I've been working on a fair bit of marketing-related stuff over the last few years, and we've been spending pretty big on online campaigns. Much of this money has gone on display advertising, the kind of stuff you'll see on Fairfax and News Corp sites. We've had a whole range of problems with this stuff, and I've come to the conclusion it's more trouble than it's worth. Certainly if you want to measure response based on sales, the rate is pitiful. Here's why it's broken.
When you buy a search ad, you can be pretty certain the person is actually interested in the topic represented by the keyword you're buying. So when someone types "wireless keyboard", it's a good bet they're probably interested in buying a "wireless keyboard" of some sort.
When you buy a display ad, you get pretty loose categorisation. Perhaps your technology product might end up in the technology section of the site, so at least you're being exposed to people with an interest. Or you might be in the "general interest" pool, in which case you're getting exposed to the people who clicked on "celebrity shows boobies" links. Just the people you want. But regardless of the type of category you end up in, you're getting people who are expressly there for something other than finding something about a very specific topic.
So really, you can't measure the results of display advertising by expecting people to buy immediately after clicking, it's more for branding. Or so the salesmen for these mass media properties will tell you. So really you're getting your brand exposed to roughly categorised people.
Except it's completely unmeasurable, and the mass media sites have only themselves to blame for this. You see, if I want to have my brand exposed randomly to roughly categorised people, I need to have some pretty solid statistics on how many have seen it. I'm an Internet advertiser, so I'm used to pretty solid stats, not based on diary entries like TV or circulation surveys like newspapers (yeah, like most copies of the newspaper get read by more than one person, every day).
The problem is the major media in Australia specifically make
any "impression" numbers meaningless, by adding a little line of code
like this to every page on their site, this from news.com.au:
<meta http-equiv="refresh" content="0300" />
This line of code means the page gets reloaded every 300 seconds, or 5 minutes. Fairfax have recently started doing this in a slightly different way (checking if a video is playing first) but they do the same thing.
By reloading the page every five minutes, they've made their impression numbers completely meaningless. People regularly leave their browsers open with a page: the news, the weather, whatever they were just reading. So they go off for a half hour lunch and three "impressions" tick over, except there's nobody home to be impressed. There's no way to quantify how often this happens, so the impression number isn't just inaccurate, it's some unknowable amount completely wrong.
So you see, measured by the only accurate measurement, click throughs and subsequent sales, display doesn't make sense. And as a branding exercise it's not worth playing around with because the publishers have specifically taken steps to make the numbers meaningless.
And they wonder why newspapers are dying, when they go out of their way to devalue the one thing of value they do produce?
11th June 2009 addition: this "study" by News Digital Media just confirms my suspicions. Surveying less than a thousand people when the ad had 300,000 impressions, yeah there's something you can use to draw conclusions! These being people who had just seen the ad (possibly, given what I said above).