curve because, given a sufficiently high tax rate, the incentive to work, report income, or live in a particular country declines. This leads to the argument that lower taxes are net beneficial: if we’re past the peak, they’ll raise government revenue, but even if we’re not, they can produce economic growth.

Facebook doesn’t have to care about “economic growth.” Their goal is not to grow the companies that exist on their platform, just to profit as much as possible from them. It might seem that this implies taxing at the Laffer maximum, but that’s not quite true: Facebook can afford to maximize the net present value of their future income, and to set their tax rate accordingly.

In practice, that will mean low (or zero) taxes as long as a company using Facebook’s platform is still growing. But once their growth slows, it shifts the Laffer curve to the point that taxing them more produces extra income even if it slows down their growth (and discourages other companies from using Facebook’s platform).

Since Facebook can easily prevent tax evasion, and since corporations have a fiduciary duty not to be lazy, the only mechanism by which the Laffer Curve can operate is emigration: at some point, Zynga could theoretically decide that Facebook isn’t worth it; they’ll take their bigger slice of a smaller pie elsewhere. I’m not sure what it costs Zynga to move one person from Facebook to a non-Facebook platform for the same game, but it’s probably a painfully high number when you multiply it by, say, the 82 million active users of Farmville. This might also explain why Facebook is okay with stickiness—it’s bad for business, but good for keeping existing business from going elsewhere.

Taxation or Nationalization?

Fred Wilson argues that Twitter can be free to ‘fill in the holes’ in their experience, but not to go after new verticals. In the Laffer context, the question is not which features a platform will duplicate and which it won’t, but which features it will tax and which it will expropriate instead.

Facebook, and more recently Twitter, have also done a form of “nationalization”: they turn someone else’s product into a feature of their own product. This seems to work best when the product works as a feature, and doesn’t have a clear monetary value: the classic example being Internet Explorer and Netscape Navigator; soon after IE came out, selling a computer and charging for the browser would have made as much sense as selling a car and charging for the steering wheel.

The “features they should have had” versus “value-added extras” dichotomy is false. The real question is which is worth more: taxing a platform user at the Laffer maximum, or subsuming them in order to shift the Laffer curve in a more profitable direction. At first, this sounds like an academic distinction—but it implies that the more people make money with Facebook apps, the more desirable it is to nationalize rather than tax other apps: Zynga’s $300 million in revenue is worth taxing; but if Zynga produced $1 billion in revenue, it could easily be worth giving up the chance to tax other companies, in order to make it easier to tax that billion dollars at a higher rate.

Twitter isn’t monetizing in the same way, but they do face the same equation. In their case, they’ve nationalized more, but it’s easy to see where they could start taxing instead. If you’ve looked at many financial sites that monetize through ad networks, you’ve seen who can bid the most for finance-related ad space: penny-stock promoters. That’s why Stocktwits has been so successful: it culls the spam. If Twitter sells ad space to the highest bidder, Stocktwits will get spam, and it’ll be official spam, too. A nasty outcome, but possible (though Stocktwits has mitigated this risk by building their own non-Twitter income sources, and their own independent platform; more than anyone else, they’ve figured out how to launch from someone’s platform, without being anchored to it.)

Why To Stay Bullish

Thanks to the power of Metcalfe’s law, squared, platforms have an incentive to encourage new applications, and to make them rewarding. Like a country that encourages people to start small businesses, these platforms will get more activity without having to try very hard.

So for small startups, this is irrelevant: they don’t have enough of an income for it to be worth Facebook’s time to determine the optimal tax rate, and they’ll produce more taxable income if they grow fast. Once a company hits the saturation point, though, there’s little reason for Facebook to let them earn excess profits. Sure, it will discourage VCs from funding some companies on Facebook’s platform, since they won’t be able to participate in quite so giant an IPO. On the other hand, the independent programmers who put together a simple app over a couple weekends won’t be too discouraged; Facebook is likely to tax only when it’s already fantastically profitable.

And even companies like Zynga won’t be left behind. There’s still no cheaper source for new users. Converting a small fraction of them from one platform to another is still easier than trying to get them to start out on the Zynga platform itself. Meanwhile, it makes a $25 billion market cap sound a bit more believable.