Ultimately time, resources and money are the three criteria by which society and organizations decide what gets build, how it is built and by whom and in consequence, the importance of having credible and reliable estimates cannot be overstressed. Unrealistically small budgets as well as unnecessarily large ones result in misallocation of resources and wasteful spending.
The effort, time and resource estimates for a software project are the range of values within which, an organization believes it is possible for it to achieve the objectives of the project with a defined probability and without jeopardizing its viabilityEstimates, are typically used to:Determine the economic feasibility of a projectEvaluate alternativesEstablish a project budgetAn estimate is merely a prediction of what is most likely to happen. There is no implication that the estimator will attempt to shape events so that the estimate is materializedAn estimate and a budget are two different things
What are possible explanations for this?
Cognitive bias -> The estimates are biased but we don’t do it on purposeIndividual and organizational behaviors -> The estimates are not bias but we don’t observe the overestimations because the behaviors result in the spending of all the effort associated with themPolitical behaviors -> Strategic intent
The more important and bigger the project, the less the cognitive bias can be used as an explanation for bad estimates. The reason I make this point here is so you understand the importance of creating a defensible – self evident estimate
We examined software project bids made by 35Norwegian and international software companies. Thebidding companies consisted of large, medium, andsmall development companies operating in Norway.All bids are in NorwegianKroner (1000 NOK is about $140 – October 2004).The mean value of the bids was 220 000 NOK, rangingfrom 21 000 to 560 000 NOK. As far as we haveobserved in other bidding rounds, this range of bids isnot uncommon when the number of bidders is high.
Who prepares the estimates? The advocates (Olympic committees)
The Freiman curve links the actual cost to the budget selected for the project. Frank Freiman, its inventor, was for many years the head of cost estimating for RCA and the developer of the FAST cost-estimating system. Succinctly stated:The greater the underestimate, the greater the actual expenditure;The greater the overestimate, the greater the actual expenditure;The most realistic estimate results in the most economical project cost.Small budgets (prices) may land a contract or result in a project approval, but they also frequently lead to financial loss and business failure. Initial project plans of staffing, scheduling, machine processing, tooling and materials' forming, etc., are not achievable. Though the project plan is established to realize the underestimated cost, the project mid-point management begins to realize that milestones and schedules are slipping. In response, there is reorganization, replanning, and possibly the addition of personnel and equipment. Delays and reorganization invariably increase costs.The cost to the organization is also high in other ways, including poor morale and the loss of capable and trained staff. Projects that suffer significant cost growth are often projects scheduled, planned, and staffed based on early underestimates, that eventually lead to a detailed project plan that simply cannot be realized. Underestimates threaten an organization's ability to survive.Needlessly high budgets serve an organization as poorly as the underestimate. Rather than resulting in greater profits, as one might hope, overestimates conjure up Parkinson's law: the money is available, it must be spent. Unless there is firm management control, the estimate becomes a self fulfilling prophesy and the organization becomes weak, unable to deliver a good product for a reasonable price.Realistic estimates result in the most economical cost. They remind managers to control the excess resources. Good estimates let the organization's resources work in harmony.The Miranda curveThe point here is everybody can produce an accurate estimate by making it arbitrarily large. However if the budget is to high you might jeopardize the feasibility of the project, in other words your client chooses another provider or you don’t get the funds.Budget choices influence the cost of the projectOn the left part of the curve we see that the effort necessary to maintain the commitments made is greater than if the work had been planned from the beginningOn the right size part of the curve we see the consequence of choosing an unnecessarily high budget. The project becomes too expensive and in consequence it does not get funds or another supplier is selected
Eduardo Miranda (Carnegie Mellon) Overruns or Underestimates? A Political Perspective on Software Cost Estimation