The idea is to use Planning poker showing range by using “likely” and “worst case” estimates.
As Steve McConnell says “A single-point estimate is usually a target masquerading as an estimate.” Further he says “There is a limit to how well a project can go but no limit to how many problems can occur.“ To mitigate this, people should present ranges of value rather than single values. The ranges can be used to understand where people think there is a lot of risk.
One way of doing this in planning poker is to ask people to offer up two cards instead of one when estimating where the first card is the “likely” estimate and the second card is the “worst case” estimate (note that not everything should be just worse case of 100). You can use the variability to either produce a better single point estimate (more informed) or track both estimates.
How does this work. If I were to read a user story and ask the team to estimate it, I might have 3, 5, 5, 5, 5, 8 as the responses. From this you would think that people are pretty comfortable with the result and would not be surprised to see a 5 as the consensus estimate. But if the team were to use 2 numbers as above you might get something like (3,40), (5,8), (5,5) , (5,40), (5,20), (8,13). This says that at least 3 team members think there is a lot of risk and after discussion you would not be surprised to see the consensus estimate to be 8 or 13 (assuming you only record one estimate).
Note that this range approach can also be used on the detailed task estimates. Also this “range” approach can be used against a Product Backlog to discussion what is likely to be released by a certain date.