backcast to fixed time horizon

To backcast to fixed time horizon is the most rigorous future focused means of decision making. It requires one to choose a time horizon at some absolute point in time in the future and to envision a set of possible future states which will exist then - as a set of scenarios. Then to step back towards now by cutting that time horizon in by typically two-thirds or halves, and repeat the process in each of those time horizons trying to envision what must be true at that time if the future scenario is also to be true, i.e. what intermediate events must occur or be avoided. Ultimately one can derive from measuring regret in this way, a single constrained best next step to take, after which one (theoretically) can repeat the entire analysis, in order to choose the next step after that.


The simplest application is the TIPA structure that simply adds a time horizon by which issue/position/arguments must be elaborated so that some kind of decision can be made - even if it is only which argument/evidence/source/authority to elaborate into a full TIPAESA structure.

Consider for instance a wiki meeting: the meeting will take place at a known time in the future - the time horizon. By that time, at least the agenda will have to have been settled, probably by determining which issues are the most contentious - which ones could not be settled in a debate by edit or debate by comment.

risky applications

In very risky applications much longer time horizons are used.

The method itself is not as high overhead as the degree of consultation required. Accordingly it can be used on a bi-annual, annual or quarterly or monthly basis to choose a series of steps or a basic direction, rather than every step in a process. The evergreen business plan, Living Agenda and position protocol methods all visit assumptions on such a regular but not continuous basis, or when extreme resistance implies that a particular decision may now be "on the wrong track."

without time horizons

Simpler backcasting methods exist that assume infinite or steady state horizons in which the exact endpoint does not matter. These however make many background assumptions and are much more amenable to time horizon mismatches.