NONE: Re: ONLINE-ADS>> CTR drops over time
Re: ONLINE-ADS>> CTR drops over time
Ivan (egben_at_dockside.co.za)
Wed, 1 Jan 1997 09:17:26 +2000
> "Ivan" <egben_at_dockside.co.za> asserts:
>
> > It doesn't matter (statistically) how many millions of adviews were
> > included in a study - if the study was not conducted within the
> > framework of an experimental design, enabling isolation and
> > statistical testing of determinant variables, it's worse than useless
> > - not the least because decisions may be taken on invalid
> > conclusions.
David Beroff replies -
> Well, yes, scientifically "pure" numbers are certainly "better"
> than those derived from a slightly less careful methodology. But,
> quite frankly, these numbers are better than nothing at all. ...
> ... those of us with *some* information generally make more money
> than those with none, and there's never anything that's a sure
> thing. As Ms. Frizzle (kids' science program) says: Get messy!
> Take risks! Make mistakes!
I agree that some information is better than none - however it's not the
quantity of information but the quality which is of concern. The
marketing case books are riddled with big number research which is
just plain wrong due to biased samples or sloppy methodology. The
most famous of these was the forecast of a massive win for Dewey in
the post-WW2 presidential election, based on a sample of 4 million
readers of a magazine. Of course Trueman won. (The magazine had
mainly Republican readers).
One reason I queried the doubleclick numbers is because I suspect a
heavy bias towards banners placed on search engines - which may well
be the worst place for banner advertising. The other reason is
because I feel an underlying negativity, almost a sense of
hopelessness in much of the correspondence to Online-Ads, and much of
the negative news needs a closer look!
There really is a need for fundamental research. 1996 probably saw
about $100 million spent on web advertising, which is a drop in the
ocean. Forrest Research forecast $30 billion a year by the year 2000.
However much media buyers speak of creative and qualitative media
selection, the harsh reality is that advertisers base most of their
media buying on hard numbers and valid information.
> > Our own recent experience with ILE does not indicate a drop off of
> > CTR if banners are rotated. We rotate four banners. During the past
> > three months on 56000 impressions the CTR is 21.6:1, our current
> > banner (due for a change) on 8600 impressions has a CTR of 21.4:1.
>
> You and a few others have mentioned rotation. While this certainly
> makes sense to me if the number of banners or the number of web pages
> were very small (i.e., combat the boredom factor), what's confusing
> me is that the ILE has very large numbers of each. The chance that
> any specific person sees a specific banner twice at all is extremely
> low. Thus, rotation shouldn't have any significant effect. Any
> thoughts as to why it seems to?
>
Yes. Either the notion of rotation on ILE is an illusion, or ILE
banner placement is not random. We are testing the first by holding
our present banner way beyond the point where we would normally
change (so far it's holding up well). Only ILE can enlighten us as to the second.
Ivan Weltman
TUDOGS: The Ultimate Directory Of Gratis Software.
http://www.tudogs.com
