Thursday, December 31, 2009

Musing on luck & Thank You

I was visiting a venture capitalist friend of mine a few months ago, talking about what we thought the future held and what kinds of companies intrigued us. He made a comment regarding a company I liked and he didn't. He said "I only invest in companies I know will be successful." On further questioning he admitted that not all of his investments had been successful: sometimes management had screwed up, sometimes customers did not get the value of the product quickly enough, etc. But in all cases, he felt he had invested in surefire plans; if the company failed, it was due to execution mistakes. And sometimes, even, he had made mistakes. But it was not that the validity of the business plan wasn't knowable, or even that he couldn't know--sometimes it was just that mistakes were made.

This is also how I thought, a few years ago: that I could know. The idea that we can always find answers and act on them and thus change our environment or the world not only underlies our technological and business culture but is explicitly taught in our best schools and finest corporations.

A few years ago, someone I worked with teased me because I never said "I don't know." Ever. I would say "I'll get back to you on that" or I would attempt to answer even though I didn't have the answer, or I would deny that the question had significance. Occasionally I would aggressively change the subject or grow angry with the person who wouldn't agree with what seemed so clear to me but that I couldn't explain. This annoying tendency of mine came from years of feedback in the classrooms and management meetings and boardrooms where "I don't know" was not an acceptable answer. Not even close. People who said "I don't know" did not get the good grades, they did not get the promotions, they did not get the bonuses, they did not get the slaps on the back. Not knowing was weak.

One thing that venture capital teaches you, though, is that you don't know. There is probably no other area of life--aside, perhaps, from going up to bat in professional baseball--where you are constantly and ruthlessly reminded that you are not only not always right but that you are usually wrong. And even, that when you did everything possible--all the due diligence, all the research, all the number-crunching, all the strategic thinking--there was no way, a priori, to know. Many venture capitalists do not accept this. Almost all people do not accept this. But it's an inescapable fact. And, it's not only true in investing but in all parts of life--it's just more obvious in investing. The proper lesson to learn from this is probably humility, but successful people rarely became successful by being humble.

This decade I've been too often reminded that a chance meeting or an idle phone call can turn what looks like inevitable failure into wonderful success. Or vice versa. That is, that there is never any way to know for sure what will end in success and what will end in failure. For me luck has not been risk; it is not beta, it is not some fat-tailed distribution. It has been the unexpected and, yes, the unknowable.


I think about luck a lot.

There's an old story, related by Alan Watts, about the farmer whose horse ran away. His neighbors say "Such bad luck!" The farmer says "Maybe." The next day his horse comes back, with two wild horses following him. His neighbors say "Such good luck!" The farmer says "Maybe." The next day the farmer's son falls off the horse while trying to tame him and breaks his leg. His neighbors say "Such bad luck!" Etc.

Good luck and bad luck come and go, and there is no distinguishing them, either before or after the fact. There is no beginning or end of things because there is no way to isolate a single causal event; causes effect widely. We technocrats would explain this by saying, simply, Things are Complicated--we have a Math for that. But this is not enough: this interconnectedness through time and across events could more easily lead to a dampening, an unchanging condition, but it does not. Watts explained this by saying that there is a cycle of remembering and forgetting. Anybody who's been paying attention would have to say that, as is often the case, the zen analysis is more useful than the scientific.

This decade I've had a lot of bad luck end up well and a lot of good luck end up poorly. If I knew then what I knew now, there would have been a lot less Good Luck and Bad Luck and a lot more Maybe.

I've often felt that my comparative advantage in the work world is my willingness to take the risks than the next person would not want to, to weather more sheer unpleasantness, deal with more complexity, and in general gladly volunteer to walk into the knife-storm that my professional life has often felt like over the past ten years. That this sort of thing did not bother me much, while others found it intolerable, is a valuable trait, it seems. It worked for me. But if my neighbors came up and said "What good luck!" I would have to say, Maybe. Because while things worked out at work, they didn't at home. My thinking that I could isolate my home life from my work life was part of the reason. But nothing is isolated.


