I could choose to call the image at the top of this page a tennis ball – it is about the right shape and is a bright colour, but if I chose to do so all that results is confusion! The name we call things matters because it communicates what we are talking about to our audience.
Over the last week we have been dragged into a number of Linked-In discussions focused on questions such as ‘Do you use PERT?’ Partly as a result of our latest blog on the PMBOK 5th Edition and also because Mosaic Project Services (and particularly Patrick) has a high profile writing about scheduling history.
The overriding conclusion from the debates we’ve witnessed is no-one knows for sure who is saying what. Each user of the term PERT may be referring to a Network Diagram, a Monte Carlo simulation, some other simulation or the actual PERT technique developed in the 1950s. If the basic premise of the debate is not clearly defined the net result is a lot of noise and no possibility of reaching a conclusion of consensus – in exactly the same way as playing tennis using the ‘ball’ pictured above, all you end up with is a mess.
I’m not sure why so many organisations and people chose to use names that have a very specific meaning completely out of context but it seems increasingly commonplace:
- it may simply be a lack of awareness, assisted by seeing many similar incorrect usages of the name;
- it may be a desire to look clever by using technical jargon, which of course backfires big-time as soon as someone who knows hears the misuse;
- it may be an overt commercial move to trade off a well know ‘brand’ to make a tool or offering seem better than it is.
What is important to consider though, is apart from the first option none of the other factors are ethical behaviour and all of the factors destroy effective communication.
To help bring some level of knowledge into the discussions around PERT, we have published a White Paper today on ‘Understanding PERT’. This paper outlines exactly what PERT is and was, identifies the shortcomings in the technique and delineates what PERT ‘is not’ and the reasons why. After everyone has understood what PERT is and is not, let the debates continue but this time with the effective communication of ideas between the protagonists, ie, a communication in which the receiver actually understand what the sender is meaning.
The alternative was effectively described by Robert McCloskey, a US State Department Spokesman several years ago ‘I know that you believe that you understood what you think I said, but I am not sure you realise that what you heard is not what I meant!’ Effective communication needs a mutual understanding of the terms used.
Pat Weaver
Lynda Bourne
Mosaicproject's Blog 
A couple things of interest. Program is misspelled ;>)
The reason the PDM prevailed is the tools used before computer diagramming of the network used push pins in slotted rows and colored string. The connections between the “pinned activities,” could be moved around inside blocks of time. So the arcs could not have information on them.
Until the mainframe plotting programs came along the maintenance of the “board” was the responsibility of the junior schedulers.
PERT of course has a sorid history, having be reverse engineered to fit the needs of the Navy. Ignoring for the moment the “fixed” standard deviation in the denominator (a+4b+c)/6, it can be up to 27% unfavorable (optimistic) of anything but the simplest networks. This is a government contracting the Monte Carlo Simulation is used (DI-MGMT-81861) and not PERT. http://archive.ite.journal.informs.org/Vol6No1/Williams/
This is a very helpful contribution. It is also worth looking at the original paper, which can be found at http://mech.vub.ac.be/teaching/info/Ontwerpmethodologie/Appendix%20les%202%20PERT.pdf.
A few observations:
1. It is clear from the original paper that the authors understood the limitations, but justified the technique on the basis that doing it properly would not have been computationally feasible at the time.
2. While the paper is not very clear on this, I think they applied the formula for the expected durations _before_ calculating the critical path, not after as implied by Mosaic’s white paper.
3. My understanding is that they chose the particular subset of beta distributions to have a kurtosis similar to that of the normal distribution, and that the parameters of the mean and variance formulae follow from this.
4. The Mosaic white paper mentions some other distributions commonly used in schedule risk analysis, but omits the lognormal. There is some empirical evidence that the lognormal is often the best fit to historical data.
Thanks for the heads up Tony.
Our White Paper has been updated to align with the 1959 original publication and the ‘PERT critical path’ is indeed based on the ‘expected’ duration not the ‘most likely’.
The original 1959 paper makes a fascinating historical study.
Pat.
Having been involved in the very first days of Critical Path Method working for BTM a British computer company being involved with Kelly and Walker and their work for Du Pont and shortly afterward simultaneously being involved in PERT and development of a number of Critical Path / PERT software packages, it is nice to see that the need to keep to the record of what happened is maintained in an age where many people seem to call a shovel a spade.
For some reason today, renaming or changing meanings is quite common,… yes I know langauge changes with time and different countries have different spellings for common terms …
I had the good fortune at the age of 18 to be involved in the creation of some of the first software produced to calculate Critical Path and then a year or so later working on the PERT system, which the wonderfully named Booz Allen and Hamillton in conjuncion with RCA, Lockheed Missiles and Space company developed PERT as we know then called Programme Evaluation and Review Technique.
As is well documented he PERT system came about as a result of the US Navy being tasked with the development of a missile to be fired from a submarine, which as we all know was named Polaris.
Booz Allen and Hamilton developement of PERT was (I suspect) based initially on Kelly and Walker’s original work using arrow diagram, but because of the nature of the project and with the uncertainty in estimating durations ,the beta distribution (better known in the US as a Bell Curve ) was used to created an “Expected Time” from a) Optimistic, B) Most Likely and C) Pessimistic times. Thus these were known as “Three Time Estimates”, which both Microsoft and Oracle for some reason now call Three Points
Pat and Linda are quite correct about calling a spade a spade, and as Patrick well knows I get quite toey at original names getting changed for change sake.
Anyway Pat’s paper is another good contribution to the maintenance of our history, even if Glen is upset at Australian spelling !!! 🙂
Raf
Forgot to mention that the USN issued a spec around the PERT system back in 1957 or so numbered Mil-P-23189A PERT/TIME for industry to take up the concept
Raf
Fascinating updates…I have been reading Malcolm, Roseboom & Clark (Operational Research 7(4), 646-69) and Fondahl (PMJ XVIII(2) pp.33-6) which compliment the contributions to this blog. As one interested in the intricacies of project management history, I do appreciate these.
In that original paper by Malcolm et al they promise a follow up paper by Clark on how to do the analysis properly. That paper can be found at http://www.eecs.berkeley.edu/~alanmi/research/timing/papers/clark1961.pdf. He has to assume normality, in contradiction to the beta assumption of the original paper, and tries to justify this approximation but I believe the justification to be bogus.
The math is quite intense and as he says it becomes totally intractable with more than two merging paths. (You can always treat a merging of more than two paths as a succession of merges of two, but this compounds the error due to the normal approximation, since even if the original paths have normally distributed lengths, the merged path won’t.)
All in all it is an eloquent argument for Monte Carlo!
Thanks Tony – I’ve updated the White Paper at http://www.mosaicprojects.com.au/WhitePapers/WP1087_PERT.pdf with your feedback.
It reminds me of a quote by St. Augustine (354 – 430) “We should beware of mathematicians and those who make false prophecies. The danger is they have made a pact with the devil to darken the spirit and confine man in the bonds of hell”.
This work was incredibly valuable and ‘ground breaking’ in its day but does seem to have confined poor mathematicians to ‘the bonds of hell’ ever since computers became powerful enough to run Monte Carlo simulations.