Michael Katz wrote an opinion piece for Adotas critiquing an auction marketplace for advertising and recommending an open CPA marketplace as a better solution. Everyone is entitled to their opinion of course, but as someone heavily involved in an auction ad marketplace with Yield Manager, I’d like to discuss a few of his comments.
The auction-based model is an incomplete solution though; a truly efficient system exists only when there is an open marketplace based on pay for performance.
First, I don’t think he ever really proves in the entire article why an open network based on performance is truly more efficient.  Second, I don’t think it’s really true. Pay for performance has some great advantages, but it also isn’t always the best solution for publishers. They’re the ones taking all the risk in a 100% pay for performance scenario.
The only people enjoying the efficiency of this program are the owners of the network, though. The advertiser has spent more than they needed to on traffic that did nothing but increase their overall spend and probably did not spend enough in the places where they should have to increase the delivery of highly converting traffic.
How does an advertiser find out what sites will convert without testing all the sites in the network? I don’t see how it’s possible to ONLY spend money where the traffic converts as Mr. Katz suggests. Once the system does find where traffic is converting, it spends as much as possible there.
The publisher, similarly, potentially missed out on a higher yielding ad.
No, the way an auction works is the publisher gets the ad that will pay the most for that impression. That is the highest yielding ad.
By utilizing an open network, advertisers see exactly which publishers comprise the network, view performance by site placement and can adjust bids on a site specific basis to ensure that no impressions are wasted.
I am a fan of open network concepts, don’t get me wrong. However, I also know that the majority of advertisers don’t have the time or knowledge to do this themselves as effectively as an automated system. Advertisers are people, which means they get busy, or they have biases that may cause them to do funny things when manually adjusting bids.
As an advertiser, if you could see that a handful of sites are providing most of the return on your investment but you are not getting all of their traffic, wouldn’t you want to re-allocate the dollars you are spending on low performing sites and pay a slightly higher rate to ensure you are maximizing the high performing placements?
Paying more to get all of a publisher’s traffic isn’t necessarily a good idea. It could work well, but increasing the frequency of your ads on their site might not lead to the same ROI you’re currently getting. As you spend more and more, you may get less ROI out of each impression. Regarding spending money on low-performing sites, because of the way Yield Manager works sites that don’t perform don’t earn much because they aren’t performing. This means advertisers are not spending a lot of money to get volume on these sites, if they end up still serving to those sites at all.
Finally, the network diversifies its risk by running a higher number of more profitable campaigns instead of relying on a handful of advertisers to make up 90% of their revenue.
I don’t see the direct relationship between allowing advertisers to manually adjust their bids per site to a higher number of more profitable campaigns showing up. I think that’s an assumption that advertisers would just come out of the woodwork if the model was changed.
The problem is that at the end of the day, a CPM deal offers almost as much risk without the upside of a CPA deal though. If the traffic is not converting, the advertiser will cancel and most of the time its with 24 or 48 hours. If you want to retain them as an advertiser, you will have to provide some sort of make good, lower the rate, or possibly even switch to a CPA model. In any event, the advertiser now has the negotiating leverage.
The dynamic CPM pricing type advertisers can use in Yield Manager solves this problem well. The upside for the publisher still exists, if their traffic converts well advertisers will automatically bid higher to get access to their inventory. It also mitigates their risk, because they aren’t giving away all their volume to CPA campaigns hoping that they’ll convert at some point. Mr. Katz says that in a CPM or CPC model the advertiser ends up with the leverage. However, I”d argue that in a CPA model, they also have the leverage as they have no risk.
In summary, by taking on a higher quantity of advertisers and matching the best performing ads with the best performing sites in an open market on a pay for performance basis, the network achieves three results. First, publishers make more because the advertiser allocates more money to the quality sites to ensure more traffic.
A good auction system does that automatically. Advertiser money IS allocated more towards quality sites.
Secondly, the advertiser maximizes the utilization and efficiency of its spend because it is paying more to high performing publishers who will now allocate more impressions on a macro level to the network and on a micro level to that advertiser.
Again, this already happens. In a good auction marketplace, the advertiser does pay more to higher performing publishers.
Lastly, the network diversifies its risk and builds relationships with advertisers and publisher by showing it is committed to providing the best and most efficient results possible.
This could be accomplished just as easily by just saying a network should take on more advertisers. An auction marketplace is committed to providing efficiency and the best results possible.
I like open markets, I’d like to see more auction marketplaces go that direction. I’m just not sure pay for performance only is the solution, or that auction markets need to be bashed to promote the idea of an open market. And why can’t an auction marketplace be open as well?
