Welcome to Bill Gurley Notes. Together, we’re taking a step back in time to the early days of the internet and into the writings of Bill Gurley, the famed 6’9 venture capitalist and author of his aptly titled blog Above the Crowd. We started with his maiden issue and are working forward in chronological order—welcome to the 10th issue! To follow along with the un-notated versions, check out my friend KG’s comprehensive collection of Gurley’s writings here.
Perception vs. Reality: The Increasing Complexity of Measuring PC Demand
4/3/1995
“I would rather be vaguely right, than precisely wrong.”
—John Maynard Keynes
Over the past few weeks, we have witnessed a noticeable increase in inquiries about the current pulse of worldwide PC demand. This seems to have become a seasonal spring ritual among investors, and we are somewhat amused by the extreme confidence (and lack of inquiries) that existed just three months back, in early January. However, putting that aside for now, we do recognize the critical importance of PC demand as it relates to high technology investment.
Historically, investors have used standard deductive reasoning to discern short-term PC demand trends. Using this approach, investors determined the buyers of PCs, checked with the channels through which they bought those PCs, and then aggregated the data into a picture of demand. From this, diligent investors determined whether or not the current earnings forecasts were accurate and adjusted their investments appropriately. This approach was simplified by the fact that the customers, products, and uses were quite homogenous. Determining demand involved simply having an acquaintance at ComputerLand.
ComputerLand: defunct since 1999.
With this post, Gurley is back to the theme underlying his initial few posts: the importance of poking at information flows and market perception.
It is our belief that the usefulness of this deductive technique is waning, due to a broad based increase in the complexity of the PC market. Today, most PC companies have three major product categories (desktops, portables, servers) instead of just one. Additionally, international sales now represent over half of worldwide sales. How many distributor contacts do you have in Singapore? The variation among customer types is also increasing. Ten years ago, all users were typical business users. Now, consumer sales represent one-quarter of all PCs sold. Lastly, the number of worldwide outlets distributing PCs has grown to include electronics stores such as Circuit City and discounters such as Wal-Mart.
What is the source of information; is your sample representative?
Let's take a look at a recent example that may help support our argument. This past December quarter, Compaq reported worldwide revenues of $3.25 billion. However, in the weeks prior to Compaq's year-end earnings announcement, excitement and anticipation around the Street led to a "whisper" number of $4.0 billion. The difference between this expectation and the ultimate reality was most likely responsible for the stock's dramatic fall. What is currently of interest to us is the realization that the market was capable of believing the $4 dollar figure. That represents a 23% error!
What failed us here? Why didn't one of those surveys or channel checks figure this out? Maybe it's just that a survey of 200 American dealers is inconsequential when a company delivers products in 100 countries through 38,000 marketing partners. We do not think this concept is that peculiar; it is really just a matter of industry scale and maturity. After all, we doubt many beverage analysts walk into their local Seven-Eleven to check on the demand trends of Diet Coke.
The 90s was an insane decade for PC growth. In 1989, manufacturers sold 21 million PCs worldwide and 9 million in the U.S. In 1998, manufacturers sold 93 million, with 36 million in the U.S. All according to Dataquest.
You would think that having a chance of being 23% wrong would be bad enough. However, the fact that this game exists inside a market adds another layer of intricacy. A famous metaphor suggests that the stock market is like a beauty contest; however, the judges pick not the prettiest contestant, but rather the contestant that they think is most likely to be picked by the other judges.
Think about how this complicates our demand prediction problem. Not only do we have several investors playing with inaccurate information, but the entire game is warped. We shouldn't be trying to predict actual demand, but instead should focus on where we think demand expectations will be in the future. Of course, we are aware that demand visibility is extremely low, but it is not inherently obvious that this helps us outguess our misinformed opponents.
Right, makes sense. Unless you have an information edge, your guess is going to be every bit as wrong as theirs.
Should we abandon all hope? We do not believe that abandonment of hope is a good idea, but maybe abandoning the old deductive "call the reseller" type of analysis is. A large amount of scientific work is being done in the area of replacing deductive type analysis with inductive analysis. In simple terms, inductive analysis is the process of formulating theories about how the world works, testing those theories in practice, and then adjusting your model over time. Of course, this heuristic approach should not seem that foreign, because this is how the human mind deals with everyday life. What is new here is that scientists are beginning to recognize that some systems are so complex that the normal method of deductive analysis is unreasonable.
