Guest Blog Post: It’s A CPM World After All…

By Susan Waldes
Director of Client Services, PPC Associates

CPC-based advertising is a very elegant model of determining value and ROI for advertisers and brings an amazing amount of control at a very granular level. It’s much easier to get our heads around, easier to sell, and easier to optimize to than a CPM-based model. It’s also relatively fair.

But it’s not the way the digital advertising world works.

As digital advertising and digital marketing have become more and more trackable and adherent to direct-response metrics, you rarely hear CPM come up as a “thing” in CPC channels. Reality is, though, that it drives the whole machine. Lurking beneath all the seeming “fairness” of a CPC payment model is a big, gnarly, machine-algo-driven world of CPM that is still controlling most everything that happens in the digital space.

In this post, we’ll explore an example of this hierarchy, how it affects marketers, and why search isn’t immune. It might not be the most uplifting read, but at least we’ll know the score.

New York Times is way into CPM.

Last week, a colleague posted in our agency project management space that he’d seen GDN traffic from nytimes.com drop dramatically for one of our clients in the last few days. He was wondering if anybody else had seen the same pattern on other accounts.

There are a few possibilities for why this might have happened:

1.  NYT blocked the advertiser
2. The client got outbid by competing advertisers
3. NYT shifted some or their entire impression inventory to another ad marketplace
4. Google “decided” to stop giving the advertiser as many impressions for non-competitor-bid reasons.

All four of these possibilities are driven by the same reason: CPM is a priority. Let’s pick 1, 2, and 3 apart (we’ll leave Google out of it for now).

1. Aside from the advertisement being offensive in some way (it wasn’t), why would the NYT block the advertiser? Because they may have determined the rev share for them was not as favorable as opening the space to other impressions.
2. Outbidding by a competitor seems clear and more CPC-based, but it really isn’t; it works in tandem with the non-bid reasons mentioned above. It’s not the literal bid, but the bid and the CTR working together; $1 bid with 0.3% CTR is worth about 3x more on the impression level than a $3 bid with a 0.03% CTR, hence the whole eCPM thing.
3. Why would the NYT shift inventory away from GDN? ’Cause GDN is remnant inventory and almost anything else offers the NYT better CPM; if the buys are there, they shift impressions from GDN to garner the best CPM possible.

This creates a huge issue for marketers: testing banner ads in a CPM world is foolish.

One challenge that we frequently see is our clients wanting to run banner split tests in GDN. It is constant. Sometimes you can even get some compelling metrics, but you are never executing a clean test.

You can only run a clean banner test if you are using a traditional display effort with an ad serving platform that is 50/50-ing your creative after the impression is already “bought” – and that is enabled by the fact that you are paying CPM in the first place. The fact is that even if you set your Google GDN campaign to rotate evenly, even if you set a 50/50 experiment to rotate your creative evenly on top of that, Google will not do it.

They would be stupid to do it, in fact. They would be losing money by 50/50-ing your creative when one variation is going to have a better CTR and therefore make them more CPC money. The algo will always favor that known better CTR ad, try and “guess” the CTR of the less known ad, and only let enough impressions through on the less-known ad to “learn” if it in fact might make them more money than the control ad.

If and only if the less-known ad’s CTR success is established, the experiment ad will “take over” the lion’s share of the impressions and start to balance your results to look like something 50/50-ish happened. But it really didn’t.

There are other CPM ramifications marketers should know.

So, lack of control on managed placement activity and lack of clean GDN banner ad tests are 2 things influenced by our CPM world. A few others include:

1. It causes huge fluctuations in Facebook daily performance – Facebook’s algo is showing your ad to your target only when it determines Facebook wins in the CPM game. Time patterns factor hugely into this whole thing.
2. The narrower your targets, the higher CPCs you need to get any impressions. In GDN, Facebook, etc., the smaller the pool of impressions you are targeting in the first place, the fewer auctions you are eligible to even play in, so the algo “finds” little to no impression where your CTR and CPC win the CPM game – especially if it’s a newly launched initiative with no history – unless you bid really high. Better to start with the broadest target possible and trim from there.
3. DCO and CPA bidding work well, but…  It’s like buying the remnant inventory within the remnant inventory, BUT it’s just the trash that happens to be one person’s treasure. (Yours.)
4. You should only switch out high-performing (CTR in particular) banner creative in GDN if you are seeing fatigue patterns. If you have locked into something the works, keep banking off the impressions it wins you until fatigue sets in.

This is a search thing too.

If you think this issue is limited to display-based channels only, you are wrong. Yes, the impact is often less visible in search because there are more ad slots, thus more true CPC position-based auction dynamics to play with. CPM is still a huge factor, though.

How?

1. It’s the whole reason “Quality Score” is a “thing.” It behooves Google to predict your CTR to influence position to drive better CPM for them. This is also one reason I’d argue that you should ignore the QS metric and look at the ones the Google algo is really looking at like CTR.
2. It’s why Google so aggressively keeps adding and testing ad extensions. Ad extension increase CTR and encourage top positions – and both of those things increase Google’s earned CPM.
3. Enhanced campaigns force more tablet advertising adoption. This generates a bunch more impressions in the system, therefore more impression-driven revenue for Google.
4. It lets Google offer click to call free on enhanced campaigns. Why? To incentivize mobile search advertising because the free click-to-call cost is surely nothing compared to the value of pushing so many impressions on mobile, with higher web-based CTRs, simply due to the extension being present and an incentive to be in top spots to get the extension to show. Kinda brilliant, really!

For every institution that “controls” impressions on any level – Google, Facebook, The New York Times, or even smaller AdSense publishers – winning means getting the most money out of every impression that is allowed to happen.

So whether it’s driven by very sophisticated algorithms (like at Google) or subjective mom-and-pop AdSense publishers blocking site ads that don’t pay out agreeably, seemingly-CPC advertisers are really operating in CPM worlds. It is less evident in google.com search because of the standard multiple ad slots that allow the bid to factor more prevalently, but it happens in search channels too – and in some limited ad slot situations (hello mobile!), the impression battle becomes more evident.

Susan Waldes has worked in the search engine marketing industry since 1999. She joined PPC Associates in August 2011 after serving as the in-house SEM manager for e-commerce retailer ivgStores. Susan has handled a multitude of lead generation, branding, and e-commerce clients in her previous roles at ROI Revolution and Rockett Interactive, as well as a stint as an independent SEM consultant. Susan has a BFA from Savannah College of Art and Design, and has contributed insights about SEM and client relationships to the PPC Associates blog and other highly regarded outlets, including Techipedia.com.