Primavera Risk Analysis: Liquidated Damage (cost impact) or Recovery Plan (cost/sched impact) ?

Benjamin,

I looked for replies to your question with an interest and expected that somebody will suggest to use conditional networks: if the milestone is achieved after certain date then remaining activities will change (software selects alternative scenario). I don't know if this option is available in PRA but expect that it is natural and shall be available.

Hi, Benjamin.

I really can't answer your question re Primavera, but, if I may observe, you really have focused on one of the shortcomings of the entire project management methodology: Ben Franklin wrote 265 years ago that time is money, yet for whatever reason, project management refuses to recognize this cost (of which LDs are but one type). If I could wave a magic wand and make one change in PM practice, it would be to force every project to start by quantifying, up front, the cost/value of delay/acceleration earlier or later than a given target date.  And then the software should track that info as part of NPV, as is done with every other investment.

First, as far as Monte Carlo simulations are concerned, you should know that they suffer from the "Garbage In, Gospel Out" flaw:

• Changing activity distribution shapes can substantially alter forecast duration. For example, if you run the simulation using the triangular distribution default across the three estimates for all tasks, you can pull in the 80% confidence point by 10-15% simply by switching to the Beta (PERT) distribution.  Ain't computers great? Just switch the default and you'll shorten the schedule and avoid LDs!
• If you are using lags, the MC software will NOT vary the lags.  What this means is that if Task A is an SS5 predecessor of Task B, and Task A's three estimates are 6, 12 and 25, the software will set the lag at 5 whether it's assuming the duration is 6, 12 or 25. Makes a lot of sense, huh?  (This is one of the reasons that Mike Testro's policy of eschewing complex dependencies and lags is often justified.) So, if you're using lags, you might not be running late even if the MC software says you are!

The above are two little flaws of which most risk software users are unaware.

To return to your other point, the issue is the True Cost (TC) of a given activity.  Activities on the critical path have something called "drag" -- it's the amount of time by which the activity is delaying project completion (or a key milestone, if we compute the critical path to an interim milestone). And the corollary to critical path drag is drag cost: how much time that activity is costing (e.g., LDs) by delaying the critical path. 

And although finance and accounting departments have no clue about this, the True Cost of any critical path activity is the sum of its resource costs plus its drag cost, with the latter often being MUCH greater than the former. 

It seems to me that your process should be to:

1. Determine the drag of all your critical path tasks;
2. Compute how much each of their drags will cost in LDs (drag cost); and
3. Decrease the true cost of each of those tasks by either using their drag cost to justify the cost of additional resources, or changing the logic to reduce duration.

If you are interested in more information about critical path drag and drag cost, try this article in the Jan/Feb issue of Defense AT&L Magazine. Unfortunately, Primavera does not compute critical path drag, so you will either  have to compute it by hand (doable with all FS relationships but VERY tricky with complex dependencies!) or get one of the two software packages that compute drag: Spider Project or the add-on to MS Project that can be downloaded (I think still for free) from Sumatra.com here.

Good luck.

Fraternally in project management,

Steve the Bajan

Ben;

IMO you are better off using your scheduling software package to modify A1 rather than using PRA - add resources to modify the duration.  Then evaluate costs (overtime/additional resources) vs. benefits (not paying LDs).

Now if the question was; "Given a duration with Optimistic, Likely, and Pessimisstic durations and you wanted to know the likelihood of reaching your milestone on time", then you would use PRA.  Do you have such a range?  We can talk about profiles later, if you do.  The costs should fall out depending on how you resource loaded your activity(ies).

AA;

I am not sure I understand the question.  You seem to be asking a deterministic question best modeled by 'what-ifs' in P6 (or whatever scheduling software you are using).  PRA would answer the probability of LDs and a new finish date given a specific risk or task(s) uncertainty.  Maybe an example of the risk event/factor would help to clarify.