What does all this gibberish mean for high tech investors? Well, for starters, the way we approach long-term investing will remain unchanged. Inductive type analysis and fundamental research are one and the same. This is because demand trends will undoubtedly be exposed over the long term. The only problem is that, without accurate short-term demand measurement, one may have to ride through many perception-driven stock price swings on one's way to success. It will be hard not to be spooked out of a position during these dark times.
In the absence of "call the reseller" accuracy, we think short-term investors may need to spend a little more time wearing their inductive thinking caps. Why have these stocks acted so seasonal over the past five years? What drives the annual expansions and declines in relative price/earnings ratios? What are some plausible theories that describe what on the surface seems like irrational behavior?
Even if fully predictive theories are found, the formulation process will no doubt add a ton of clarity to an investor’s understanding of a given market.
With this in mind, we will now formulate our own theory as to why these stocks exhibit seasonal behavior. We will call this our inductive thesis. It is our impression, or theory, that the severe lack of demand visibility is the key culprit which creates high tech stock price seasonality. The bottom line is that we have little idea how many PCs were sold last year and have an even weaker picture of how many were sold over the last 90 days. Therefore, to play these stocks in the short-run, you must be able to forecast the perception of PC demand and not real demand itself.
The more information becomes inaccessible, the more savvy investors must play expectations. That’s the only thing left.
Keep in mind that the growth of the home PC is causing industry sales seasonality to increase every year. This makes fourth quarter/fourth quarter comparisons easier and second quarter/second quarter comparisons tougher. Investors who fail to adjust their expectations will likely be surprised in December and disappointed in June. The other major factor is our gullibility regarding demand in the spring versus the fall. This past November, the book-to-bill ratio came in 500 basis points below expectations, and no one cared. Why? Well, we all knew demand was strong; after all, Christmas was upon us. In the summer, however, we continually allow marginally performing companies to use demand as an excuse for earnings below expectations. This unchallengable defense of "summer weakness" sure makes a nice scapegoat. The problem is that we allow the use of this defense by companies, and we also allow the excuse to affect our own perceptions.
Book-to-bill is ratio of orders received to units shipped. In November, this low ratio doesn’t affect perceptions because of the newfound seasonality of the consumer-led PC market: investors know demand is currently high.
One thing I don’t understand: if this ^ is the case for investors, why are they also surprised by high fourth quarter sales?
One easy approach is to be a contrarian -- "Buy em when they hate em and sell em when they love em." The other way to approach this problem is to ask yourself, "what news will come out over the next 90 days and how will this affect the perception of demand?" We are trying to determine what is believable. With interest rates sharply up for the first time in the history of the industry, and Windows 95's late arrival, we think the possibility of demand excuses and therefore jitters is quite high. And just imagine what will happen if the perception of how fast Windows 95 can deploy falls.
As noted in the previous issue, Gurley’s bearish Windows 95 outlook did not play out.
Takeaways
Gurley is singing a familiar tune for public market investors: become well acquainted with the air between market perception and reality. The less information is available in a given market, the greater the importance of understanding and playing market perception becomes.
The notable twist in this post: his dictum that investors employ inductive reasoning rather than deductive, heuristic-based observations. This is a a call to take a first principles approach to the problem rather than responding reflexively to announcements and rumors.
His application of an inductive approach to understanding PC demand is his strongest section. Let’s take a closer look:
Observation: there is seasonality in PC stock pricing.
Why? The market has demand invisibility. As the market has grown, old tricks for discerning demand no longer work.
So why, in the absence of real information, does price seasonality exist? Well, maybe investors have failed to account for the higher share of consumer PCs, leading to higher-than-expected sales over the holidays and lower-than-expected sales in the summer. (Note: summer weakness should not be allowed to be used as a scapegoat by otherwise-underperforming companies.
This is great because we get from Gurley a technique for looking at the market: ask why. Use those answers to construct a theory of market expectations. The process of doing so is no doubt just as valuable as—if not more than—the theory it produces.
See you tomorrow,
DS