Had lunch yesterday with my friend Alan Gould who co-founded IAG Research and recently sold the company to Nielsen.  We talked a lot about the various types of measurement available in Internet advertising and how they influence ad buying.  Alan is working hard to build a measurement platform for brand impact of display advertising and I believe that this is a crucial missing element.  Despite Wenda Millard getting some flack the other day for complaining about the absence of “art” from a world of keyword CPC advertising she is addressing the same point.

Measurement and incentives are closely related.  If you can measure some things well and others poorly and you provide incentives based on those things you can measure you can easily wind up with dramatically skewed results.  A classic example of this are sales commissions.  It is easy to measure the sales dollars (just add them up), but hard to measure customer satisfaction.  Therefore whenever companies aggressively jack up sales commissions they wind up with a short term increase in sales which is then often followed by lasting damage to the brand as dissatisfied customers pile up (following overpromising by the sales folks).

We have a similar skewing occuring at the moment in online advertising and Google is the biggest beneficiary.  CPC is easy to measure.  The benefits from brand impact of display are very hard to measure.  Especially in an economic downturn when ad budgets are scrutinized, ad buyers have no incentive to go with advertising for which they cannot point to a clear ROI.

There are several possible ways to measure the brand impact of display advertising.  Alan and IAG are using a panel based approach, wihch they have honed in testing the brand impact of TV advertising.  While that is likely to work well for some of the very largest display campaigns out there, it is hard to see how it would scale.  But it might be possible to automate the tracking of display’s brand impact.  Using a cookie one could aim to detect how many display ads for a particular brand someone has seen when they arrive at a checkout page.  One could then analyze purchase likelihood based on the frequency of exposure.  That of course does not get at any offline purchases.  For that – other than panels – the only hope might be to run geographically restricted campaigns and attempt to correlate the results.

Bottomline is that shifting towards CPC is where the incentives point to based on measurment but this may be a short term gain for a long term loss.  Anybody who can help fix that by making the brand impact of display more measurable will have a very sizable business.