But we should appreciate a neglected cost of buying large companies, a cost we might call the complexity premium. This refers to the inherent, relative complexity of any large publicly-traded company. Some small caps are indeed "pure plays." Few if any large caps are pure plays. We sometimes read that conglomerates are historical artifacts that have given way to more focused companies in recent years, where spin-offs are generally held in better esteem than acquisitions.
But in reality every company is a conglomerate. And your average large-cap company is a multi-dimensional amalgamation of dizzying complexity and breadth. An investment that you can never fully comprehend. But that’s okay. As self-doubt strengthens faith, so does skepticism encourage onion peeling. I’ve argued before that Buffet’s principles need to be reinterpreted in light of (i) exponential technological change and (ii) the fragmentation and business line blurring that ensues from Friedman’s Flatworld. Buffet succeeded, in part, by projecting economic relatively stable profits forward in time, as a function of a company’s economic moat. This analytical game is harder nowadays. Few companies can build solid moats around an stable income stream; much less project them forward with high confidence.
For this reason, if you are evaluating a large cap company, in most cases the key financials (e.g., revenues, margins, returns) are just a starting point. For most large caps, corporate-level metrics are aggregates that combine the performance of several different units. Much of the success of the initial Advisor portfolios was due to the fact that, almost regardless of the size of the company, we analyzed each company as a set of several companies-within-the company. The market, being a voting/weighing machine, needs headlines; it will reduce most companies to a story that can be summarized in a banner.
In our experience, the best companies are the ones that are incubating high-growth units within the overall organization. And you won’t find those high-growth units in the corporate metrics. So for example, we saw that Ceradyne (CRDN) was not really a defense contractor but an advanced materials firm. (they continue to be compared to the wrong peers, generally). That premise then had an implication: for the long term, we cared more about their materials businesses than their body armor sales.
A great place to start in the segment data. Per SFAS 131, companies are required to disclose reportable segments. A reportable segment is defined by SFAS 131 as a part of the company that accounts for at least 10% of any one of the following:
* Total revenues
* Operating profit
* Identifiable assets
Take for example, Intel’s (INTC) latest quarterly filing. The average analyst is concerned with whether Intel’s gross margin will fall below 50% next quarter (as the company has guided, a level that would be a symbolic breach). Given the breadth of Intel’s product line and the strategic importance of truly new markets, it is not clear what you can really do with this overall gross margin number. It is an average. It is more important to identify Intel’s different markets; then to evaluate the respective importance of each of these market; then finally to evaluate Intel’s respective prospects vis-a-vis each market. For example, without offering an opinion, what is the relative importance of the desktop segment; i.e., versus mobile and other platform opportunities? If it is important, what are Intel’s prospects here?
Here is a chart of the Q2 2006 reported segment data. Intel reports three segments, digital enterprise, mobility and flash memory.
We can see big differences. Digital Enterprise is 70% larger than Mobility, but contributed fewer operating margin dollars ($931 million versus $946). Flash memory deepened its loss; natural for a high growth unit (except it didn’t grow). If we look further, we can find vivid illustration of the competitive chip market. Consider this note about the Mobility Group:
Net revenue for the Mobility Group operating segment increased slightly by $67 million, or 3%, in Q2 2006 compared to Q2 2005. We experienced significantly higher microprocessor unit sales, and to a lesser extent, higher revenue from sales of chipsets in Q2 2006. These increases were largely offset by lower microprocessor average selling prices. These increases were largely offset by lower microprocessor average selling prices. Higher operating expenses, and higher unit costs each contributed to the decline in operating income in substantially equivalent amounts.
To recap this note, the Mobility Group sold "significantly" more product but that volume increase was met by an (almost) equivalent price deflation. And to make matters worse, both operating and product costs increased. You would think they’d get more credit for growing sales, but this is a fiercely competitive business.
I am not rendering an opinion on Intel -- to do so requires far more analysis than I’ve done here. Rather, I wanted to illustrate the idea that you can start with segment reporting as a way to understand the business better.