So, this isn't one of those What a Great Decade posts, although it was a great decade for me, for a lot of reasons, work and even personal. It's just my annual reflection, being realistic about the good and the bad. For me, it's always easy, upon being told "What good luck!", to say Maybe. The hard part is saying the same when told "What bad luck!"

Ten years ago my goal for the decade had something to do with doing well and, perhaps, even doing some good. In retrospect, my major accomplishment was learning how to say "I don't know" and then stopping to listen. (I do not consider my wonderful children accomplishments, I consider them gifts.)

For the next ten years, my goal is to be able to answer Maybe to bad luck with the same equinamity I do to good luck. If I could do this, then everything else would be easy.


But I do want to say a little bit what a great year it's been. I want to appreciate all of you reading here. And thank you for commenting and connecting outside of this forum. I use this space to think out loud, it helps me order my thoughts to know that other people will read them. This is valuable to me, and why I started writing in the first place.

What I didn't expect was that in addition to the benefit of writing, I get the benefit of connecting. I've met a few great people through this blog, I've stayed connected to a few other great people, and I've learned an awful lot from the people who read it. I am thankful. I haven't quite quite found the invisible college, not yet, but it's a start.

So, thank you. And in the new year, may good luck find you and may you transform all bad luck to good.

Tuesday, December 22, 2009

Pinch and Flurry Merging

Pinch Media and Flurry, the two best providers of analysis services to mobile phone app developers, are merging. This is a huge step in enabling the better monetization of these apps, and I think Pinch/Flurry will be a key enabler in the growth of this market.

A year and a half ago I sat on the floor in the back of the TWC Borders* with Greg, talking about the mundane details of getting a company started. It's awesome to think how far Greg and Jesse have taken Pinch from there, and how bright the future looks with them teamed up with Simon and the Flurry team.

Maybe it's just me but--even in our more democratized era--the celebrity status of successful companies and their CEOs causes some cognitive dissonance when I'm sitting on the floor of a Borders in t-shirts and sneakers talking to an entrepreneur about building a great company. It's worth remembering that most great companies started similarly.

Congrats on a big step, Greg, Jesse, Simon and the rest of the Pinch/Flurry team.

* In NY we don't start companies in garages. The attendants frown on it.

Sunday, December 13, 2009

Information and markets, 3

An anecdote.

A few years ago a private equity firm asked me to help them look at some lead-gen companies. One of the target companies had a network of many thousands of small lead buyers and were known to produce high-quality leads. They had worked years to build this reputation. They got good prices for their leads. But high quality leads are expensive to generate--and there aren't as many of them out there--and building and maintaining a large distribution network is also expensive so while they had historically made OK margins, they weren't blowing the doors off and their growth was steady but slow.

Then the founders started thinking about selling. They hired an investment banker a few months before the PE firm started looking at them. In that time, revenue had started to grow faster and margins had started increasing. The lead-gen company credited improved technology.

The PE buyer was enthusiastic. They asked me what I thought. I called a couple of people who were in the same lead sector. This is what I heard: "they're stuffing the channel, everybody knows that." Maybe everybody in the lead-gen business, but obviously no one in the PE business.

The company had decided that, to boost their valuation, they were going to generate lower-quality leads, and more of them. Since their customers generally bought a few dozen leads at most, they did not in the few months after the change notice when the average conversion went from about 5%-10% to something much lower*. This allowed the lead generator to raise revenue and margins. I warned the PE company that the buyers of these leads would not be fooled for much longer, that this way of doing business would come back to bite them**. And that is what eventually happened; the lead-gen company lost half its customers over the next six months, as the customers became aware that they were paying for high-quality leads and getting low-quality leads.

Any smart buyer knows that uncertainty exists: quality changes over time, for many reasons. The natural response to this is a concentration of buying. Larger buyers have more information, so will notice and respond to changes in quality much more quickly. Any market where quality information is not available will favor large buyers over small, causing the exit of small buyers. This will then cause the marketplace itself to suffer: a large buyer does not need an external marketplace, sellers will come to them. Without a competitive marketplace, the sellers suffer, both from lack of pricing power and lack of information about what is working for the buyers (reflected in varying price levels.) This then causes inefficiencies in production, leading to higher production costs.

