Commentary
The Efficient Frontier of Display ROI: UGC, Porn, & Torrent Sites?
- by Paul Knegten , December 1, 2009
I'm liking a lot of the comparisons I'm hearing of media buying and selling to stock trading on Wall Street. Whether you choose to call them pork bellies or diamonds, impressions are commodities that carry with them the same properties of their counterparts on the trading floor.
Like a good MBA student, I took the requisite finance classes and studied modern portfolio theory. The thesis is that no single stock purchase is efficient: it exposes you to undue risk relative to its expected return. The more diverse stocks you buy, the more your risk/return ratio gets smaller, and the more you approach the "efficient frontier."
I like the sound of that (looks like somebody else did too and founded an SEM agency with that name). I'm certainly not the first to claim that the parallel exists between exposing yourself to undue risk with a single stock and exposing yourself to the ROI risk of running an ad on a single website. But let's focus on that efficient frontier now: on Wall Street its index funds and ETFs. What's the equivalent here?
Ad exchanges (like AdBrite, RightMedia, DoubleClick, etc.) are the soft answer. But what does this mean? What stocks am I actually buying here, and which ones are making me rich?
This is where we depart from the Wall Street analogy a bit. I claim that the efficient frontier of Display is the stuff we never thought we wanted, and the stuff I always get asked to exclude by clients that ask for that kind of thing: user generated content (UGC), porn sites, and torrent search sites.
Here's why: this stuff is dirt cheap, as would be expected. But what's not expected is that these sites are heavily trafficked, and are now becoming the wild west of some really, really cool retargeting. Suddenly a $0.0001 impression on a UGC site becomes immensely valuable when you know the user looking at that impression did a search for laminate wood floors yesterday. Here's an experiment you can do: do a few searches on travel sites, on retail sites, and on the home improvement example I just gave you. Then, go to a site like imageshack.us and refresh the page over and over for a while. It's the last place you'd expect to see the next generation, science fiction ad campaigns but here they are (among some really lame ads as well, bear with me).
What's happening here is that by giving the campaign access to the entire market (akin to an index fund or ETF), the market is determining where the best place is to show your ad, in this case based on wherever your customers are browsing. So, in that sense, this is indeed modern ad portfolio theory in action. But, I would reject the idea that the whole web is the efficient frontier; I would modify that theory to apply to online advertising and say that an impression on the New York Times is actually far less valuable (on a strict ROI metric) than the über-cheap long tail sites. The marginal cost of an impression on a premium site will never allow the ROI possible with the marginal cost of an impression on a long tail site.
So what about targeting? Back to MBA school: segmentation, targeting, positioning! The idea that seemingly generic sites that provide no targeting information of their own are the most efficient sites to target your customers is more than a bit counterintuitive. Fair enough, but we're imposing our age-old conception of content targeting (from print) upon a web that contains far more user data than print media. User behavior and algorithmic media bidding are quickly changing the rules.
Excellent.
Well-said. If you connect these ideas with post-click marketing, you're on the road to computational marketing as analogous to computational finance. (The commentary the other day on "When Algorithms Collide" is relevant to this discussion too.)
Paul,
I understand the analogies that have been made regarding media buying and stock trading. In fact, in the context of all of the emerging tools in the display space (DSPs, exchanges, analytics), I agree with you that there are several parallels that can be made.
However, I must disagree with the statement that user generated content (UGC), porn (adult) sites, and torrent sites should be included on any type of media plan (re-targeting or otherwise). Despite the low CPMs, the content on adult sites is a completely inappropriate environment for any legitimate advertiser (DR- or brand-focused). Torrent sites are widely known to facilitate the practice of illegal file sharing - ask any marketer’s legal counsel about their view of sponsoring that content. And finally, two clicks in on imageshack.us provided some content that was at best extremely NSFW. The CPMs on these sites are low for a reason.
However, there is one thing you were dead right on: I was amazed at the ads I was seeing after a few refreshes on the page. One of which was for a 200 year old, high-end men’s clothing retailer whose flagship store is right up the block from Undertone headquarters on Madison Avenue. I’d be shocked if the brand manager for this conservative company was happy with this placement.
The good thing here, from my perspective, is that it seems that many advertisers know what’s going on, and are specifying that these types of sites be excluded from the plan from the start, as you mention. However it seems that some, as evidenced by the results of my imageshack.us experiment, are not.
It’s simply incorrect that advertisers should have to give up control and environment to drive results. There are companies (like Undertone) that are focused on delivering not only results, but in the highest quality environments. In the spirit of your finance analogies, call them the Berkshire Hathaways of online display: focused on only the best companies (sites) and, more often than not, beating index funds by a wide margin.
Eric Franchi
Interesting thoughts but I'd say it's quite expected that UGC, Porn and Torrent sites are heavily trafficked and are the Wild West of the Web. Interestingly if these sites are so valuable then a market system will drive the cost of their media up, no?
The curveball is that the cost of the impressions across these sites is not really what drives ROI -it's CTR. This is why much of this type media is CPC or is backed into CPC. With so much inventory even (and especially) on top tier publishers, I'm not sure how long this super low CTR frontier remains more efficient.
Great post Paul. I have to agree with Eric that porn, torrents and objectionable content are not good places to run mainstream advertisers even when the price is right. You are certainly correct that there are huge inefficiencies in the market where good advertising opportunities are being incorrectly priced. But there is so much volume available in non-objectionable places that there is no need to risk an advertisers brand by running on objectionable content.
Yeah, I agree with you guys that brand sensitivities need to be taken into account. If I were on the other side of the table I'd request my brand be clear from porn sites too.
And there's where I'm admittedly a hypocrite :) If I see a brand advertising on a porn site, no sooner should I an utter a "how DARE you, favorite brand of mine!" than I should ask myself "well what are YOU doing actually browsing the porn site?" I'll leave you with a quote a podcast guest of mine (that shall remain nameless) said "Sure I look at porn. But I also stay at the Four Seasons..."
Theoretically this is all true, value, particularly for a DR campaign, can be understood from targeting on these kinds of sites, but ultimately someone has to answer that call from a client asking why their executive saw an ad for their product on that site (other than asking why the executive was on that site). It's going to depend on the brand, and transparency of the exchange/network/publisher. This is certainly an important issue to consider these days rather than the knee-jerk reaction we've seen in the past.
placing adverts on a variety of sites may work differently for different brands. it will be cheaper and might even have a better CTR, however will it have a better conversion rate. at the end of the day, the business decisions arent made because of CTR, they are made because of which site gives you the most conversions.
if there is a direct link between lower CPM = more impressions = more clickthroughs = more acquisitions, then the logic is sound. if the maths doesnt work out, great - if it doesnt, you may need some blue chips in your portfolio.
@Nevo - Oh definitely. No mention of CTR in this editorial and rightfully so.