Predicting Program Performance

[Phil]

Can Project extrapolate the start and finish dates of future tasks based on
performance to date?

Most Program Managers want to know how badly the end date or critical
milestones of their program will be impacted by less than perfect
performance. Do they have a major or a minor problem to deal with and where
should resources be applied.

I understand there are some serious assumptions in this calculation and that
the correctness of predecessors and successors will determine the outcome. It
will still have a lot of utility.

Can Project do this?

 

[Andrew Lavinsky]

The answer probably depends on what you're looking for.

For instance, if the schedule is created with the appropriate logic in MS
Project, then past performance will shift future tasks, and thus give you an
extrapolation of future schedule based on current performance. If you're
doing Critical Path Scheduling though, unless you've put buffers into your
schedule, you're most likely extrapolating the most optimistic outcomes from
where you are now. This approach does predict the future, but it may not
automagically take into account past performance issues.

If you wish to take past performance into account, you then have a couple of
options. Earned Value is one approach. This will give you an SPI, which you
can divide future task duration by to derive some sort of estimated
completion date. To do this on a task level isn't really supported natively
within MS Project, but could probably be done fairly easily with a macro to
drop it into Excel - or with a macro to drop that calculated duration into
one of the PERT durations.

You could also take the same approach but with an Earned Schedule
flavor(www.earnedschedule.com), which also has some strong adherents - and
some MS Project plug-ins.

And finally, a lot of folks would recommend you just use a Monte Carlo
add-in to MS Project. This doesn't necessarily take into account past
performance, but allows you to predict the unknowns remaining in the project
based on past performance, and then extrapolate future tasks from that.

 

[Jim Aksel [MVP]]

Adding to Andrew's informative post.
To apply earned value, you will need costed resources assigned to all
discrete tasks and a baseline. This will give you an SPI once you set a
status date and appropriate %Complete on the tasks. You can then use a
spare custom duration column formula as [Remaining Duration]/[SPI] Be
careful if SPI falls below about 0.8 ... think about what happens when you
divide by a very small number.

As for Monte Carlo, that is probably the soundest approach since you are
forced to think about the maximum likely remaining duration. The only
Add-In that I know for Project MonteCarlo is Delek's Risk Plus
www.deltek.com. Other more robust products are available but are stand
alone (expensive) solutions.

Jim

 

[Rob Schneider]

Adding to Andrew's informative post.
To apply earned value, you will need costed resources assigned to all
discrete tasks and a baseline. This will give you an SPI once you set a
status date and appropriate %Complete on the tasks. You can then use a
spare custom duration column formula as [Remaining Duration]/[SPI] Be
careful if SPI falls below about 0.8 ... think about what happens when
you divide by a very small number.

As for Monte Carlo, that is probably the soundest approach since you are
forced to think about the maximum likely remaining duration. The only
Add-In that I know for Project MonteCarlo is Delek's Risk Plus
www.deltek.com. Other more robust products are available but are stand
alone (expensive) solutions.

Jim

Can Project extrapolate the start and finish dates of future tasks
based on
performance to date?

Most Program Managers want to know how badly the end date or critical
milestones of their program will be impacted by less than perfect
performance. Do they have a major or a minor problem to deal with and
where
should resources be applied.

I understand there are some serious assumptions in this calculation
and that
the correctness of predecessors and successors will determine the
outcome. It
will still have a lot of utility.

Can Project do this?

I think the best approach for looking at the future is to do it with
Monte Carlo simulations. I say that because setting it up makes you
really think about the reality of the project. It is not just numbers
and statistics.

In addition to Jim's comment above, there appear to be a lot of these
products each with their own small markets and followers. More keep
cropping up into my visibility all the time. I've been using the @Risk
for Project product for many years. It's an add-in to Project. I also
like Risky Project but it's a stand-alone.

However, as Philip Alacabes concluded and wrote in his superb book
"Dread": "We should stop listening to scientists who claim to be able
to foresee the future." Project forecasting is a) required, b) risky.