Cracking the crystal ball

Date of publication: November 7, 2004


Approaches were investigated for forecasting in the developing South African cellular phone market at the aggregate level (market share), and individual level of switching behaviour. While the traditional first order Markov process model was not found to predict accurately, an augmented first order Markov process model incorporating new users provided good predictive ability of the direction of changes in market share. At the individual level, models were able to provide large improvements over chance prediction of individual level switching behaviour a year ahead.

  • PDF
  • This could also be of interest