• What is the value of prohibiting Negative Lag if you can work around it so easily by modeling lag as an activity, a common workaround when using poor risk software that cannot model probabilistic lag?
• Without breaking the rules but still going back in time!

Rafel 

I don not konw the software that Andrei is using and I can not see any dates so its hard to analyze the use of a negative lag from a screen shot. I will send you my example of why it does not work especially in Primavera P6.

Zoltan,

float values calculated by P6 are wrong in many cases, so what?

An approach does not become wrong just because some poor software has limited capacity.

SF links and negative lags shall be used when they represent real life constraints.

Project model is used for decision making and what if estimates and so it shall reflect the real life, isn't it?

Of those links mentioned in the article perhaps the most prone to cause problems is the Maximal Links.  When you have two activities linked by SS with minimal lag, SS with maximal lag, FF with minimal lag and FF with maximal lag if the duration of any two of the activity changes you might get an impossible set of conditions.

These links might be needed to model some scenarios.  In some industries these scenarios are not uncommon; these links shall not be banned.  In any case when suspicious what shall proceed is to require a justification on their use. 

We shall do what Ekrem is doing; ask for justification of their use and existence.

You can start by looking for knowledge at Micro-planning as well as other sites that might have different views.  My favorite is Planning Planet where I am confronted with different opinions, where we can debate the issues.  Where I am given for free the links to such valuable references.

SF LINKS

Ekrem,

The following is another example of how SF link is used to control the changeover between two processes, this would be my second example.

• http://www.mosaicprojects.com.au/PDF/Links_Lags_Ladders.pdf
• Start-to-Finish links (SF) constrain the finish of a task based on the predecessor starting. The finish of the succeeding task is delayed until after the start of the preceding task.  If a lag is nominated (say three days), the finish of the succeeding task is delayed until three days after the start of the preceding task. 
• This type of link is used to control the changeover between two processes, if a business is changing from a security system that uses key cards for access to one that uses bio-metrics, the use of the key card system cannot finish until after the start of the bio-metric system.  If both systems are required to run in parallel for a time, a lag is added to the SF link.

My third example would be when a business is changing accounting system, usually for a minimum of three months the new and the old accounting systems must be run in parallel. 

All my examples are for a changeover between two processes, the classical use of Start-to-Finish links (SF). 

To Zoltan and everyone else,

• If you do not believe on Start-to-Finish links so which is your approach to these sample scenarios?

FS LINKS AND NEGATIVE LAG

To all,

In the lack of volume lag the only solution to Andrei Sannikov sample scenario is not only to use FS links but also Negative Lag.  Yes FS LINKS + NEGATIVE LAG !

• If you want to mobilize your hydrotest pumps shortly before your piping is complete (say 5 days) to make sure you flood your system on time then you need to use FS and -5 days lag - everything else is illogical and would give you false results:  if you use SS with a huge lag your pumps might arrive either too early or too late depending on the actual duration of your piping activity; FF would not work either because I want pumps to be ready shortly before completion of piping.
• If you do not believe on FS links or on Negative Lag which is your approach to his sample scenario?
• My solution is to use SS + % of volume lag.  In this way the pumps will arrive just after a certain percentage of the piping installation, never early, never late.  If for whatever reason pumps arrive late then flooding of the system will be delayed as expected.
• I believe in allowing the end user to select his software even if it lacks volume of work lag.  I would never ban P6 or MSP even when I consider them archaic and not good enough for my needs.  My claim on allowing the use of FS links is to allow those using software not capable of modeling volume lag to create some model that makes more sense than the other options they have on hand.

PDM VS ADM

• Some schedulers claim PDM with its new relationships has provided more flexible realistic project representation in schedule networks. 
• Some schedulers claim PDM networks are easier to develop as there is no need for the infamous Dummy Activities.
• Others argue ADM is superior because of the “Lag Drag” issues in contiguous models and claim non-contiguous models are superior.
• Others argue contiguous models are superior because they give you control on activity performance.  In case of “Lag Drag” you just simply identify its occurrence and take control on how activity performance is going to be executed by splitting the activity.
• Some schedulers argue CPM and ADM critical path are different representation of the same.
• http://www.microplanning.com.au/about_xpert.php
• Some schedulers argue CPM and ADM critical path are not the same.
• Network Scheduling Techniques for Construction Project Management - By M. Hajdu

I say it is up to the end user which method to use.  If you like ADM then you should not be banned from using it.  Micro Planner X-pert is an excellent option for those who have a preference on ADM.

Best Regards,

Rafael

• http://www.calveyconsulting.com/files/PS02_Great_Negative_Lag_Debate.pdf
• http://www.planningplanet.com/forums/primavera-project-planner-p3/414072/negative-lag

Paul H Harris warns about some issues with negative time lag such as; issues with calendars, issues with risk model and issues with percentage time lag, to name a few.  None of his workarounds can tackle the scheduling scenario presented by Andrei Sannikov.

Please be reminded that in more advanced software such as Spider Project lag calendar is independent of any activity, lag have its own calendar, fixed volume of work lag and percentage volume of work lag can also be modeled. 

Also Spider Project can model activity duration automatically adjusted by the software as a function of assigned resources and their production rates.

Lag can be represented by an activity as required by archaic risk software that cannot model probabilistic lag, so they suggest using an activity for this purpose.  But this is not possible for the case of negative lag. 

In some situations such as when volume of work changes then activity duration changes so in these situations the option to use % volume lag is in no way any dangerous.  Spider Project provides both options for volume lag as well as for time lag.

To avoid any such issues just use software capable of modeling dynamic activity duration, independent lag calendar, probabilistic lag, volume lag as a fixed volume of work as well as a percentage of the total volume of work allowing you to make better models.

Hi Rafael

Surely the same effect can be created with a FS link.

Your linkage would only work with a Powerproject buffer task.

Best regards

Mike Testro

no because you cant go backwards in time 

Here's the same answer I gave over at stackexchange.com:

One is tempted to do these to manipulate dates for just-in-time work:

1. Install pump -SF-> Deliver pump;
2. Construct platform -SF-> Platform pre-construction meeting;
3. Commission Eqpt -SF-> Energize equipment.

Although this gets the dates where you want them, the logic flow is reversed from reality. This messes up critical path and other logic analyses. Some people have pointed to the second example as an acceptable use case, though I would suggest it is just as bad as the rest. You can accomplish the same thing with correct logic flow and a zero-free-float constraint (if you have it available in your software) or intermediate/dummy milestone (if ZFF is not available, e.g. MSP.)

I do think Rafael's example is a good one, if problematic in practice.