From the time of Galton's experiment (1907) professors with many students had the easy task of using a large glass jar full of marbles or beans, collecting everybody's estimates, and calculating the average. They could verify that the effect is repeated constantly, with an impressive accuracy of the estimate, usually over 97.5%. Its use was soon extended to other fields. The collective average is almost unbeatable by each individual participant.

 I once had 160 of them before me, in Deauville, and the calculation of the collective average was instantaneous, thanks to remote controls and a computerized system. The people (the most experienced RasBank consultants) were truly amazed at their systematic underestimation of uncertainty. Moreover, the fact that the errors of each person, once aggregated to the average of 160, bordered perfection was even more incredible. Everyone was wrong about four "forks", but not a single one when they were a group. They were all extremely interested in this amazing phenomenon (Legrenzi, 2008).


Ford, the inventor of this method, called it Shang after the binary oracles of the dynasty that ruled China until three thousand years ago and who provided a response to binary questions. For example: after a hypothesis was made ​​on the arrival of rain, they indicated a date before or after, and progressively narrowed the range of variation.


• In the first query the participants formulate minimum/maximum numerical estimates.

Then they calculate the arithmetic means that become the initial constraints for the next query, where the central value is submitted.
• Then the participants say "major" or " minor" and the most frequent response becomes the new benchmark.
• So they proceed in the next queries and soon find an narrow interval for the estimate sought.


The method is efficient; in creating it, I considered the main critical aspects (listed in the book), but also other valid opinions. As a matter of fact, the method:
• starts from a single mass estimate, without predetermined interval (Wisdom of Crowds); only later are the estimates of minimum and maximum, larger and smaller asked for (Shang); and the same method is used for the group of experts, meeting face-to-face for the completion (Pfizer - Galleri);
• In this way we respect Surowiecki's four criteria (2010): each person has their own opinion, not influenced by others or driven from the top, and the data are aggregated correctly;
• we respect also the Janor Lanier restriction (2006) because we use only estimates requiring single figures or values;
• in its specialist version this method can be applied with rigor and high competence in scientific researches;
the expert supervision - trained with the original method (80 hours) - Prevedere per Decidere - neutralizes the remaining concerns.