In the aggregate, forecasters may be “roughly as accurate as a dart-throwing chimp,” but some forecasters are particularly and consistently far better than average. Credit Suisse reports that the book Superforecasting: The Art and Science of Prediction by Philip Tetlock and Dan Gardner provides important insights into how to improve forecasting skill, 
perhaps by as much as 60%. In other words, there are measurable differences between run-of-the-mill forecasters and “superforecasters,” and these differences can be a guide to improving forecasting skill.
Drawing on Tetlock’s work, the Credit Suisse report, which was written by Michael Mauboussin and Dan Callahan, provides four key ingredients for managers seeking to improve forecasting:
- Find the right people;
- Manage interaction;
- Train effectively; and
- Overweight elite forecasters and extremize estimates.
Finding the right people means employing forecasters who think in ways that are correlated with above average forecasting. “Foresight ‘is the product of particular ways of thinking, of gathering information, of updating beliefs.”
Superforecasters far more readily engage what Daniel Kahnemen calls “thinking slow,” or “system 2” thinking. They do not over-rely on intuition, but actively seek to challenge their own beliefs and counteract common psychological biases and heuristics (“rules of thumb”) that may limit forecasting effectiveness. They are open to new information, spend a lot of time thinking about their own process, and embrace feedback in order to improve. They operate in what may be called “perpetual beta” and tend toward more specific use of probabilities (i.e., more finely grained probabilistic assessments).
Tetlock’s work revealed that teams were on average 23% more effective than individuals. While there may be some challenges to utilizing this information in a corporate setting, “[t]he main lesson is that interaction among a diverse group, especially those with a profile of a superforecaster, can be very effective if managed properly.”
Training designed to improve forecasting includes two main elements. For individuals, training focuses on awareness of common psychological biases and ways to counteract them. For groups, training focuses on improving the ability to work together, which may be collaborative or competitive.
Finally, Credit Suisse suggests that a lesson of Tetlock’s work is to overweight the predictions of superforecasters and to “extremize” results provided by multiple forecasters (whether individually or in groups). The latter point is aimed at capturing “unshared information” by using an algorithm to push results further toward 0 or 100%.
The report also emphasizes the importance of good questions, which are measurable questions with a reasonable time horizon, and of actually measuring the success of forecasts (“keeping score”). In addition, it emphasizes that while forecasting itself requires a certain kind of thinking, acting on good forecasting requires characteristics often associated with strong and bold leadership.
