Leveling; getting unexpected results

Question

I have spent hours reading up on the leveling algorithm and trying various alternative approaches but I have a problem I simply can't figure out. I'm using MSP 2016.

My schedule consists of around 1250 tasks made up of about 250 separate and independent projects each with about 5 tasks. Each project and its related tasks has been assigned a unique Priority Score; eg Project A and all its tasks are 950 whereas Project B and all its tasks are 949. However, the higher ranked project may be further down the ID hierarchy; eg Project A with score 950 may have starting ID of 525 whereas Project B with score of 949 may start at ID 275.

My leveling options are Month by Month with Can split tasks. Initially I chose Priority, Standard as the leveling order.

This worked pretty well but I found a few tasks that were inexplicably leveled to start AFTER lower priority unrelated tasks, even where there appeared to be available resource to start on or sooner after the original Start date.

In a particular case the lower ranked task was delayed by leveling to start on 1 Jan and finish on 31 Jan, whereas the higher ranked task, which had a longer duration of 2.5 months and which should have started on 17 Jan was delayed to 1 Feb.

So that is my first question; how could this happen? Why was the lower priority task not split to allow the higher priority task to use the available resources and start on 17 Jan? In fact, why was this task delayed at all when there appears to be sufficient available resource?

I then sorted my plan by Priority and then ID, and then permanently updated the ID values to this new sort order. Then I tried leveling by ID only. The results were similar (not surprisingly) but not exactly the same, and still had the problem with some tasks not being prioritised correctly.

I then changed the leveling order to Standard and found that the problem task now started on the correct date but now had a longer split mid task and that many higher priority tasks were delayed more than previously.

I have tried to replicate these issues in a simple generic plan but have been unable to do so, so I assume they are an artifact of the complexity of my plan. However despite its size it really is not complex as each group of 5 or so tasks is independent of all other groups.

This is really causing me some problems, can anyone shed any light on the causes of these apparent inconsistencies?

Thanks

Eric

Answer 1

This is not really a possible canonical answer but more of a suggestion. I have found that a tool's leveling feature consistently produces results that were both unexpected and seemingly weird. I personally avoid allowing the tool to level my project's schedule. I instead use it to see possible solutions for leveling that I may or may not use as I manually level the schedule. In other words, use what it does that makes sense and ignore what it does that makes no sense, then punch your baseline. Also, I avoid trying to get to "perfect" leveling with zero over- or under-utilization since work is always probabilistic. It's a precision that is only valid until two minutes after you start work, then it's off.