• starting from the conclusions and working backwards through the decision tree to find the resulting decision
    • conclusions … trade-offs at end of branches (leaves)
  • allows to filter out clearly bad outcomes and maybe early on rule out certain decision paths entirely (especially with more complex decision trees)
  • computationally expensive
    • grows exponentially with choices and depth