Craig Kolb, business analyst at Ask Afrika, claims that discarded models such as the Markov model, which saw its heyday during the sixties and seventies until it was discredited by other researchers, have shown to be applicable to markets not considered by the old critics: "Unfortunately, many research agencies have a large contingent of client-facing staff, but only a handful of statisticians at the core to handle the complexities of tailor-made analysis.
"Sometimes, however, one needs to question the status quo view on things and dig below the surface. While the Markov model doesn't apply to fast moving consumer goods markets, our initial findings indicate it may after all be a useful framework for understanding behaviour and forecasting in markets such as the cell phone market."
Kolb explains: "As South African cellular phone penetration growth continues, operators are hard pushed to grab new subscribers entering the market while simultaneously preventing existing ones from switching to competitors.
"Additionally, in a dynamic market, operators need to know whether competitors' claims that they will reach 25% market share in a year are realistic. Isolated reports of churn, or intention-based estimates of churn are insufficient. A model that can provide a more holistic behavioural framework is needed - which is why we tested an adaptation of the Markov."
Ask Afrika also conducted an additional study to find out what happens to consumers after a survey is said and done, a factor which, Kolb says, is rarely considered by market research houses: "We were curious. Do consumers do what they say they are going to do? This is an important question because many business executives rely on standard models applied using this type of data to identify the drivers of consumer behaviours.
"The results were surprising. Of the 600 consumers followed up, two thirds of those who said they would switch cell phone operators over the next year ended up staying. Not leaving it there, we took another approach and found an alternative which improved substantially on this dismal record.
"We've also found that the academic literature has tended to be ignored - literature which indicates there are better alternatives to intention scales, and in some cases alternatives which are equivalent to expensive proprietary models. In one interesting case, a publicly available alternative was reported to perform just as well as a proprietary alternative available internationally. This highlights the need to not accept proprietary models at face value - they should be proven."
