The subjective and unknown nature of projects often lead to use of subjective techniques and estimating is part of the process. Estimates are a key activity of any project, specially more so in the beginning of projects when things are still cloudy and murky. Estimates are subjective because they tend to fall victim to bias, and the high degree of variability associated in an initial estimate. Estimates are mainly done for cost and durations involved in schedules and it is also used in doing feasibility studies before projects begin.
Before we dwell more on estimates, let’s discuss about estimate qualities.
- Validity: If cost basis is valid, resulting estimates will be valid
- Reliability: methodical process instead of an unorganized effort
- Accuracy: Always mention the accuracy to let readers know the range of error
- Precision: The significant figure used in making estimates (to the nearest hundredth, etc.)
Quality of estimates improves over time, as more project elements and well defined and refined. Based on this level of definition, the accuracy and effort required to make estimates is mostly as given below
|Class||Level of project definition||Purpose||Method||Accuracy relative to class 1||Effort relative to class 5|
|5||0-2%||Screening or feasibility||Stochastic or judgement||10 to 20||1|
|4||1-5%||Concept study or feasibility||Primarily stochastic||5 to 10||2 to 4|
|3||10-40%||Budget authorization or control||Mixed but primarily stochastic||3 to 6||3 to 10|
|2||30-60%||Control or tender||Primarily deterministic||2 to 3||5 to 20|
|1||50-100%||Check estimate or tender||Deterministic||1||10 to 100|
Two approaches are adopted when making estimates. If you have a Work Breakdown Structure (WBS) in place, you could do the more exhaustive bottom up estimating, which would yield a more accurate and objective result. Such estimates usually require resources, time and patience. Bottom up estimates are also known as micro estimates. Bottom up approaches are mostly deterministic.
There are many different bottom up approaches to choose from.
- Detailed estimating: Requires work components to be subdivided into Work package details. Estimators should be familiar with each of these work packages
- Template method: Detailed estimate from previous projects are used. Difference in scope is adjusted for in a template, typically a spreadsheet. This bottom up approach is quick and easy
- Parametric or unit rate estimating: Use organizational (or industry) standards for each work component. Its is an extension of the template method, but instead of considering one detailed estimate, this method tries to evaluate all the available estimates to come with unit rate
- Range estimating: Used when work packages have high level of uncertainty. This method provide three estimates based on low, medium and high uncertainty. A risk level is associated with each work package based on low, medium and high occurrence. Combination of these two dimensions helps in coming with a final estimate
- Probabilistic estimating: Arguably the most accurate type of estimating available and allows to provide estimates with a confidence level. Separate software is used to simulate the result based on uncertainty and correlation of all related work packages. Monte Carlo simulation has almost become synonymous with this approach, even though it is not the only method.
But bottom up approaches are tedious work and only works when the project definition has significantly progressed. What happens when you have to make estimates at conception stage of the project. Without a WBS, none of the bottom up approaches work and we have to depend on the more stochastic top down approach. Top down approaches are quick, easy, requires less people but has the disadvantage of being inaccurate. They are also known as macro estimates, ballpark figures or guestimates, because usually the estimate is just guesswork. These estimates are stochastic, meaning they are inherently subject to changes
Similar to bottom up approaches, there are many top down approaches to choose from
Comparative or analogous estimating: Compare projects with previous similar projects based on standard or grade. Any difference is adjusted for
- Consensus methods: A group of experienced people sit down and compare estimates and arrives at a consensus. A popular consensus method is the Delphi method which is a more refined consensus method
- Parametric or Ratio methods: Another quick method where project parts are compared to other similar works
- Apportion methods: Is similar to parametric or ration methods, and based on projects of similar scope and size
- Function point method: Primarily geared towards IT projects, one could argue that the method can be adopted for other industries. Function point analysis is a widely scrutinized method but yields good results
- Learning curves: A mathematical curve used to predict pattern of cost and time reduction as certain works are performed over and over again. These curves form the basis of the estimates for different components
Whether you use top down or bottom up approach will largely depend on time, available knowledge and resources. One thing is clear. Estimates should always be refined, continuously, even during implementation. The improvement in accuracy is highlighted in the diagram belowSome other things to consider when preparing estimates:
- Who is responsible for the estimates?
- Who will be involved in the process?
- Assume normal conditions, eg; work hours, work days
- Units of cost and time
- Contingency should not be included in estimates. This is to prevent double counting