During his time working at Vivendi Universal, Fabrice Cavarretta, a PhD candidate in organisational behaviour at INSEAD, says intuitively it felt that the company would either do extremely well or very badly. But it was not clear whether anyone could have predicted which way it would go. I became fascinated by Vivendi's top management team's composition, which was so homogeneous one could feel the situation turn out excessively well, or be a complete fiasco - one extreme or the other.

Drawing on this and other experiences, Cavarretta developed his research on the performance of teams within companies. However, as he explains the traditional route of studying teams' performance has been to see how it can be improved with the burning question, 'do diverse teams improve or reduce performance?' The findings from research related to that traditional question have been ambiguous. Some research shows that diversity can improve output because it brings different skills to the table. On the other hand, diversity could reduce performance if team members don't get along. So when measuring the results on average, these two effects combine into a flat and unclear outcome.

Cavarretta says he takes a different approach. My angle is to make a statement about the range of performance - that diversity in fact influences that range between the highest and lowest outcomes. Conventional thinking explored whether diversity moves that range up or down: I'm trying to figure out whether diversity grows or reduces the breadth of that range. In practical terms, this matters because the range determines exceptional outcomes - both low and high - which often matter more than average outcomes. Cavarretta cites the example of firms searching to make an exceptional public offering, such as Google, a context where exceptional performance is the ultimate goal. Alternatively, from a governance perspective where one is trying to avoid corporate fiascos, such as in the case of Enron, exceptionally low outcomes matter the most. In both instances, average outcomes do not matter much and predicting the range of outcomes may be more suitable.

Cavarretta has three key predictions. Highly diverse teams will be particularly risky, because of what he calls a 'social' hazard. If the team members don't get along, they will clash, which will lead to low performance, Cavarretta says. However, if they do all get along, this can lead to outstanding performance. Highly homogeneous teams are also risky since they face another hazard - this time, 'informational.' Their performance can be extremely high if they carry out a task in their competence set, where their cohesion will serve them well. Unfortunately, when such teams confront changing or complex conditions, they then dramatically lack the proper information, and perform outstandingly bad. These two ideal cases, highly homogenous and highly diverse teams, are both risky in their own sense, and can be contrasted with mildly diverse teams. Those are not going to be exceptional good in terms of their cohesion, nor the information available to them. Mildly diverse teams are therefore low risk, rarely outperforming but also avoiding fiascos.

This linkage between team composition and extreme performance is of importance for businesses, in particular industries such as aviation where reliability is a key concern and where extremely high or low performance is not acceptable, as this could result in an accident. In airlines, your prime concern is whether you will hit rock bottom! There's a saying in the aviation business that 'there are no good pilots, only old pilots.' Because that context is driven by risk reduction, it's essential to find the balance in the composition of the team that will minimise risk, which should be a mild level of diversity. The second example he cites is when it's important for teams to reach an extremely positive outcome. Venture capitalist companies invest in many firms, and in a portfolio of ten firms, three die quickly, and two to three are likely to deliver an exceptional exit such as an IPO. In the VC world, the firms in the middle that are only doing average, even if doing a bit better than others, are called 'zombies,' reflecting the fact that slight improvements above average are not interesting, compared to the possibility to make it big. In that context, knowing that team composition could influence the range of outcomes is important. Cavarretta's research may explain why an unusual team composition - for instance very homogeneous or extremely diverse - occurs more in nascent firms, because such a team may be more likely to deliver the exceptional performance often sought in entrepreneurial situations.

To support this research, Cavarretta studied the academic performances of 200 MBA student teams. By observing trainee future executives, Cavarretta says his field work supports his predictions that the team's composition influences the range of performance by regulating the two hazards of organizational life: whether team members get along (social hazard); and whether they have the right information (informational hazard).

His prediction, he says, shows there is no 'one size fits all' team composition - this depends on the context. When risk is to be avoided, mild diversity is the least problematic and helps to keep outcomes under control. In situations where risk is rewarded, both very homogenous or very diverse teams have a greater chance of achieving exceptional outcomes.