Three-Point Estimating - Useful or Not?

This forum is for members to share and gain knowledge of Project Management. Got a question about project management? Need help with a problem? Wish to offer tips and advice? Post here.
User avatar
dhaughey
Site Admin
Site Admin
Posts: 495
Joined: Sat 19 Dec 2009 4:39 pm
Location: London

To date, I've relied on task decomposition and brainstorming for my project estimates. This has proved fairly reliable when done in a team meeting. Recently, a colleague suggested I use three-point estimating and gave me this formula:
  • Best Estimate + 4 x Most Likely Estimate + Worst Case Estimate divided by 6

    E = B + (4*M) + W/6

It's meant to give a weighted average that considers both the most optimistic and most pessimistic estimates.

I tried this with my team as part of task decomposition, but the task durations are too short for it to make any real difference. I've now gone back to brainstorming the durations.

How accurate do people find three-point estimating and are there other formulas people use when preparing estimates?

Regards,
Duncan
abrougui
New Member
New Member
Posts: 1
Joined: Mon 04 Oct 2010 8:25 pm

Hi,
Three point estimate and brainstorming are not exclusive... Usually on my project, i use them both. It just give me a more accurate task duration specially when all members of the team do not agree on a specific task duration...
Raouf abrougui
satisfactionuk
Expert Member
Expert Member
Posts: 132
Joined: Wed 08 Sep 2010 1:38 pm
Location: Westminster - London

A lot of industries have books of standard schedules of rates and times for costing purposes. Where this is not available there are numerous time and motion studies from practically every industry that one could access on the net. Failing this, one could conduct a sample time and motion study to ascertain how long and how much effort a sample task would take. Only after exhausting the above resources would I opt for a guesstimate, which is only a rough opinion at the best.

F W Taylor in his studies proved that people would never willingly of their own accord operate at full capacity for any sustained period of time. Herzberg's theory of motivation and Maslow's hierarchy of needs indicates that people need to be motivated and the motivation factor has to exceed their comfort zone and trigger a higher need that causes the increased effort to be sustained. Pareto’s Law suggest that people will instinctively strive to work at 80 percent of their normal capacity and will when asked automatically make a 20 percent allowance in their performance estimates. In industry this level is accepted as being a lower 65 to 70 percent of personal capacity whereby manufacturing firms often class 65 to 70 percent of sustained effort as equating to 100 percent production target and are often willing to pay 30 to 35 percent bonuses in order to achieve 130 to 135 percent of production target or that is in fact a true 100 percent sustained employee effort from the production workers.

This clearly shows that there is no point in asking a person who has a vested interest in a work package to estimate the time or effort needed for its completion. If one does, then one cannot rely on the information without further scrutiny and or arbitrarily adjusting the figures to allow for the above peoples natural tendencies. This also raises a problem with antaganism when the project manager (through his/her experience) having adjusted the time down to a realistic level is then confronted by the worker saying the time is unrealistic, and through determined effort sets out to prove it.

Surely then, it is far better to seek this information from experienced people (internal or external) who are not going to be directly involved in that part of the project. Here, experienced project managers who have an extensive and diverse professional contacts list to call upon for advice have a clear advantage.

In any case, the adage 'archer without a target hits nothing' applies, every person working on the project should have a challenging deadline to work to that also addresses the problems associated with Pareto’s Law. Addressing this problem can be in the form of a time allowance when determining the 'effort time' associated with the task. Additional allowances may have to be made if one was to put the allowance on the 'task duration' where more than one person was working on the task (the multiplier effect). Of course when considering Herzberg's theory of motivation and Maslow's hierarchy of needs one would need to seriously consider the person doing the task to determine what type of motivator (carrot or stick) to use in order to get the job done on time and to the quality required.

Finally there are two pertinant points to make, the first is that project manager must insist, if not demand that workers time having been agreed and allocated by line manager for the project task is not encroached upon e.g if you have been allocated 2 hours of a persons time, you want a full 2 hour and not 1hr 55 minutes; and the second point, as a safety precaution, one could add an adequate contingency time element to the task (undisclosed to the worker) which if not used will cause smiles all around and bring your task in under the project time constraint.

