Varying Resource Efficiency

N

NickK

Your insight and ideas on this scenario are greatly appreciated. I’ve
tried a couple of different ways to achieve this in Project but have
yet to find the appropriate way, if achievable at all. At the core,
it’s how to handle the progression of a project from proposal to
actual and the corresponding unknown versus known resources.

Example:
During the Planning stage of a customer engagement we develop the
software requirements and document these. We develop a draft task plan
utilizing generic “senior” resources. Essentially software engineers
that have a 100% efficiency. We put an estimated fixed price on the
project.

We move forward from the planning phase and get ready to start the
development. We revisit the task plan and can now assign actual
resources. Some of the resources are “senior” 100% efficiency 1:1.
Some of the other resources are mid or junior level and thus a 1.5:1
or 2:1 factor. Tasks are based on fixed work. So a 4 hour task may
take resource A (100%) 4 hours, but resource B (50%) 8 hours. I need
the work and duration to accurately represent the resources efficiency
and corresponding allocation.

Naturally over the course of a 2,000 hr development project the
resources may change and thus the work for the corresponding tasks
needs to be adjusted.
The only way I’ve thought of to handle this is to manually make the
adjustments as “real or known” resources are assigned, or re-assigned.

Modifying the resource’s calendar or units has been used in the past
as kind of a “hack” for projecting a duration representative of the
resources efficiency. Going forward this won’t be an acceptable
solution. I need to be able to use the resource’s calendar or units
for their intended purposes.

I appreciate your thoughts and insights on this. I could only find one
previous post related to this matter from 2003.
http://groups.google.com/group/micr...read/thread/3c3f478adf184141/f3d704485ebbc145
 
B

Bob C.

It almost sounds like a PERT analysis would help. That way you can leave some
allowances for best/expected/worst case scenarios (i.e. 100%, 75%, 50%, etc.)
in the plan, then adjust the schedule when you know which actual resource
will be assigned to the job.

I don't really see a way to accurate predict efficiencies / output if you
don't know which actual resource(s) will be used, therefore I think that
*some* guessing will be required.

-BC
 
J

Jim Aksel

Project has no way to know that Senior Engineer will take 4 hours but "New
Engineer" will take 12 hours. The best you can do is make the adjustments
manually when you assign the resources. I suppose you could write some type
of macro to alter the amount of *work* assigned to the tasks based on the
resource group assigned; that would get quite complicated if more than one
resource type is assigned to a task.

Personally, I would make sure you bid the work using the intended resource
mix. I also understand this may not be possible since you are probably
bidding multiple projects, only some of them will land, and who knows on
what date... etc.

You should monitor costs carefully. I assume you are using costed resources
as loaded into the schedule.
 
R

Rob Schneider

Sounds like you should consider "Monte Carol Simulation" of your project
plan. Can do with Project with add-in tools, or there are some bespoke
tools that import Project MPP files.

The idea is that you define a plan, then you look at the defintion of
the plan from a risk perspective and apply your estimates of possible
variation on some or all definitions. Typically these variations are
defined as a probability distribution with a min, mean, and max. The
shape of the distribution is also of interest, e.g. "normal, vs.
triangle, vs. Weibull) but perhaps only to statisticians. The important
thing is that variation is allowed and single point estimates are no
longer a limitation--and a risk! Then, after defining these variable
definitions on tasks (assigned to say work, start dates, duration, ...
whatever) then you tell the computer to re-compute all combinations of
all tasks to give you output expressed as probability distributions.
For example, you will get a computation of the project finish date from
a probabilistic perspective, e.g. the date that has probability of being
on or before at 20%, 50%, 80%.

My rule of thumb: Promise the 80% date to sponsors/investors and based
the initial target for 30-40% probable (stretch). As the project
progress, re-assess and bring the two schedules into unison.
 
R

Rob Schneider

Oops. "Monte Carlo Simulation".

Rob said:
Sounds like you should consider "Monte Carol Simulation" of your project
plan. Can do with Project with add-in tools, or there are some bespoke
tools that import Project MPP files.

The idea is that you define a plan, then you look at the defintion of
the plan from a risk perspective and apply your estimates of possible
variation on some or all definitions. Typically these variations are
defined as a probability distribution with a min, mean, and max. The
shape of the distribution is also of interest, e.g. "normal, vs.
triangle, vs. Weibull) but perhaps only to statisticians. The important
thing is that variation is allowed and single point estimates are no
longer a limitation--and a risk! Then, after defining these variable
definitions on tasks (assigned to say work, start dates, duration, ...
whatever) then you tell the computer to re-compute all combinations of
all tasks to give you output expressed as probability distributions. For
example, you will get a computation of the project finish date from a
probabilistic perspective, e.g. the date that has probability of being
on or before at 20%, 50%, 80%.

My rule of thumb: Promise the 80% date to sponsors/investors and based
the initial target for 30-40% probable (stretch). As the project
progress, re-assess and bring the two schedules into unison.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Top