This HBR article from a couple of years ago has some good techniques for helping make better bets about how the future might evolve for a specific outcomes.  They would be useful when you’re at the pointy end of a scenario exercise, rather than at the start.  The entire piece is a worthwhile read, and my three main relevant takeaways can be summarised as:

1. When estimating data points that may occur in the future, make three estimates – one high, one low, and then, by extension, one that falls in the middle.  The middle estimate is much more likely to be accurate.
2. In a similar fashion, make two estimates about future data points, then take the average.  Note that it’s important to take a break between making the two estimates in order to avoid bias.
3. Create a premortem i.e. imagine a future failure and then explain the cause.