Conclusion: Three-point estimating is a good idea but in my opinion, from the above discussion one should not gain the estimates from the people who have a vested interest in doing the task. Therefore get the estimate from suitably qualified people who do not have a vested interest and then sell the final time allowance to the task team members.
User avatar
dhaughey
Site Admin
Site Admin
Posts: 495
Joined: Sat 19 Dec 2009 4:39 pm
Location: London

With most developers I find I get three types of estimate:

Padded: Where they give themselves extra time by increasing the estimated duration to ensure they can complete the task on time.

Accurate: Where they give an honest opinion of how long they think the task will take.

Squeezed: Where they are worried I won't like the truth so reduce the estimate to a level where they are almost certain to fail.

You get to know who gives which type of estimate.

And then you’ve got the bosses who thinks all estimates are padded and slash them to the bone.

Tricky!
satisfactionuk
Expert Member
Expert Member
Posts: 132
Joined: Wed 08 Sep 2010 1:38 pm
Location: Westminster - London

Wouldn’t it be better if the project manager just gave a true and realistic estimate and costing for his project and then addressed the risk in the place where it should be addressed i.e. in the contingency quotient? That way if it goes over budget or time but stays within his predetermined contingency he could analyse the project properly to find out where he went wrong. Any other method just confuses and confounds the situation and brings the project manager into disrepute resulting in nobody trusting his judgment, capabilities or even honesty in future projects.

To be honest, if an executive or director slashed the budget on my project, I would just smile at them and say fine, tell me what you want me to leave out of the project and I will bring it in at the new costing.

My attitude is simple; I never ever set myself up, or allow anyone else to set me up, to fail.

Project managers must always remember, its their intelligence, skill, integrity and reputation that they take to the next job, never let anyone compromise any of them very valuable personal attributes.

Here’s an afterthought, assuming you have done the above to the best of your ability and been honest in your appraisal of the project, if you concede to a cut and slash manager who says to you 'Reduce the time and or the cost of the project' without reducing the scope, then you are admitting to that person that you got it wrong in the first place and that action alone brings your whole professional being into disrepute. You are not a child handing in homework, you are a professional that should demand the respect and trust of other professionals.
Stephan Toth
Lyaschenko
New Member
New Member
Posts: 1
Joined: Mon 18 Oct 2010 5:15 am

I am generally agreed with previous suggestions but want to add some comments:
- 3 points estimation is good as it helping to collect more accurate estimation comparing with just one estimation. Simply because an estimator spent more time to thing about what could stop to complete a task within best duration. However to achieve better result you would need to educate your team of how to provide estimations. Each difference between optimistic, pessimistic and most probable estimation has to be explained.
Generally speaking it is a risk that something happens during execution. PERT formula has been developed in NASA many years ago for big aero projects and not applicable as a general rule... Even it is included in PMBOK. In same time it could be used as a good start point to develop you one formula for your type of projects.
- An estimation could be provided by same person who will execute a task or by external expert. Advantage of external estimator has been described in previous posts but the issue is that productivity has to be considered. Real duration would vary for different team members. So I would suggest to use internal estimator. However, as mentioned before, based on experience from previous projects you should understand whether a person tend to underestimate or overestimate a work.
- There are scheduling tools which could calculate a schedule based on 3 points estimations and probability of success. Optimistic schedule does not include any negative risks and probability is ~0%. Pessimistic schedule should include all risks and probability ~100% these tools not using PERT formula at all.
satisfactionuk
Expert Member
Expert Member
Posts: 132
Joined: Wed 08 Sep 2010 1:38 pm
Location: Westminster - London

Where other more reliable estimating methods are not available, estimates using the three point method could be enhanced by using MS @Risk for Projects or Crystal Ball whereby you enter any number of estimates that you have been given into say a triangular distribution and run say 1000 or more iterations using either the Latin Hypercube or Monte Carlo simulation algorithm in order to determine a more accurate mean time for the task or tasks under consideration.

