"TOC delivers results or you don't pay"

VISITORS

 
SO WHAT HAPPENED?

Obviously the results of each unique game are different, and it is best to do multiple trials to get a sense of the likely results over time. Most of us are surprised by the outcomes―I know I was the first time we ran this exercise. Below is a sample game result we can use as a basis of discussion. The rolls went as follows:

Task 1 - 5

Task 2 - 9

Task 3 - 3

Task 4 - 8

Task 5 - 6

This example is actually not very bad in the end. We expected an average finish date to be 21 days (each task on average would take 7 days and there were three tasks on the critical path), and the actual result was just a little over at 23 days. 

Of course it is easy to attribute this to the variability of the dice rolls, “better rolls better results, it all averages out.” To test this further we can determine the average roll and show why we were late. If we sum the rolls above (5, 9, 3, 8, 6) we get 31, divided by 5 rolls the average roll was just a little over 6 at 6.2! The rolls were actually better than the average we needed to bring the project in on time and it was still almost 10% longer than required. So who was to blame? Perhaps one of the resources really let us down, let’s see what their averages were:

Blue (Tasks 1 and 4) delivered in 5 and 8 days = 13/2 = 6.5 average

Red (Tasks 2 and 3) delivered in 9 and 3 days = 12/2 = 6.0 average

Green (Task 5) delivered in 6 days = 6.0 average.

It seems every resource performed better than the average expectations and still we were late—there is no one to blame! It is our experience that about 4 out of 5 iterations will result in project durations in excess of 21 days and only 1 in less.

The reason for this is known as the Cascade Effect: delays multiply and gains do not add up. It is apparent in the above example in a couple of places. For instance, on Task 1 Blue finished early, in 5 days, but could not start the next task until day 10 because of a delay by Red on Task. Neither could Task 3 begin because the Red resource was still occupied with Task 2. Not only was Blue’s 2 day gain on the average lost, but because it took Red 9 days to complete Task 2, the next tasks were delayed. Resource contention caused Blue’s gain to be washed out, and Red’s delay to be passed to all of the subsequent tasks. When Red recovered by doing Task 3 in 3 days, it too was lost because Green had to wait on Blue’s 8 day completion of Task 4—again the gain did not add up and the delay multiplied through the project.

The Cascade Effect impacts any non-linear set of activities, and any system where resources perform multiple tasks. Unfortunately nearly all projects utilize shared resources (either within the project or between projects) and have multiple integration points (where more than one task or chain of tasks must be completed before the next task can start). It is not hard to imagine what the performance would be if instead of just five tasks, three resources, and one integration point (at the start of Task 5) we had the dozens, or hundreds, of tasks resources and integration points present in most projects—the outcomes would be far worse. And if we added in the impact of resource sharing across projects, the situation would worsen further.

These three situations are unavoidable in projects. Doing things one task at a time produces excessive project lengths, and it is not economical to pay for extra resources to eliminate all dependencies. (for more on the Cascade Effect click here)

Since the Cascade Effect stems from the interactions between tasks and resources, efforts to manage individual tasks more closely or to press resources to do more with less deliver little benefit to the project as a whole. As in the example here, there is no resource to blame, everyone’s performance is better than the required average. Pressing on individual resources to better meet their task delivery dates can even have the opposite effect. The more we hold people accountable for delivering on a “date” the more they will increase the “certainty” of their estimates.

The Cascade Effect is at the root of why so many projects miss their target dates, and why projects can take much longer than we intuitively sense they should. One alternative to missing dates is to work with much higher confidence task estimates—in the case of the game if we were to estimate 12 days for each task (the maximum roll of two dice) we would always deliver in the planned 36 days or less. Obviously this is a far from optimal result. So it is inevitable that if we want to achieve anything approaching aggressive project durations, and still deliver on-time, that we must deal with the Cascade Effect head on. For more on this subject click here.

©2007

The TOC Center, Inc.

19 Long Point Road, Stony Creek, CT 06405

Phone: 203-315-2123