China
Africa
Explore chapters and articles related to this topic
Cost estimating accuracy
Published in Dennis Lock, Shane Forth, The Practitioner Handbook of Project Controls, 2020
The idea behind reference class forecasting is to overcome the optimism bias during the early stages of project definition and improve the degree of confidence in the cost estimate and project schedule. It comes from Daniel Kahneman, an Israeli-American psychologist and economist and Amos Tversky, a cognitive and mathematical psychologist. Kahneman also has a lot to say about how the tendency towards optimism can frustrate any attempt to produce a cost estimate that has a reasonable degree of confidence. See Lovallo and Kahneman (2003).
Schedule delays of major projects: what should we do about it?
Published in Transport Reviews, 2021
As a recommendation to address this cognitive bias in planning under uncertainty, Kahneman and Tversky proposed that decisions be supported with reference class forecasting (Kahneman & Tversky, 1979). Reference class forecasting, also known as the outside view, encourages to evaluate the experiences of a class of similar projects, to create a rough distribution of results for this reference class, and then to place the current project in that distribution. Doing so prevents the decision maker focusing on projects which are successful and close in time and space to the decision at hand and further suspends the natural and often cognitively flawed judgments. Reference class forecasting is more likely to generate accurate forecasts, Lovallo and Kahneman say, and far less likely to deliver one that is highly unrealistic (Lovallo & Kahneman, 2003).
A cross-national comparison of public project benefits management practices – the effectiveness of benefits management frameworks in application
Published in Production Planning & Control, 2020
Terry Williams, Hang Vo, Mike Bourne, Pippa Bourne, Terry Cooke-Davies, Richard Kirkham, Gordon Masterton, Paolo Quattrone, Jason Valette
Although both optimism bias and gaming were well-recognized concepts, no common solution emerged from our study. Some used sensitivity analysis and some used external estimating teams, but the UK was the only country mandating reference class forecasting. As with ex-post learning discussed above, a pre-requisite for reference class forecasting is an organization that collects and maintains the appropriate benchmark data for generating the appropriate forecasts, and this is recognized outside government projects (Fouché and Rolstadås 2010; Ochieng et al. 2016). In the UK resources have been dedicated to achieving this but future research should focus on why this approach has not been adopted more widely.
Impacts of human communication network topology on group optimism bias in Capital Project Planning: a human-subject experiment
Published in Construction Management and Economics, 2019
Two interesting observations in our experiment deserve further exploration. First, our findings highlight the importance of showing the assumptions and considerations previous estimators held when making estimating decisions in the reference class forecasting process, in addition to the historical information of previous projects. The reference class forecasting method requires decision-makers to identify a similar project and use the estimates of the similar project as the baseline for new estimates. Literature suggests that providing information about similar projects will help control optimism bias when making predictions on the present project (Flyvbjerg 2007, 2008). However, it should be noted that due to the technical uncertainty of capital projects, it may be difficult for decision-makers to evaluate the similarity between two projects. For example, as illustrated in Figure 8, for complete piping, even when projects are of similar magnitude (LF – linear footage), the production rate could vary by more than 300%. In our experiment, this means that even though the students could find a similar historical project from the historical database (Figure 4, Zone C), they still need to pick a number from a wide distribution (Figure 8). Moreover, our previous work also found that even when more parameters were introduced (e.g. job site configuration), the similarity between two projects was still extremely difficult to measure, probably due to the vast search space (Du and Bormann 2014). As a result, the main assumptions and considerations pertaining the selection of similar projects should be incorporated in the reference class forecasting process as they indicate the dimensions of similarity measurement (Flyvbjerg 2008).