You can use seasonal shifts to model true productivity and true work hours + adjusted activity duration using Productivity DPH type.

You can then use other cost functions to estimate rate increase and also to consider overtime, nothing is lost, everything represents reality, no need to use workarounds. Use cost periods for its original intended use.

http://www.mediafire.com/download/1lrowvroc3ns5c7/RD108.001.sprj

Costs are easy, the challenge is on modeling complex resource allocation that are real. 

The same volume of work increased from 3000 to 8000 in the sample provided, the details can be seen by looking at the file, you can play with all variable such as different rates for the two excavators etc. You can add idle cost of equipment etc. You can increase labor rates as to make 6 hours in winter equal 10 regular hours, it all depends on the rules of what is to be modeled.  

I am hooked on resource modeling because during my PM days on the field we always had issues with resources, every single day we would be relocating some resources as to put into work idle resources left unassigned by the poor software that was available on these years. In reality even worst, to start activities with idle resources that were on hold/delayed by the software because a small percentage of the resource was unavailable, insane software of the 70's. 

You can use Monte Carlo to simulate weather events and the impact of the events on schedule probabilities.

Just remember to apply Monte Carlo to the same model and same engine you are using to manage the job for it to be valid. Some functionalities are not equal such as how different engines deals with some constraints but the most difference can be seen when resource leveling engines differ.  If you have to export your schedule to other software then most probably your model if using resource leveling will run hundreds or thousands of simulations with very different results.

Also remember the statistical methods are limited because are based on many assumptions, use them as a guide in particular to watch for relative changes/trends in probabilities of success. Another added value might be that they disclose which of the near critical activities have high probability of becoming critical. It is not uncommon for the deterministic critical path be different to the most probable critical path. 

Most of us have to deal every day issue on rain delays using deterministic schedule no matter if Monte Carlo is used or not and this is where we shall first become proficient. 

Hi Rafael

I was recently commissioned to prepare a baseline programme for a cladding contractor working on a new building in Siberia.

The loss of productivity during the winter period - October to April - was on a gradation of 6 productive hours to 3.

so I set up 6 work patterns and applied them to the outside working crew along the calendar.

The events that drove the work over the winter period had a spectacular effect on the end date.

Unfortunately for me the parties settled before the Arbitration Tribunal got underway.

Best regards

Mike Testro

It is not unusual for weather to have additional impacts other than idle time. At times the whole activity requires a different crew and work hours during different seasons. For this we model seasons as long duration shifts and resources using whole crews as skills for which availability is dependent on season. 

In the following screen Team 1 is the Summer team with twice the productivity of Team 2, the winter team. 

Definitively weather can have significant impact on some activities.

An activity might be delayed and moved from rain season to dry season, it will take less calendar days and therefore the delay impact is somewhat reduced. 

Rob,

You can download the following file for Puerto Rico rain data as you will see how much estimated rain days per month might vary depending on how you define it.

http://www.mediafire.com/view/4wjh53i3hv1179x/NOAA%20SJ%20Rain%20Data%20...

I do not believe in feeding buffers, critical chain is myopic to critical path as it assumes it does not change and as soon as it changes feeding buffers will be out of place. There are other issues but it is a separate discussion. 

Contractual rain is best modeled via calendars, so if the activity that can be affected moves/delayed to another month the duration will be adjusted automatically. 

I recall very detailed rain scheduling specifications that specified a final rain buffer, if I can get a hold on them I will post them. It should be interesting to compare methods. The specification required to deduct on every update a portion of the buffer. 

I use 2 schedules one a tight schedule that does not includes any buffer and another we use to submit to the client. We use the tight schedule to target for early completion without any buffer in the hope we will overcome student syndrome and Parkinson law, what critical chain theory is after but using different approach. By not showing any buffer if it does not happens we will not miss any early date.

Best Regards,

Rafael

Usually weather is a risk that is shared between owner/client and contractor as stated under contract terms. The most common distribution is that normal weather impact is to be absorbed by the contractor while the unusual weather is to be shared, extension of time will be granted and each party will absorb his own costs.

At times the contract conditions are clear as to how many days of expected rain per month are to be considered by the contractor. At other times the contract conditions make reference to NOAA data but fail to define what constitute a rain delay day creating much controversy; if a day with rain over 0.25 in., if over 0.50 in. or any other number.

The scheduling specifications usually require for the contractor to consider normal rain impact on his schedule and leave out above normal impact as to use the schedule for delay analysis. In such case if the contractor/client wants to model above normal weather event he shall do so using the same cpm engine but as a separate scenario.

Weather is a risk event that can be modeled using probabilistic models such as Monte Carlo but the contract conditions usually favor use of a deterministic model for normal rain even when source is a statistical data.

So let assume we know how many rain-days [work] per month are to be included in our contract schedule.

Usually wet weather affects only activities on the outside so for a start the calendar with the contingency shall be applied only to those activities. Then the issue becomes how to spread the expected idle time due to rain.

Some assume at random some specific days to be rain days but I favor spreading the idle time across every hour. 

For a 20 work day’s month with expected normal rain delays of 3 days based on work days, 80% of the month/days/hours are expected to be available for work. Using the hour distribution means we shall use a 51 minute hour.

Note the conversion of work days to calendar days is tricky and specifications usually provide to convert a work day to calendar day using a fixed ratio formula, not perfect but easy to interpret. 

If your software is not capable of defining hours per minute but every 15 minutes you will not be able to use this method that makes a better [more even] distribution.