Following on from yesterday's post on Advertisng Game Theory is this timely reminder of the level of certainty of what we are dealing with when it comes to Online Ads right now.
Greg Sterling's
Screenwerk notes that 2 respected research companies, e-Marketing and Veronis Suhler, have somewhat divergent views:
e-Marketer's 2011 forecast shows the US Online Ad market is $ 7.8bn
Veronis Suhler's Stevenson's is $19.2 bn
A mere 2.5 x difference.......but the key question is:
There are similarly divergent numbers for mobile advertising, which in a way is more understandable given how new and generally speculative the medium is at this point. But how can the VSS and eMarketer forecasts be so far apart? Are the assumptions wrong? Are the underlying data incomplete? Is the methodology flawed?
We do this sort of work ourselves all the time, and the answer is not so much error as choice of assumptions. The most likely thing is that they have most probably made different interpretations and assumptions on (typically similar) existing underlying data sets, and maybe used different methodologies to drive the rates of change over the 5 years.
For example, what method would you use to define rate of growth of online Ads as going forward? You could use historical data of similar industry growths in the past. you could map Ad revenue to something like the online Attention forecasts and proxy that, you can try a Delphi technique and sample a lot of experts in the area. You can build simulation models, nice simple system dynamics ones or huge event based ones. You can do all the above - and guess what - they wont match - probably will miss each other by a mile.
You can even
get numbers wrong!
To be fair, in this case they start off with different positions in 2006 - $2.1bn vs $6.1 bn - so growth rates assumed are not too dissimilar anyway - the issue is why the discrepancy in what is being counted as US online adspend?
The only thing you can say for certain is, to quote one of my Uni lecturers, "all forecasts are wrong, and the further out it goes, the wronger it gets". Or in other words:
- Firstly, that forecasting is an inexact science, especially about teh future - and especially in emerging industries. Small shifts in assumptions and methodologies can drive huge shifts over 5 year projections.
- Secondly, that there is truth in what Ronald Coase noted, ie : "If you torture the data long enough, it will confess" (to say what you wanted it to say)
So what do you do?
On ething is to map your past forecasts to actual to see how good your approach is - Veronis claims they are no more than +/- 2% over the last 9 years. (I'd like to see how that is calculated, if that is forecast v actual on a 5 year future number it is....um..well why waste time writing research reports, there is a stock market waiting.)
The real error is to think there is "one number" for a new industry like this 5 years in the future. Too much can change.
Our usual approach is to use an estimate plus upside and downside scenarios, with "what you have to believe" pointers for both - the idea being that as data comes in you can see which scenario's "beliefs" are most true and can start to track that trend.
Easy way out mayhap, but we think that as the "what you have to believe" approach starts with the concept of a range of error, so it at least starts with the correct frame of mind.