First off, this article was not meant to attack anyone who uses YM but clearly you do not understand the meaning of efficiency or market place. I will address your question before I obliterate your argument.
Q: How does an advertiser find out what sites will convert without testing all the sites in the network?
A: I never said I was against a testing a network. Test periods are imperative for optimization but they should be limited, often times you can tell what is going to work very early on. Furthermore, (if you have the means to afford it) you can give yourself a better chance through using comScore data and other demographic targeting approaches. The idea is that you spend money efficiently by testing systematically and then once you find your sweet spot, running only those sites. If its a blind network, you do not know what sites are working. In an open network, you see exactly which sites are working AFTER the test period.
The definition of efficient:
1. Acting or producing effectively with a minimum of waste, expense, or unnecessary effort.
2. Exhibiting a high ratio of output to input.
Here is how a blind CPM buy can equate to waste:
Say (for simplicity purposes) a network has 3 sites on it. The advertiser is paying 1.00 CPM and has a desired CPA target of 15.00.
Site 1 generates 10 conversions on 100,000 impressions
Site 2 generates 0 conversions on 200,000 impressions
Site 3 generates 15 conversions on 100,000 impressions.
The advertiser has spent $400.00, their eCPA is about $16.00 which would be tolerable to the advertiser but if they saw which sites were performing they could have spent either $375.00 on a CPA buy (or could have eliminated all of site 2 traffic early on and spent closer to $200.00.) The advertiser is satisfied with the results they got and the network is happy because they know they eeked out an extra $200.00 here, but what if the advertiser knew they could have gotten the same results or better by spending half the money. Do you think they would be as satisfied? Note that the advertiser could have seen even better results because site 3 could have been making $1.50 instead and could have possibly given the ad better placement within its rotation and in turn generated even more conversions. You may argue, isnt this the same as using a dynamic CPM? The answer is no b/c the advertiser would still lose out on Site 2 albeit at a reduced rate.
An open marketplace where the advertiser pays only for performance is more efficient because 1. an advertiser is only paying when their ads produce the desired effect and they are minimizing the waste by not paying a CPM rate, 2. the publisher, if their traffic matches well with the advertiser’s offer, will make more money in terms of an eCPM and give better placement to that advertiser.
Hi Michael,
Thanks for engaging in a discussion.
First, your example is somewhat correct on a blind eCPM buy, but that isn’t how dynamic pricing works in our situation.
Site 2 wouldn’t get that disproportionate amount of volume if it weren’t generating conversions, and the eCPM they’d be getting would drop signficantly or they’d be entirelycut off to eliminate that waste. As you said, SOME testing is required, so in this case that’d have to be acceptable to determine if the campaign worked for that publisher.
Sites 1 and 3 would see a raise in eCPM and more of the advertiser’s budget heading in their direction since they generated conversions at a solid rate.
I agree with your basic premise that pay for performance (CPA) eliminates advertiser waste, but it CREATES publisher waste. You have publishers running lots of impressions that aren’t monetized at all, unless they end up with a solid conversion rate.
So what happens in your example above if it were CPA only, is that Site #2 quits your marketplace after a couple of days and never comes back before you’re even given a chance.
There are 100% CPA networks out there doing fine, but there’s a large group of publishers who have tried and failed or will never work with those networks because of the waste they get when they aren’t paid for impressions.
Reality is that publishers don’t always want to take all the risk, and it’s hard to minimize waste on both sides of the equation. We feel that our system does so with dynamic eCPM that tests and adjust quickly and automatically, combining the strengths of CPA and CPM ad buys.
Publisher waste is not a bad thing as far as I’m concerned, it is a self correcting situation. Publisher waste occurs when their traffic is of poor quality, there is no publisher waste when the traffic converts at a high rate.
The fact is that there are several ways publishers can obtain visitors to their sites and some produce less than desirable results for the advertiser. Take Hula Direct as an example. They were buying traffic to their [ridiculous] sites by having networks and adware company serve their URL as a full landing page then monetizing by putting up to 10 different ads on their site. Sites with good, quality traffic who obtain it through mastering skills like SEO and link swapping with similiar sites should be rewarded. Webmasters who take shortcuts should not be rewarded.
Well, maybe I’m publisher-focused being a publisher who works with publishers, but I don’t see publisher waste as “not a bad thing”. It’s not necessarily self-correcting, often times publishers have to do the self-correcting by switching to a new ad network or provider because the current one just isn’t monetizing their inventory for whatever reason.
A high quality publisher can just not meet up well with a particular network’s CPA offers, resulting in a poor eCPM. The only self-corerction is they leave at that point.
I agree with your example of not rewarding bad publishers, which is why dynamic eCPM bids very low or not at all for inventory that is not providing value.