This decline in overall efficiency of the ecosystem affects the sellers first, but also eventually affects the buyers. The process, though, can not be avoided by the buyers, even if they are aware of it: they are stuck in a prisoner's dilemma. The only solution is to have more open and robust quality information available.

* Junior year I took a semester abroad. My alma mater did not believe that any other university, anywhere in the world, could possibly educate me nearly as well as they. As such, they would not allow me to include any class I took anywhere else on my transcript. The most they would do was allow me to take a test and place out of classes I took abroad. Because of this dynamic, I spent my time in London doing anthropological studies of the pub culture. When I got back I did pass the required tests and so placed out of statistics and a year of German. I now find that I can neither speak German nor do statistics unless I have had several pints of bitter. I was going to tell you how many leads a lead buyer would have to buy before they would have a good idea that quality levels have changed, but I wrote this post Sunday morning and I couldn't get my hands on a sufficient supply of ale.

** Oddly, they did not believe me. This turned out badly for them.

Thursday, December 10, 2009

Information and markets, 2

Those most familiar with the cattle trade agree that there often exist wide differences between the actual selling price of cattle in the market and the previous estimate by the feeders sending them forward as to the prices they should bring. The small feeder, who seldom follows his cattle to market, has a poor chance to learn market conditions and requirements, but the regular shipper has an excellent opportunity to do so. Feeders must rely largely upon the market reports for their knowledge of the condition of the cattle trade... Inability on the part of the feeder to interpret correctly market quotations places him at a decided disadvantage either in selling his cattle to a shipper or in shipping to the open market.--Market classes and grades of cattle with suggestions for interpreting market quotations, Herbert Mumford. (1902)
Beef isn't assigned to quality grades--like prime, choice and select--to help buyers know what to buy. It's assigned quality grades so producers know what to produce. Mumford's groundbreaking work led, eventually, to the voluntary grading of beef and other commodities by the USDA.

Raising cattle that has a higher proportion of well-marbled, tender, Prime muscles is more expensive than raising one full of chewy, touch Select muscles. So even more than needing to know at what price they can sell cattle, cattlemen need to know at what price they can sell different quality cattle, so they can figure out whether it makes business sense to spend the money to raise high-quality beef.

Markets not only consume information, they generate information: primarily information about demand at different price levels. Price information, of course, is what drives efficient allocation of resources*. So, lack of public information about quality causes both
  • Problems for the buyer when the seller knows quality but the buyer doesn't, and
  • Problems for the seller, when the buyer knows quality but the buyer seller doesn't**.
The first case is relatively straightforward: as per Gresham's Law, bad commodities drive out good. The second case is more subtle. Because the buyers know what a quality product is worth to them but the sellers don't, buyers will tend to price goods at the lowest possible price level that ensures they will keep being produced.

In the lead-gen world, I heard over and over from lead generators that it took at least a year before a newcomer could make money. Not because they couldn't generate quality leads at a decent cost, but because it took a year to realize just how much the companies buying their leads were screwing them on price.

In the display ad world, the buyers have access to all the information they need to judge quality--context, customer, behavior, etc. But the sellers, the publishers, have let themselves be isolated from the information they would need to link quality and price on any inventory they sell through markets***. It shouldn't be surprising that the prices they get are rock bottom.

* I'm sure you've read about it, but if you haven't actually read it, do: Hayek's The Use of Knowledge in Society. It's a good read, and short. It's also explains the key concept in how our economy works.
** If neither knows the quality of the good being sold, a robust market is still possible. The stock market, for instance.
*** Including the ad nets, the ad exchanges and the ad optimizers, all of whom run a type of market.

