A recent RFP for consulting services regarding strategic platforms for SAP from a major European company which included, among other things, a request for historical and forecast data for all the relevant platforms broken down by region and a couple of other factors, got me thinking about the whole subject of the use and abuse of market share histories and forecasts.

The merry crew of I&O elves here at Forrester do a lot of consulting for companies all over the world on major strategic technology platform decisions – management software, DR and HA, server platforms for major applications, OS and data center migrations, etc. As you can imagine, these are serious decisions for the client companies, and we always approach these projects with an awareness of the fact that real people will make real decisions and spend real money based on our recommendations.

The client companies themselves usually approach these as serious diligences, and usually have very specific items they want us to consider, almost always very much centered on things that matter to them and are germane to their decision.

The one exception is market share history and forecasts for the relevant vendors under consideration. For some reason, some companies (my probably not statistically defensible impression is that it is primarily European and Japanese companies) think that there is some magic implied by these numbers. As you can probably guess from this elaborate lead-in, I have a very different take on their utility.

My firm belief is that detailed market share statistics are not critical factors for an individual decision on a technology platform. What is important to understand is whether the technology in question has sufficient market share and customer base to represent a “safe” alternative, i.e. whether or not it will continue to be an active and profitable investment for its supplier. Within a wide range, say the top three or four suppliers, depending on the industry and segment, the exact market share and even the ordinal position of the vendor is not as critical as other factors:

1. Specific ISV support
2. Current and future enterprise (user) technology strategy
3. Current skills, processes and investment in related technologies
4. Anticipated vendor strategies and external forcing factors 

Likewise, the forecasting of future market shares is a useless investment of time and the client's money, since they are:

1. For the reasons mentioned above often irrelevant.
2. Usually wrong. None of the supposed expert forecasts for server volumes and market shares in 2000 anticipated the collapse of 2001 – 2003 and the sudden deflation in Sun Microsystems from a run rate of $20 billion per year to barely $10 billion per year. Likewise, nobody predicted the changes wrought by the recent recession, and forecasts as recently as January of this year had no way of factoring in the potential but probably significant effects of Oracle’s precipitous decision to attempt to cripple HP’s server business by discontinuing development of Itanium software. And when the next quarter’s forecasts come out they are guaranteed to be wrong, since nobody yet can accurately predict the impact of this decision. 

I’d be interested in any counter arguments, but absent a compelling case that I am wrong, my personal wish list might include a request that people spend more time focusing on relevant decision criteria and less worshiping the art of elegant statistics.