September 28 & 29, 2020 | Virtual
Please login or join to subscribe to this thread
Consider painting a house:
Analogous estimation: House A is about the same size as house B so it will probably take about the same amount of paint as the previous project.
Parametric estimation: A gallon of paint covers X square feet. Based on a house of a certain size, we may calculate the amount of paint required.
Analogous estimates can also be cases such as a hotel is similar to an apartment building, so we can base our hotel project estimate off of what it took us to complete the apartment building project.
I like that analogy, Keith!
I do consider the difference as qualitative vs quantitative.
Keith put this nicely in an example.
Analogous estimating is basically comparing one project to another similar in size and complexity at high level and is used when there is limited information available.
Parametric estimating is based on unit rates per activity. It is much more accurate than analogous estimating when there is data available.
I would probably differ a bit from Andrew’s opinion, as I consider both to be quantitative and the reliability of the results depends on the quality of the data and assumptions.
The difference between both is that Analogous Estimating is more of a Top-Down Approach while Parametric is somehow Bottom-Up unless you’re dealing with one activity only.
Hope this helps.
Keith explained it well, how the both are different in application. A parametric estimate requires an algorithm. An algorithm is developed by researching and taking many (100s, 1000s) of projects and trying to find correlations. An example is CoCoMo by Barry W Boehm, who took a huge book to explain a quite simple formula to estimate mandays for a SW project. Or function points.
The problem is that not all of these projects would be considered analogous, so an algorithm might not produce a relevant estimate for a specific project.
Anyhow, if you apply more than one estimation techniques, it gives you a re-assurance or a reason to dig deeper if the results are to widespread.
I have seen a article by Ricardo Vargas, explaining how to use parametric estimating not with a fixed algorithm, but with AI (neural networks)
I do agree with Keith.
You have stated the answer. Parametric analysis uses an algorithm (typically Monte Carlo) to produce a data set with accompanying probabilities (e.g. levels of confidence). It is a statistical methodology, whereas analogous is typically a single point estimate with no accompanying statistical probability or level of confidence.
It is also worth mentioning that both analogous and parametric estimating can be applied at any level of decomposition (i.e. overall project, work stream, work package, activity). They also apply to resource, time & cost estimates.
I'd also suggest that you can start to provide analogous estimates based on a smaller historical data set than parametric ones unless there is a simple relationship between the x and y variables for the parametric formula.
I came up with a few bad examples before a good one. Between that and reading the comments of my colleagues, if teaching a class I would explain the difference as follows: An analogous estimate is based on an analogy, while a parametric is based on a correlation between 2 or more variables (parameters).
An analogy compares 2 different things that are similar in some way. We can approximate what we need to know about one thing, from what information we have about the other thing. It can be a rough parametric such as, “This job is about equivalent to doing that job twice.”, but not as precise as, “This job is equivalent to 2.14 X that job.”
A parametric uses one or more parameters (variables) to calculate different ones. It could be based on a logical relationship such as Mass = Volume * Density. Those variables are related, but entirely different properties. Knowing 2, we can calculate the 3rd. They can also be a heuristic (something we use because we know it works), based on experience such as many data points.
My favorite example of a heuristic: The cost of some products is primarily driven by the engineering design effort. It turns out the cost correlates very closely to the product weigh however, so if I know the weight, I can calculate both the cost and the engineering effort to design it. There is no logical reason why it works, but it does so we can create a precise estimate, based on a completely unrelated variable.
What I notice about the clarifications is the word “parameter” is not a part of analogous explanation. Although PMBOK includes it and causes confusion.
So say you want to use the labor hours from a past similar project. That will be analogous right?
Please login or join to reply