Predicting the Path of Technological Innovation: SAW Versus Moore, Bass, Gompertz, and Kryder
Emory University - Goizueta Business School
University of Southern California - Marshall school of Business
Gerard J. Tellis
University of Southern California - Marshall School of Business, Department of Marketing
University of Michigan at Ann Arbor
July 22, 2012
Marketing Science, Forthcoming
Competition is intense among rival technologies and success depends on predicting their future trajectory of performance. To resolve this challenge, managers often follow popular heuristics, generalizations, or “laws” like the Moore’s Law. We propose a model, Step And Wait (SAW), for predicting the path of technological innovation and compare its performance against eight models for 25 technologies and 804 technologies-years across six markets. The estimates of the model provide four important results. First, Moore's Law and Kryder's law do not generalize across markets; none holds for all technologies even in a single market. Second, SAW produces superior predictions over traditional methods, such as the Bass model or Gompertz law, and can form predictions for a completely new technology, by incorporating information from other categories on time varying covariates. Third, analysis of the model parameters suggests that: i) recent technologies improve at a faster rate than old technologies; ii) as the number of competitors increases, performance improves in smaller steps and longer waits; iii) later entrants and technologies that have a number of prior steps tend to have smaller steps and shorter waits; but iv) technologies with long average wait time continue to have large steps. Fourth, technologies cluster in their performance by market.
Number of Pages in PDF File: 54
Keywords: technology evolution, innovation, SAW model, Moore’s Law, Kryder’s Law, Bass Model, technological predictionAccepted Paper Series
Date posted: July 22, 2012
© 2013 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo1 in 0.641 seconds