Page 206 - Untitled-1

MANAGING SCHEDULE CONTINGENCY 185

group of tasks come together, providing some margin in case one of the tasks
slips. Similarly, a safety allowance is added by each level of management. Finally,
on top of this, everyone knows that the total duration will not be accepted. They
expect to be pushed for a 20 percent reduction, so they add 25 percent to the al-
ready inflated estimate.

The Shared Contingency Idea

Essentially, Goldratt’s solution might be called shared contingency (my term), and
is applied in several stages. First, he locates the critical path and reduces task du-
rations to be consistent with a 50 percent probability rather than 90 percent. Half
of the removed duration is added at the end of the path, as a project buffer.

Next, the feeder paths are located and treated in a similar manner, and half
of the removed duration is added at the end of each feeder path, as a feeder
buffer. The overall project schedule is reduced. Emphasis is placed on monitor-
ing the project buffer and feeder buffers (for shrinkage), rather than managing
the critical path.

The moving of the inflated portion of task estimates to a collective buffer has
always been an option in traditional critical path programs, and does not require
the abandoning of such programs just to adopt the shared contingency protocol.
However, commercial products that support critical chain have extended func-
tionality to address buffer analysis and additional critical chain features.

Trap If you were to adopt the full critical chain philosophy
and support programs, you would also have to adopt the full
set of rules and processes associated with critical chain, and
abandon many of the important features of traditional CPM,
such as earned value and milestones. So be sure that you want
to do this before changing over to CCPM.

Managing Schedule Contingency

Now that we have defined several ways to improve task estimating and to create
schedule contingency, we have to provide for the management of such contin-
gency. It would be wasteful to just move inflation values to a buffer and then as-
sume that the buffer duration is up for grabs. In fact, that would have a reverse
effect. Float, slack, or buffered contingency are not slop time. They must not be
treated as extra time available to waste.