Instead of No Estimates
Instead of no estimates, we should consider adjusting our approach to estimates that eliminate the abuse, and still allows for the answers to the business questions, “does this project improve our bottom line” allowing the business to determine if the company really wants to undertake the project, and if so, do we have the talent and resources to undertake this project. Answering these two questions initiates the next steps to actually create a project and being planning and doing the work.
Besides the techniques below, we can estimate from top down, estimation comes from managers and executives, or bottom up, that is those doing the work or closest to the work, provide the estimates. There are draw backs and benefits to each of these approaches.
There are many techniques for estimating. Experience suggests organizations may not use much more than the least helpful, expert judgement.
- Simulation (Monte Carlo Analysis)
- Expert Judgement
Instead of estimating each time like it were new, we can use our past work experience that is close to the work we are exploring. We look at similar work products and similar projects to estimate what is required to complete this work. For example, our company has developed instrument clusters many times, we may look at the aggregate of that work to make estimates for this new instrument cluster projects. To make analogous estimating work, requires tracking the past projects and historical data associated with those projects.
Parametric estimating uses historical data and statistical analysis of that data to derive a correlation between certain variables and the estimate of the quantity of work. You see this parametric estimating in action when you get a quote for creating an addition to your house when the contractor tells you the prices will be a certain dollar amount per square foot. This estimating approach requires understanding the relationship of the costs and parameters based upon past historical performance.
Program Evaluation Review Technique (PERT) is also known as there point estimating. In this technique we attempt to fit a range of estimates to a normal distribution curve, expressing the duration as a range of possibilities (duration – time) with an associated degree of probability (%) which corresponds to the degree of confidence. It is important to note that this is essentially a continuum of possible duration – the greater the estimate, the more probable.
Monte Carlo Analysis
Monte Carlo Analysis is a simulation technique used to evoke the probabilistic distribution of event(s) through the use of random and repeated sampling and simulation. The simulation can start with those distributions we generated in the PERT method. Through multiple simulations we generate a range of completion dates and probabilities, providing us with a range of possible answers (time) along with corresponding probabilities (% / confidence). In so doing we are able to select the value that represents the risk we are willing to accept. It is important to note that this is essentially a continuum of possible duration – the greater the estimate, the more probable.
Sometimes we can use expert judgement, through mechanism like Wide Band Delphi and to a lesser extent planning poker (planning poker is a team engagement and as such an egalitarian form of Wide Band Delphi). For specialized areas, and when we may not have appropriate historical information to support the estimates, expert judgement is another form of estimating.
Other Things to Consider
It is more than the method, it is also the processes or ways we think in association with the estimating work. These areas below are also worth considering, especially the
- Estimates are subject to revision – constant adjustments – every time we learn something that has implications on cost or time, we update the estimates.
- Focus on the immediate – and keep track of the progress results (Earned Value Management Techniques)
- Do not take months or weeks to detail plan – we delude ourselves when we think we can detail plan months out into the future
- Model and historical data – historical data provide fodder for model development, models make it possible to consider many variables and interactions in multiple explorations (statistical)
- Those estimating are not “responsible” for the veracity of the estimates – we cannot hold people accountable for these estimates as these are estimates and there are times when the scope of the work is new and requires learning to improve veracity.
- Earned Value Management – these project management techniques provide a clear comparison of the planned expenditure and rate of accomplishment with the actual. Paying attention to these metrics (Schedule Performance Index, Cost Performance Index, Cost Variance, and Schedule Variance) provide data that allows for constant adjustment to the estimates based upon actual performance.