Premium
The accuracy of project cost and duration estimates in industrial R&D
Author(s) -
Norris K. P.
Publication year - 1971
Publication title -
randd management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.253
H-Index - 102
eISSN - 1467-9310
pISSN - 0033-6807
DOI - 10.1111/j.1467-9310.1971.tb00091.x
Subject(s) - duration (music) , statistics , estimation , cost estimate , econometrics , constant (computer programming) , mathematics , pessimism , operations management , economics , computer science , art , literature , management , programming language , philosophy , epistemology
. This study of the accuracy of cost and duration estimates, both initial and intermediate, for industrial R&D projects, is based mainly on analysis of the records of 475 projects in four varying research organizations. Mean ratios of actual to estimated cost ranging from 0.97 to 1.51 are obtained, and mean ratios of actual to estimated duration from 1.39 to 3.04. The results are generally similar to those of other British and American studies, with which they are compared. There is no evidence that the information gained as projects progress enables their future cost and duration to be estimated more accurately; at best, the accuracy of such estimates remains constant. No effect of project size on estimate accuracy is found; effects of project length on accuracy, and of time (i.e. increasing experience) on accuracy are found only in one organization each. The pattern of expenditure over time is examined, and found on average to be not far from linear, though with wide individual project variations. When the individual project ratios are reduced by constant factors representing optimistic or pessimistic bias (derived from the mean ratios, and assumed to be characteristic of the firm), and then subjected to a log transformation to make their distribution more symmetrical, the remaining variation, which is a measure of the inaccuracy of estimation for the individual projects, is closely similar in the four organizations, and not greatly different in the organizations covered by other studies when the data are similarly treated.