Google Chrome OS, Meet Mr. Markov

The Web is ablaze with commentary about last week’s announcement by Google of their Chrome OS, a new operating system scheduled to be released in several months. In the newsletters and blog feeds that I receive, I probably counted two dozen entries on the day after the announcement. Articles fall into two distinct camps: this spells the end of Microsoft; and Chrome will prove to be just another “also-ran” in the many attempts to unseat Microsoft from its dominance.

Partisan commentary aside, it is an interesting question that we face in marketing all of the time: how will a new product entering an established market fare? Can it acquire enough marketshare? How long will that take? Or will it barely make a dent.

One very useful tool in such analysis and planning is Markov modeling (named for the Russian mathematician Andrey Markov). Markov models are simple in concept (although there are a great  many variations), and while a thorough explanation is not possible, here is a summary. A Markov model consists of a number of “states” and a set of probabilities that define possible transitions from one state to another. Here is a simple model that we see in often in marketing:

In this example, two percent of prospects convert to first-time buyers per month and of these 60% proceed to a second purchase. You can see the transition probability matrix in the table below the diagram. In this case, the transitions are largely linear and one-way, but in most Markov chains, transitions would potentially apply between all states.

Markov models have many more features than I can cover here, but they are a very powerful methodology for understanding and optimizing marketing systems. There is plenty of online material if you want to learn more…or feel free to send me an email.

So what does this have to do with Chrome OS? The same core methodology can be used for forecasting market share changes, something that we do frequently for our clients. Here the states become the competitors in the space, and the transitions become the changes in market share. Here’s an example of a scenario we recently built:

For a few years, this market was a duo-poly until Brand C was introduced. In the year Brand C was introduced, Brand B had the highest brand loyalty, losing only 8% of its customers per year. Brand A lost 15% of its customers to Brand B, while Brand C lost 13% of its customers to Brand A.

With the entry of Brand C into the market, who should be concerned? Certainly Brand B has the most to lose, as it begins with both the highest share (67%) and highest loyalty (92%). Here’s the output of the Markov model we used to forecast compared to what actually happened (the model actually predicts 15 years, but we’ve shortened the table for the sake of brevity.

As one can see, the Markov model provides a good forecast of market share six years after the entry of Brand C using very minimal data. Overall, the news has been good for Brand B: it has lost some share, but much less than Brand A in both absolute and relative terms. But even for Brand A, the worst seems over, as the market appears to have reached equilibrium.

Of course, it is much too early to forecast the fortunes of Google’s OS – after all, it will not be available commercially for some time. And in the meantime, Windows 7 is being introduced, Apple continues to maintain its marketshare, and the many flavors of Linux are available. But I’m sure a lot of smart people at Microsoft and Google have been working on forecasts, and chances are they are using Markov models in their analysis.

Comments are closed.