Though this level of accuracy is great for large projects with hundreds or thousands of task it could prove to be a very expensive solution for small to medium projects where a standard mean test carried out on a piece of paper may provide sufficient accuracy.

Speaking as an experienced production manager in a variety of industries who is going into project management, at the end of the day an estimate only an educated guess no matter who or how many people contribute towards it. I guess in the end it really comes down to knowing your job and having a feel for how long task should realistically take. Asking other people for their task duration estimates is a way of confirming what you (should) already know and also an additional method of getting them to buy into and accept your final timing decisions.

Stephan Toth
IDF
New Member
New Member
Posts: 1
Joined: Tue 19 Oct 2010 5:37 pm

I am sure this was a typo, the formula is not quite E = B + (4*M) + W/6

but E = (B + (4*M) + W)/6.

I came across it some years ago though did not use it until fairly recently.

I had a contract to get a project back on track. So there was some work to do before we had a complete list of outstanding tasks that we felt comfortable with.

As some others have pointed out, getting good estimates can be tricky. I got everyone together for half a day and we went through each task with a B,M,W for each task, about 140 in total. Reviewed them, then got the weighted averages (and rounded up). All work was broken down so that eventually no task was more than 5 days. It was a very transparent process and I got buy in, as result I think.

There was a 6 weeks duration and we finished on time.

Naturally there were other PM tasks to undertake during that period.

In effect, what happened with this technique was that the over estimates for some tasks compensated for the under estimates for others. And we had a sufficient number of tasks for it to even out.

SO, was this good estimating, good technique, was it the buy in or good luck? Or a combination of the four?

I think it was a combination of the four.

Wishing you all good estimating!
satisfactionuk
Expert Member
Expert Member
Posts: 132
Joined: Wed 08 Sep 2010 1:38 pm
Location: Westminster - London

IDF you certainly know how to sell a sausage lol, I guess a good estimate meets two criteria, 1) it comes in on time and cost and 2) the customer is happy.

Seems you achieved both.

I am still an advocate of scientific management when ever the data is available, however I am warming to your formula as a first stage prediction that could be fed into MS @Risk or Crystal Ball to determine the final mean time after say 1000 iterations.


But to do this one would realistically need two estimates showing the least probable time, the maximum probable time.

I wonder if just putting in your predicted mean would work or would one have to use a different method of prediction. That’s the problem with video tutorials, you cant actually do the experimental work on them so I'll have to wait until I buy a copy.
Kind regards

Stephan Toth
satisfactionuk
Expert Member
Expert Member
Posts: 132
Joined: Wed 08 Sep 2010 1:38 pm
Location: Westminster - London

CRITICAL CHAIN METHOD

I was reading the book 'The Definitive Guide to Project Management by Nokes and Greenwood ISBN 0-273-66397-6 and came across this interesting methodology that I believe is worth your consideration.

Creating a project buffer:

First, Determine the task effort duration by asking the team members doing the task for their end date predictions firstly at 90% certainty and then at 50% certainty.

Then, identify and add up the two columns for the task that fall on the critical path.

Then either do one of other of these two calculations.

a) Divide the estimated totals by two.
b) Take the square root of the sum of the squares of the differences between the 90% and 50% estimates for the task along the critical path chain.

The authors say that option A is easier to do and I guess its more suitable for smaller or less risky projects while option B is more rigorous and may be more suitable for larger more complex and risky projects.

The final method is to insert the feed buffers in those places where non-critical work streams join the critical chain in order to protect the critical chain from any variability in the timing of non critical tasks.

You can calculate the size of the feed buffers in the same way as for the project buffers, using the activities in each feeding work steam instead of the critical chain activities in the calculations.

It is obviously better to read the whole book in order to fully appreciate the concept of using critical chains instead of the traditional critical path.
Regards

Stephan Toth
Post Reply