"Cross-impact analysis is often thought of as an extension of the Delphi technique. Like its name entails, it involves identifying and evaluating the impact of trends or events upon each other. Cross-impact analysis is commonly used as part of an expert-opinion study, which is why it can be considered part of the Delphi technique. It is useful in exploring a hypothesis and in finding points of agreement and divergence.
Cross-impact analysis involves constructing a matrix to show the interdependencies of different events. A matrix lists the set of events or trends that may occur along the rows, and the events or trends that would possibly be affected by the row events along the columns. Respondents are then required to assess how the occurrences in each of the rows affects the probability of the event in the corresponding column."
After preliminary probabilities and interdependencies are estimated, the probabilities are iteratively recalculated using Monte Carlo sampling or another method.
It's hard to say whether cross-impact analysis improves the orginal estimates or not and, if yes, how much.
A complement to cross-impact analysis is cross-consistency assessment.