R.O.M. Auctioneering; Accurately estimate complex projects in one hour.
The process described here allows you to create a Rough Order of Magnitude (ROM) estimate in less than one hour, which has proven to be +-80% accurate to the final figures based on incomplete information. The process allows you to jump from a high level of uncertainty as visualized in the Cone of Uncertainty to a Low level of uncertainty in 7 simple steps.
This practice should be undertake when the following 2 conditions exist
1. This effort / sizing estimate must be provided soon as possible.
2. This effort / sizing estimate must be as accurate as possible.
The process is named R.O.M. Auctioneering. It is so called because we work very quickly throwing out t-shirt Sizes, much like an auctioneering quickly jumps from value to value.
We start with a Work Breakdown Structure (WBS) diagram drawn on whiteboard. Its easy to convert Agile Themes, Epics and even User Stories into a WBS.
1, 2: Divergent Thinking & Negating the Anchoring Bias (5 minutes)
We then agree upon a preliminary Definition of Done (DOD). The DOD is the list of tasks that must be undertaken before a User Story (or task) is considered as complete. Analysis, Development, Testing, Documentation & Peer review are often found in the DOD. Once defined, on Pass 1 [5 minutes], we then rapidly work vertically down the parent / child nodes of the WBS, sizing each node (Epic or User Story) as Small (S), Medium (M), Large (L) or Extra Large (XL). It is important to note that we do not use numbers at this stage, as this may cause the team to fixate on a value. We then move horizontally across until each WBS node has a t-shirt size value.
3: Analogous Analysis (8 minutes)
Now that each node has a size, we process the WBS again. We ask if the sizes of each WBS that are currently listed with the same size is relatively the same size as another with that value.
Pass 2 [2 minutes]: Is this Small WBS Epic comparable to this Small WBS Epic? As much as possible, compare each Small WBS Epic with all other Small WBS Epics.
Pass 3 [2 minutes]: Is this Medium WBS Epic comparable to this Medium WBS Epic? As much as possible, compare each Medium WBS Epic with all other Medium WBS Epics.
Pass 4 [2 minutes]: Is this Large WBS Epic comparable to this Large WBS Epic? As much as possible, compare each Large WBS Epic with all other Large WBS Epics.
Pass 5 [2 minutes]: Is this Extra Large WBS Epic comparable to this Extra Large WBS Epic? As much as possible, compare each Extra Large WBS Epic with all other Extra Large WBS Epics.
As facilitator, its important to try to find the best example of a Small, Medium, Large and Extra Large. You will make use of these later. Its also important to look out for (and mentally note) the ones that the team are hesitant about. Which Small, may be a Medium. Which Medium may be a Small, or may be a Large etc.
4: Parametric Analysis (17 minutes)
Then we look to compare the different sizes with each other.
Pass 6 [5 minutes]: Is this Small WBS Epic comparable to this Medium WBS Epic? Use the ideal ‘Small’ that you made note of before. Also flip between the borderline Small WBS Epics. Test and verify each one. Don’t be afraid to change a ‘Small’ to a ‘Medium’ if needs be.
Pass 7 [5 minutes]: Is this Medium WBS Epic comparable to this Large WBS Epic? Use the ideal ‘Medium’ that you made note of before. Also flip between the borderline Medium WBS Epics. Test and verify each one. Don’t be afraid to change a ‘Medium’ to a ‘Large’ or ‘Small’ if needs be.
Pass 8 [5 minutes]: Is this Large WBS Epic comparable to this Extra Large WBS Epic? Use the ideal ‘Large’ that you made note of before. Also flip between the borderline Large WBS Epics. Test and verify each one. Don’t be afraid to change a ‘Large’ to an ‘Extra Large’ or ‘Medium’ if needs be.
Pass 9 [2 minutes]: Is this Extra Large WBS Epic comparable to this Medium WBS Epic? Don’t be afraid to change them.
5. Bounding the estimates (10 minutes)
Most developers do not like to give estimates. And with good reason. To get around this problem, after reviewing the Definition of Done again, we ask the team to provide two estimates for a Small, a Medium, a Large and an Extra Large. We ask for the Lower Bound (smallest number of that possible size) and the Upper Bound (largest number of that possible size). Note that the number provided can be either a Story Point or a Man Day Estimate. Ensure that the team agree on which beforehand. We then compare each Epics against these ranges to ensure that the Epic fits within that range. If not, then we look to either move the Epic to a different size, or adjust the range accordingly.
6: Establish the Median Average (15 minutes)
Having reviewed all of the Epics against ranges, we then sometimes go further and either determine the average or the median average of each Size and determine if they roughly fit against each Epic within the WBS.
7: Attribute numbers to Epics (5 minutes)
Finally we attribute the numbers to the WBS Epics and total them. We review the notes that were taken and determine where risk lies within the Epics. We determine if contingency is needed and usually calculate this as a percentage of the total. This provides you with a vastly more accurate ROM estimate then the traditional ROM estimate.
Resources needed:
· 1 hour
· 1 Whiteboard & markers (with the WBS already drawn out)
· 1 Scribe (to note down risks)
· 1 Facilitator (to undertake the auctioneering process)
· 1 Knowledgeable team (knowledgeable of the Themes and Epics within the WBS)
Your questions answered
Why do we work on a whiteboard or flipchart and not a computer?
Speed. The estimation process needs to move at pace. Noting this down on a PC is too slow.
Why do we work rapidly?
To foster ‘divergent thinking’ and not concentrate or converge on a specific point.
Why use t-Shirt sizes and not values such as Story Points or Man Day estimates?
We wish to eliminate the effect of anchoring, which is the phenomenon where people tend to provide a ‘random’ number close to the previous number listed or spoken.
The team keep discussing and its slowing down the actioneering process. Is this okay?
No discussion can last longer than 15 seconds. Ideally discussion is not required. ‘Divergent thinking’ requires that the team do not focus on a specific aspect. If a sizing of Small and Large is given, then note both down and move onto the next task.
The WBS node (Epic) already has a size associated with it. Can this be changed?
Yes. Absolutely. Its not uncommon for an WBS element to change 2–5 times during the Auctioneering process.
All of the ROM sized elements are the same. Shouldn’t I just spend longer individually estimating all of these Epics?
In part, this process works due to the Law of Large Numbers. It works on a large number of Epics or Stories, where the +- variances average out over the large number of Epics. If you only have a small number of items to estimate, then this process is not for you.
I hope you find this useful.