Media Mandible

I’m disappointed to see that Chris Anderson, author of The Long Tail, has made another post in which he completely misinterprets industry data to justify his ideas about the way online CPMs work. (See also Chris’ last post on this subject, and here’s my comment on it.)

Chris references some PubMatic survey results indicating that large websites are getting lower CPMs these days than smaller websites.

Unfortunately, he chose not to address a note from the PubMatic page: “The pricing data reflects net publisher monetization via ad networks and excludes ad networks’ share of ad spends as well as inventory sold directly by publishers to ad agencies or advertisers.”

Here’s what that means:

• The PubMatic survey only deals with ads sold via ad networks. Ads sold directly by the sites themselves don’t factor in.
• The CPMs reflected in the PubMatic survey are the not the CPMs actually paid by the advertisers, they are the effective CPMs earned by the publishers, after the ad networks’ cut is subtracted.

In other words, the survey results are a good indicator of the amount of revenue earned by publishers from networks per site size, and not much else.

Here’s the crucial insight. Ad space that large sites sell to ad networks is remnant inventory, meaning that it’s whatever is left over after the site sells targeted advertising. In fact, even after a site sells all its targeted advertising (e.g. ads in the entertainment section, or ads targeted to registered users who have indicated that they are women over 40), it may still sell ROS (run-of-site) inventory, which could appear anywhere on the site. ROS placements are not “targeted” relative to anything on the site in question. But they are still targeted relative to ads running on networks, because they have still been earmarked to run on that particular site, and no other.

Ads purchased on networks are, by definition, not targeted to specific sites. Usually, they’re targeted to “channels” such as “men,” or behaviorally-determined categories, such as “shotguns and pickups.” (I didn’t make that up.) These channels and categories use inventory from all the sites they deal with as part of one big goulash. The connection between the low CPMs and the lack of site-oriented targeting is actually the whole point of the ad network business model.

The truth is that large websites often can get higher CPMs than smaller sites for placements targeted to that specific site.

Now, let me take a step back and say that I’m a big fan of buying ads on smaller sites. But most media buyers don’t want to bother with them, because:

• It’s too much trouble to round up enough small sites to make the investment worthwhile.
• Small sites are off the radar of the research tools, like Nielsen’s @Plan, usually used to measure the demographic and psychographic makeup of a site’s user base.
• There is an irrational prejudice, left over from Old Media, that if you put your ad on a big site, more people will see it than if you put your ad on a small site, regardless of how much money you have to spend, or how many impressions you buy.
• It’s easier to convince a client to buy ads on large sites, because it’s more likely that the client has already heard of those sites.
• Smaller sites are less likely to sweep across an entire demographic segment. So, if you have ten million dollars to buy media with, and your goal is to reach every man between 18 and 35 who watches professional sports, then buying on CNN.com might actually be your best bet. This is usually not the actual situation, but it is often the situation that the advertiser wishes were the case, so it can be tough to move away from.

Again, small (and medium-sized) sites can be a great place to run ads. And we are increasingly seeing those sites band together in ways that make it easier to assess the value of their audience, and buy ads across many of them at once, without going through a network (Federated Media being the most successful example).

So why does PubMatic show that ad networks are paying small sites more for ad space per thousand impressions than large sites? Probably because large sites have so much more remnant inventory to sell, that they are getting a volume discount.

One more time:

There is not a trend toward ad inventory on small sites getting more expensive than inventory on large sites.

Or, if there is, Chris Anderson has not presented any evidence that effectively says so.

bird photo by t.j. haslam
boat photo by mckaysavage
dinosaur photo by GregPC