Negative lags

Why it is recommended NOT to have any lags in relationships ?

Although I use lag, I seldom use lead (negative lag), but I do use it. Here is why I avoid negative lag, if there is nothing else driving a task, then use of negative lag essentially "guarantees" the occurence of a future event. Example, I have a task that says "Execute Software Test" that is 20 days in duration. OK, I am expecting to be "in test" for 20 days (4 weeks). A successor task is "Author Test Report" set as "Execute Software Test" FS- 5 days. That is, I will start writing my test report 5 days before the ***forecast *** test end date. Really? This tells me that if I start authoring the test report on test day 15 that my action will cause the test to be complete 5 days later. Really? That is what the logic tells me ... I start authoring the report 5 days prior to the finish of the testing... if i started the the report is must be because the test is ending if 5 days.  Reality check... what is missing here is sufficeint detail in the "test" task to effectively establish the start of "write report". Someone will post, "Fine, then do it SS+15d. " Well, that is somewhat better. But if the test runs ahead of schedule (does that ever happen???) and finishes in 12 days days instead of 20, are you really going to take 3 days off before you start authoring the report? Of course not, so SS is probably not the best case either, but it is better (in my opinion) than using the negative lag.  Both situations beg for additional detail in the predecessor task. Well, at some point the schedule model has to stop with the detail and let the people do the work. I really dont think any program manager wants so much detail in their schedule model as to schedule "Execute Test 12.32.11.a" for all 1100 tests in the document, right?  To use lead or lag, you have to ask yourself a question as to what the lead/lag means. In this example, is the -5 days just to help reduce schedule duration: "I can buy a week of schedule if I start the test plan early"? Or is there really an event trigger where I can truly start the test report FS without lead or lag? Can I start the test report when I have claimed 80% (or 92%) of the EV for "Execute Test"?? That is probably the better predecessor as it will walk with the EV of "Execute Test"

I think you are a little confused ... as well as confusing others. In your example, you write:  "I will start writing my test report 5 days before the ***forecast *** test end date. Really? This tells me that if I start authoring the test report on test day 15 that my action will cause the test to be complete 5 days later. Really? That is what the logic tells me ... I start authoring the report 5 days prior to the finish of the testing... if i started the the report is must be because the test is ending if 5 days."  You are confusing which is the predecessor and which is the successor. Acconrding to your own words, the TEST is the predecessor and the REPORT is the successor (" A successor task is "Author Test Report"). If the Report is a successor to the Test, then how can the commencement of the Report (a successor) "cause the test (which is the predecessor) to be complete 5 days later?" You have it wrong, my friend.  As for your argument about the use of a SS of 15 on a 20 predecessor, my suggestion would be to use a SS of 5 AND an FF of 5. That way, five days into the Test, the Report can start. The FF will ensure that the Report still has five days to complete AFTER the finish of the Test.  What you and I DO agree on is that a negative FS is a hoaky way of tying activities together. Will the report really take two weeks (10 days) to write? Hhow much of the Test must be completed before the Report writing can even start? How much of the Report writing must take place AFTER the Test is complete? Answering these questions will provide the information you need to set the SS , the FF, and the Report duration.  Suppose we are told that, if we were to wait until the Test was 100%^ finished, we could write the Report in one week, 8 days, Then this is our Report activity duration. Suppose they say that five days of the eight must wait on the completion of the Test. Then the FF is 5. Now suppose they say that they would need to have al least half of the Test done before commencing the Report. I would ask for two durations for the Test: the expected duration, and the best case duration. They tell us that the expected duration is 20 days, but the best case is 12. Since we would need "at least half" of the Test done, then in the best case scenario, that would amount to six days. So I would set the SS to 6.  Lets say the Test starts on Day 100 and takes 20 days; thus, it finishes on day 120. Based on the SS pof 6, the Report would start on Day 106, but would not end until Day 125 (based on the FF 5).  Suppose, in the alternative, that the Test takes only 12 days. Its start would still be Day 100 and its finish would be Day 112. The Report would still start on Day 106 and it would end on Day 117. How did I get Day 117? The duration for the Report is 8 days and it starts on Day 106. If not for the FF, the Report would end 8 days later, on Day 114. But there IS an FF 5, and since the Test finishes on Day 112, the Report will finish five (FF5) days later, or Day 117.  Using the combination of SS and FF, we are able to overlap the Test and the Report in a very simple, rational, and understandable manner. Using the negative FS is convoluted and counter-intuitive.

I think this discussion brings us to this conclusion: When we locate negative lag, it is a sure sign additional planning is required. Let me see if I can clarify my point about the confusion of starting to write the test report prior to finish of the test. Using FS-5d, since the predecessor drives the successor ... the only way I should start writing the report is if I know with high probability in 5 days the testing will end. Knowing with high probability that testing will end on day 125, I can commence writing the report on day 120. Yes, additional questions come to mind. Another one to ask might be, "When it becomes clear on day 123 that testing will not end on day 125, and the report has already started, what will be done?" Questions like that help to change the logic such that someone would then say, "Well, the contract actually requires me to submit the report 15 days after completion of the test, and I just need more time to author the report." Bingo: FF+15d !!, Duration=20 days, and, you win a deadline date too. After a careful read of my prior post, I can see why it can cause confusion - sorry. I hope between this additional information and Stevans post that readers will have a more clear idea, avoid negative lag when possible and logical. In any event, the logic in the schedule model should accurately reflect how the program can be executed and what is driving what. There may be more than one way to solve the problem, each with its own gremlins.

some time ago, we suggested to Primavera to have a relation that is linked to a % done rather than days. I.e. in your example, the relation would say Start of the report is on 80% completed of the test. They accepted the concept but feedback so far. I think is has other complications when it comes to % complete type and EV. but it worth the invistgation.

I would rather use a SS link with a positive lag instead of using a FS with a negative lag. A general rule of thumb is that negative lags are un acceptable in time schedules and most likely you will receive a comment from the Engineer or the PM on this kind of dependencies. Also, a neagtive lag can cause future problems while updating the schedule during the execution phase.