Resource Leveling?

[Riko Wichmann]

Hi all,

I don't think, I understand resource leveling in MS Project.

Here is the situation:
The task is to optimize the preparation of a certain component for
further assembly. Let's say, I have 10 components coming in with a rate
of 1/week. Then I have the preparation steps planned with resources and
availability, and with dependencies. It's a linear series, when each
step as one and one one predecessor. These steps I copy 10 times and
link them with the arrival date of those 10 components. Nothing else is
set, no priorities, no constraints or anything. Doesn't sound to
complicated so far.

Now, the limiting factor is the availability of resources. That's where
I thought resource leveling would help.

If I look at the resource sheet, I select in the resource leveling
toolbox the overallocation level in a way, that all overallocated
resources get the yellow exclamation mark in the info column telling me
that I should level the resource using for example "hour by hour".

If I do that, using otherwise the standard settings, I get rid of some
of the overallocation but not all: 2 out of 4 remain overallocated.
Consecutive levelings don't help.

There are no constraints otherwise on any of the tasks which would stop
project from just shifting things until there is no overallocation. Why
does it not do that?

Next step would be: What happens if my components arrive with a rate of
2/week. You would assume, the whole process is AT LEAST as fast as with
1/week. However, after resource leveling it takes LONGER????

Using Project Standard 2002


Thanks for any hint,

Riko

 

[Mike Glen]

Hi Riko,

Welcome to this Microsoft Project newsgroup

Please see FAQ #28 - Resource Levelling Does Not Remove Overallocation
and/or #34 - Overallocation Occurring In Less Than One Day.

FAQs, companion products and other useful Project information can be seen at
this web address: http://project.mvps.org/faqs.htm

Hope this helps - please let us know how you get on

Mike Glen
Project MVP

 

[Riko Wichmann]

Hi Mike,

Please see FAQ #28 - Resource Levelling Does Not Remove Overallocation
and/or #34 - Overallocation Occurring In Less Than One Day.

FAQs, companion products and other useful Project information can be seen at
this web address: http://project.mvps.org/faqs.htm

thanks for the reference. I thought that might be the reason after I
studied the distribution of work in the resource graph at various time
scales. But it's good to have it confirmed.

Any idea, why with resource leveling Project overshoot in the end date
in second part of my problem?

It's probably because resource leveling is NOT an end-date minimization
algorithm ...

Thanks again,

Riko

 

[Mike Glen]

This is more difficult to answer, Riko, without seeing your file. However,
if you get 2 per week you've twice the amount of work to do in a shorter
period. This, I would expect to overload your resources. Thus levelling
will extend the end date. All levelling does is to delay tasks until the
resources are available. Maybe there's a clash of calendars or individual
resource working times. You'll need to examine the Resouse Usage view to
find your answer.

Mike Glen
Project MVP

 

[John Sitka]

Why would increasing the rate components are fed into the
system reduce or maintain the rate of preparation.

Next step would be: What happens if my components arrive with a rate of 2/week. You would assume, the whole process is AT LEAST as
fast as with 1/week. However, after resource leveling it takes LONGER????

It should take longer, you are doing MORE "stuff" and because if you levelled for a reason in the first place, ("the 1/week"
scenerio) which introduced delay to eliminate overallocations.
More work equals more delay. Study a bit on bottlenecks and Queueing Theory before it hrts too much.

There are two other tools that can be used with delay to get an opitmized solution, By applying weighted performance measure rules
to resources which allow resource substitution and/or iteratively try every possible combination of sequencing that eliminates the
greatest chunk of delay in the whole system. These can take a real long time to run with anything beyond the trivial case, most
often the software is written with "good enough" smarts (heuristics). These aren't available in Project. And since neither really
allows the user to control the results easily they are often used in what is called simulation. It's been my experience that former
accountants, the ones who often authorize checks for operational software, don't believe in simulations, except when it comes to
accruals..., one of life's little ironies that.

 

[Riko Wichmann]

Hi,

Why would increasing the rate components are fed into the
system reduce or maintain the rate of preparation.


It should take longer, you are doing MORE "stuff" and because if you levelled for a reason in the first place, ("the 1/week"
scenerio) which introduced delay to eliminate overallocations.
More work equals more delay. Study a bit on bottlenecks and Queueing Theory before it hrts too much.

nope, we don't feed more "stuff" into the system, because the total
number stays the same. The thinking was, that you have more flexibility
in processing the components (because more of them are available in a
certain time) and by optimizing the resource usage you might be able to
safe time. In particular, since leveling the resources in the 1
component/week case did not in fact shift the end date but managed to
level within the original time frame.

I thought like that before I actually played around with resource
leveling and how it works. Thanks to the discussion here and the links
Glen provided, I know now that we wanted to use a tool for something it
is not intended to be used for.

Thanks again,

Riko

 

[John Sitka]

Sorry Riko I didn't get the notion of the

1/week for 10 weeks

OR

2/week for 5 weeks

The thinking was, that you have more flexibility in processing the components

I agree, but Project can't do it as it dosen't run through the millions of test cases and discover the best fit.
You should be able to manually improve the 2/week case to beat the 1/week case.