PubMatic kicks us when we’re down (but gently)

As if we weren’t all feeling gloomy enough already, PubMatic has just released its Q4 2008 Ad Price Index report, which makes for sobering reading. For those of you not familiar with PubMatic, they provide “multi-network optimization” for publishers who are looking to maximize the yield on their remnant ad inventory (i.e. the inventory the publisher can’t sell themselves).

Rather than manually dealing with a handful of networks directly, the publisher hands their inventory over to PubMatic who ensures that the most profitable ad is shown, whether it comes from Google Adsense, BlueLithium, AdBrite, or another network. Since a key part of what PubMatic does is measure the CPMs for online ads, they have access to lots of ad price data, and every quarter they roll this data up into a report (available here as a PDF).

PubMatic has been doing this for 15 months now, and so far, they’ve yet to deliver any good news:

Given the economic Armageddon that overtook the world, PubMatic’s report that average prices only softened by $0.01 during Q4 actually seems like pretty good news. But then again, Q4 was the holiday season; and compared to Q4 2007, Q4 2008’s numbers look pretty horrendous.

The detail of the report contains some more interesting tidbits: for example, the average CPMs for small sites (fewer than 1 million PV/mo) are the highest, at around $0.60, whilst the average CPMs for large sites languish at around the $0.17 mark.

Before you start predicting the doom of the mainstream media, however, it should be pointed out (as Mike Nolet has done) that there is a sample bias in the PubMatic numbers – whereas a small publisher is wholly dependent on ad networks for all their revenue (lacking the resources to sell their inventory themselves), and so is likely sending all its inventory (including the juiciest stuff on the home page) to PubMatic, a large site will only be sending the inventory that they couldn’t sell themselves – i.e. the bottom-of-the-barrel stuff.

It also turns out that average prices for the largest and smallest publishers have slumped by around 50% in the past year, whilst prices for medium-sized sites have remained more solid:

I’m at something of a loss to explain why this might be – at the high end, it may be because large sites are becoming more efficient at selling their inventory themselves, so it’s only the really cheap stuff that is being passed on to PubMatic; whilst at the bottom end, small publishers are becoming increasingly crowded out by new sites.

What would be immensely useful would be for PubMatic to provide some kind of indication of the proportion of inventory from sites that is being served through them; this would make it easier to understand if changes in average prices through PubMatic are the result of a change in the mix of inventory that is being passed to the company. However, I would be very surprised if PubMatic had access to this kind of data.

One more thing…

Once you’re done reading the Ad Price Report, stick around on the PubMatic site a little longer and download their White Paper entitled “Death to the Ad Network Daisy Chain”. This little document does a nice job of explaining how an impression is passed from one ad network to another, and highlights the surprisingly high proportion of ad calls that are returned ‘unsold’ by networks. The document then goes on to talk about how ad operations folk have to manually set up ‘daisy chains’ of ad networks to try to ensure that the maximum amount of inventory is sold. As the title of the document implies, this is held to be a bad thing.

Because of the nature of the business that PubMatic is in, the recommendation in the document is that publishers use ‘dynamic daisy-chaining’ (which is essentially what PubMatic does, choosing the order of daisy-chaining based on expectations about which network will be likely to monetize an impression most effectively) to solve this problem. At one point the document states (my emphasis):

Due to the volatility of online ad pricing … creating a dynamic “chain” of ad networks, rather than a static one, is the only way for a publisher to ensure that they can get the best price possible for their ad space.

I would respectfully disagree with this statement; another way of achieving this is to use an ad network that is a member of an ad exchange, and which can therefore draw on a larger pool of advertisers than just those with whom it has a direct relationship.

But I don’t disagree with the main sentiment of the PubMatic paper, which is that publishers still struggle with significant inefficiencies in the way they monetize inventory; and I believe we’ll see the kind of multi-network optimization solution that PubMatic offers (also available from Rubicon and AdMeld) become increasingly important as the year wears on.