Sunday, May 27, 2012

What Kind of Company is Splunk? Categories and Organizations


Let me begin by answering the question in the title. Splunk (http://www.splunk.com/company) describes itself as a company founded to "make machine data accessible, usable and valuable to everyone." With machine data they mean not just any data on a computer, but rather the data generated by machines rather than humans; by accessible they mean that they are making programs to analyze data that would normally be in an unstructured form, so not in a database. Their software deals with system configuration and monitoring, security logs, servers, virtual machines, and so on. The name comes from spelunking, or cave exploration, because supervising machine data is like cave exploration when the software tools are poor.

To someone outside the computer industry, like me, most of this description is hard to follow. For those inside the industry, each function and piece of software may be easy to understand, but it is not necessary clear why a single company would be founded to take care of these services and not others. What would we call such a company, and who are its competitors? That is the question of categorization. Customers care about categories because they want to know which firms to approach for bids when there is a need for new software. Investors care about categories because they want to understand the strategy of a firm they want to back.

Usually when I look at a problem of categorization I want a clear answer. Green tea ice cream bothers me because a non-dairy product that is usually served hot is frozen and milky. There is something fundamentally wrong with three-wheel cars and four-wheel motor scooters. And, as Elizabeth Pontikes has found in recent research, I seem to be a typical consumer in that respect. If a firm is difficult to categorize because it belongs to a category that is hard to delineate or has significant overlap with other categories, then they assign a lower value to it.

The surprise came in how venture capitalists made the same judgment: they took the exact opposite tack and thought such companies were more valuable, not less. Why is that? The key seems to be that (potential) customers don't have any control over the firm, and are facing with a possibly unclear product or service that they have to take or leave. Venture capitalists, on the other hand, will invest in a firm and take a role in setting its strategy, and for them the unclear labeling may indicate novelty and opportunities to maneuver. So, it is not just the category that matters in how one reacts, but also the degree of control over it. For consumers, that still means staying with four-wheel cars. For entrepreneurs it means that the radical but still half-baked idea might be more interesting to investors than they imagine. Oh, by the way: When I looked today, Splunk was worth approximately 3 billion USD. 

Pontikes, E. G. 2012. Two Sides of the Same Coin: How Ambiguous Classification Affects Multiple Audiences' Evaluations. Administrative Science Quarterly, 57(1): 81-118.


Friday, May 18, 2012

The Late-Goal Effect: Risk-taking and Deadlines

Last week Manchester City became Premier League champions after a nail-biting match that had them lagging 1 to 2 by the end of regulation time, and so second in the Premier League after local rivals Manchester United, but then recovering with two injury-time goals to win the match and league. It was surely an effort to bring their fans to the edge of their seats, as well as the Manchester United fans. Lest we think that such late-game heroics are local the kind of football played in Europe (and, pretty much everywhere), fans of American football know that many games are decided by magical plays in the last two minutes of the 60-minute game time.

What is going on in those last few minutes, and why is it so different from the rest of the game? Looking at the data from American Football, Lehman and coauthors (2011) looked at the deadline-proximity effect. It starts with the idea that decision makers make changes and take risk if their performance is below a goal, or aspiration level. This is known in many situations, and is actually best documented for organizations making changes when their performance is low (Greve, 2003). But being below is worse just before a deadline, because then one has to do something truly dramatic and risky to catch up. And in fact, that is exactly what the football teams in the study were doing, because they were choosing a riskier play when they were behind late in the game then they were equally behind earlier.

Is such risk-taking good? For sports, it may be. Risky plays naturally have a chance of failing, but the extra risk could well be worth the chance of winning the game, and losing by more points is not such a great loss. The only problem would be the following. Risk in sports and in management is unfortunately more similar than we would like it to be, because there is the same sense of winning when getting just above a target level in management, even when the targets are arbitrary and the deadlines (usually by accounting convention) are arbitrary as well. We do not know yet, but it is reasonable to expect that some late-period risk taking occurs also in firms. Does the level of risk taking by managers and others in organizations matter enough for us to care about whether it is done consistently over time? A good person to ask that question would be J.P. Morgan Chase CEO James Dimon, who is now looking at losses estimated to $3-5 billion inflicted on the bank by some unusual trading positions.

Greve, H. 2003. Organizational Learning from Performance Feedback: A Behavioral Perspective on Innovation and Change. Cambridge University Press.
Lehman, D. W., Hahn, J., Ramanujam, R., & Alge, B. J. 2011. The Dynamics of the Performance-Risk Relationship Within a Performance Period: The Moderating Role of Deadline Proximity. Organization Science, 22(6): 1613-1630.