Tuesday, December 8, 2009

Information and markets, 1

There are many markets in which buyers use some market statistic to judge the quality of prospective purchases. In this case there is incentive for sellers to market poor quality merchandise, since the returns for good quality accrue mainly to the entire group whose statistic is affected rather than to the individual seller. As a result there tends to be a reduction in the average quality of goods and also in the size of the market.
--The Market for Lemons: Quality Uncertainty and the Market Mechanism, George Akerloff. (1970)

Markets that don't have a good way to judge the quality of the goods being sold have a problem. As Akerloff noted, if buyers can not differentiate quality, they will pay for a statistically likely quality level. This will then drive away the higher quality goods (thus lowering the statistically likely quality level!) Buyers then adjust their price down, and this downward cycle continues until only the shoddiest goods are left.

One long-standing problem in the lead-gen industry (and, in a different way, in the display ad industry) is the inability to grade quality. There are high-quality leads and low-quality leads (quality here meaning likelihood to convert into a sale.) It's cheap to generate low-quality leads (think reg path or, if you've been around a few years, free ipod.) It's expensive to generate high-quality leads.

Problem is, once a lead is generated, it's pretty hard to tell if it's high quality or low quality. You can cross check address, telephone number and email, but you can't see from the face of the lead the key unknown: intentionality. Does the lead actually intend to buy the good or service they entered their information for. Anybody who's been the person calling the lead can tell you how often they hear "I'm not interested in a new car, I just wanted the free _____." This is a poor quality lead, despite all of the information being correct.

Imagine a lead market where the leads have a random quality from 1 to 100. Leads with quality 1 are worth $1. Leads with quality $100 are worth $100. What would you pay for a lead? Statistically it would make sense to pay about $50. On average, you would be getting your money's worth. But when the price level is $50, the people who are selling the leads with quality greater than 50 all leave the market (and, probably, start generating lower quality, lower cost leads.) The average quality now sinks to 25, so the price also goes to $25. Repeat until the quality reaches the lowest increment. This is Akerloff's point, and what I've seen actually happen in lead marketplaces.

Now, ask yourself, why is the inventory trading through the ad exchanges the worst inventory above remnant? Buying and selling through an ad exchange beats direct buying and selling in every single way that doesn't involve expense account meals. Yet both direct sales and ad network/rep sales have higher CPMs than the ad exchanges, because buyers believe the higher quality impressions are sold that way. Is there an information problem here?

Sunday, December 6, 2009

CPMs rising?

So, this has been bothering me for a couple of months. I figure it may well be a figment of the data, or I am abusing the data. But I don't know, so I'll put it up... comments welcome.

This is a graph of Implied Average Online Display Ad CPM, 2006 through Q2 2009 (left axis, thick blue line.) Implied Average CPM is ad spend divided by impressions.

The right axis and thin black line are impressions in millions, as per Thursday's post. This seems to show that as display ad impressions fell in 2008, ad spend did not fall as fast. For this to happen, CPM must have increased. This is both not what has happened, anecdotally, and is hard to believe in this era of expanded access to non-premium inventory. But I hate to think I believe the data when it confirms my preconceptions and then disbelieve it when it doesn't.

Are average display CPMs really nearing $7?

Anyone know what's going on here?

Sources: Ad spend--TNS. Impressions--Nielsen Online. Both purport to be display only.

Friday, December 4, 2009

Ad impressions by month

More data. Nielsen Online's ad impressions per month, via (paid CPM display ads only*.)

First, the raw data.

Broken out by category, sorted by change from August 08 to today: largest decline on top (software), largest increase on bottom (telecommunications.)

Change in four largest categories, and total, since February 06 (Feb 06=100.)

And percentage of impressions by sector. This one's a bit psychedelic.

* Per Nielsen: "Nielsen Online, AdRelevance service uses a proprietary methodology for estimating online advertising expenditures and only takes into account image-based technologies and advertising sold per CPM. Above data does not reflect house advertising activity, strategic partnerships between publishers and advertisers, or text units, paid search, sponsorships, email, units contained within applications (e.g., messengers and pre-rolls) or performance based advertising. "