The Page View Problem: Gawker Sacks Editor
February 27th, 2008 | by Brad King |Gawker Media fired one of its editors today because she wasn’t bringing in enough page views, a killer for an online media company.
I don’t know the particulars of Gawker Media’s business model, so I don’t know whether it was a smart move. However, I’ll be at their party next week at SXSW Interactive, so let’s assume that it was a very good move.
Page views means money, right?
Not exactly. At least not in a traditional media setting. Last week while I was giving one of my presentations, a journalist asked me how my discussion fit with editors who became obsessed with page views to the point where editorial decisions were routinely made with only that in mind.
The answer I gave: the way advertising software works is that big page view spikes aren’t rewarded by systems such as DoubleClick’s DART (which comes up looking funny with AdBlock turned on) withoutb some serious human controls because the system delivers advertising campaign’s based in part upon historical page view expectations.
So editors who go for the big spike are either: 1) giving away loads of extra impressions, thus driving down their CPM and lower expectations for future advertisers; 2) delivering tons of house advertisements, which lowers your actual CPM, thus lowering expectations for future advertisers; or 3) has a trafficking manager ready to jump out the window.
This, in other words, is statistically worse than the forward pass, where at least one good thing can happen.
Now, we used a third-party tool — DART — which may have limited our ability to serve paid advertising during times of spikes; however, tracking those spikes is risky. You start making long-term decisions based upon second-to-second actions and you can very easily find yourself in the shit. These are still statistical models, so they aren’t equipped to deal with traffic spikes. They can’t take into account a story about Paris Hilton and compare that to 9/11 and create a statistical model that can deliver ads effectively.
That means editors making decisions on page views based, if they are using statistical model based advertising delivery along with second-to-second “fixes” for traffic spikes, are tacking quite a rabbit (as we say in Austin).
It also means they may not understand the technology in their business enough to use it to make decisions (which might be a bad idea anyway). Gawker, of course, likely has a different model: sponsorships possibly, regularly planned updates to train readers for better statistical models. Who knows.
I don’t know if that was a satisfactory one for the journalists in the room, but it’s a truth that I know to be, uh…true.