Sunday, May 13, 2012

What if Our Best People Join the Competition?


Readers of the business press have been watching the slow decline of the famous New York law firm Dewey and LeBoeuf, which took on too much debt and is reported to be investigating financial misconduct by one of its leaders. Most troubling for a law firm, which lives on the expertise and customer networks of its leading partners, it has experienced an increasing stream of defections by its partners to other law firms. To many observers, the doubling of departures in four months (8 in February, 22 in March, 40 in April, and 81 in May, according to Wall Street Journal) is a signal that the end is near.

What if it had not been a rush like the one we are currently seeing, but instead fewer exits? We might think that a firm would withstand a smaller stream out, as job movements are a common sight on most industries. But if these are key personnel, a case can be made that even a smaller number of exits is dangerous. So which is it? In fact, Wezel and coauthors have looked at exactly this issue (for accountant firm partners), and found some interesting effects.

First, departures do harm the firm - a plausible first result. Second, departures of individuals were much less damaging than departures of groups. This is not just because groups are "bigger", but also because their experience working together means that they add up to more than the sum of the individuals. Third, reasonably, the effect is worse when they depart to a nearby competitor. Fourth, and perhaps most surprising, the damage to the firm that they depart is greater when they start a new firm than when they join an existing one. What makes this surprising is that existing firms are often stronger competitors to begin with, and adding more expertise to current competitors seems to create a greater threat. But a startup is worse than a greater threat; it is a new threat.

This may not quite be news that rescues Dewey, which seems to have a bigger set of problems than the departure of a few partners. But, for other firms it may be useful to know that retaining key employees is not just good one’s own competitive strength; it also avoids strengthening existing competitors or creating new ones.

Spector, Mike and Jennifer Smith. 2012. "Bankruptcy Specialist at Dewey Heads for Exit." Wall Street Journey, May 13 2012.
Wezel, F. C., Cattani, G., & Pennings, J. M. 2006. Competitive implications of interfirm mobility. Organization Science, 17(6): 691-709.

Monday, May 7, 2012

Network Theory and Drone Attacks


Chances are that you will find a news story about a drone attack on militants within a week of reading this blog post. They have become common in Pakistan, and are now used in other places such as Somalia and Yemen. Drone attacks are expensive, with the missile alone costing more than USD 50,000, plus the cost of operating the drone. The expected payoff of drone attacks is to weaken insurgencies such as the Taliban. But do they work?

David Siegel has an article on repression and social networks that provides some answers on when such tactics are effective. The reason social networks are relevant is that people decide to join insurgencies based in part on what their friends are doing, because the actions of friends are helpful for learning what actions are considered good, what causes to join, and what the risks and benefits are. As a result, insurgencies spread through social networks, and they will spread faster when these networks have many connections and few places that are isolated from each other.

Thinking about insurgencies as social networks helps us understand what drone attacks do: they remove individuals from the networks, and when one individual disappears the connections that individual has also disappear. A significant number of attacks can tear the network into small pieces and stop an insurgency completely, but that requires a scale of attack much greater than anything currently seen. Instead, what the current pace of (approximately) one attack a week does is to pick of some parts of the network, leaving much of it unchanged.

Based on network theory, will that work? If individuals are equally eager to join the insurgency, and equally well connected, such a strategy has only a remote chance of stopping an insurgency, because there are too many connections that need to be severed. If some individuals are (much) more influential and connected, and the attacks hit those individuals more than others, they can be effective, especially if these individuals occupy central positions in the network. Because drone attacks target leaders or assemblies of central persons, this is in fact the theory behind them. They remove those who are most central in the network, and through that seek to remove those who are the strongest drivers of the insurgency, and break the network apart.

But what if friends of the people who are removed (well, killed) by the attacks react by getting angry and joining the insurgency? In that case, the attacks not only lose their effectiveness, they can in fact become counter-productive and can fuel the insurgency. This backlash effect is true under one, ironic, circumstance, however. Attacks are counter-productive when they hit actual leaders with many ties to others who can become angry. If they instead hit innocent people, the backlash will be less because few people in a society of villages have enough network contacts to fuel much anger.

Siegel, David. 2011. “When Does Repression Work?: Collective Action and Social Networks.” Journal of Politics 73 (4): 993-1010.