The AACEI RP 29R-03 2011 on the modeled substractive methods provide for extraction of entire activities or a portion of the as-built durations representing delays or changes from a network analysis model representing the as-built condition of the schedule.
If two overlapped activities with original duration of 10 days overlapped by 5 days lags are reduced to five days then common sense tells us that lag must also be reduced to half.
This reduction is not automatically performed by many software as they only model time lag. This time lag yields wrong modeling when true lag is volume of work lag as it is most common in the construction activities. You keep a volume of work distance that is active as long as the predecessor is not finished, volume lag correctly models this while time lag is no longer active after it is consumed by time. Time time lag does not consider in any way the volume of work required distance.
I believe if subtractive models are restrictive, always in favor of the owner why add more with a wrong model using time lag instead of volume of work?
In the following sample job pacing is being extracted and the correct model, the one using volume of work lag yields a shorter collapsed network by a duration of 3.75 days.
Is very simple but the AACEI RP is protective of the limited functions of some software, they mention Primavera, although not as much as before, but make no mention whatsoever of this modeling functionality that is available in Spider Project and that can yield better models. They do not prohibit the reduction in lag but require any reduction in lag be justified when some software can do it within the correct modeling technique, under such models there shall be no need to justify one by one each due reduction in time when there is no reduction in volume lag.
[1]
I wonder how many of you have been in need of reducing lag in your collapsed models and how you handle the issue when it becomes frequent? Do you need to justify a better modeling algorithm for every occurrence?
Please note I am not saying time lag is always wrong, concrete curing is a very correct lag when modeling concrete curing, a lag with 24/7 calendar while successor and predecessor activities might have each a calendar of their own.