Apr 92012

Do Paid Ads Cannibalize Organic Traffic?

Google has released some new research on the question of paid and organic search cannibalization.

This study is more carefully done than their earlier pass which concluded that 89% of paid search ad clicks are incremental. The challenge with that older study was three-fold:

  1. They failed to recognize that (according to their new study) on average 81% of the ad impressions happen when the advertiser has no organic link present,
  2. They made no distinction between brand ad and non-brand ads, and
  3. They did not consider the relative position of the paid ad and the organic link.

This new study takes a deeper look, resolving the first and third problem, but still not explicitly addressing the hugely important brand/non-brand distinction.

Their findings are well worth studying as they have access to more data than any of us can dream about.

The box plot below indicates that while 81% of ad impressions are served in the absence of a competing organic link, the average advertiser receives ~75% of their ad clicks without an organic link on the first page. About 12% of their ad clicks come with an associated organic link in position 1, 4-5% with an organic link in position 2 - 4, and another 8-9% with an organic link on the first page but 5 or below. {In a box plot the upper limit of the box represents the 75th percentile, the lower limit of the box the 25th percentile and the line in the box is the 50th percentile or median.}

The next box plot is fascinating. It indicates that there is some correlation between ad position and organic link position. Advertisers tend to be higher on the page for ads when their organic link is higher, lower when the organic rank is lower or non-existent. This may be evidence that advertisers and Google reach the same conclusions (in general) about what traffic is best suited to their website. Advertisers bid more when users are happy with what they find and are therefore worth a premium. Google seems to do a pretty good job reaching similar conclusions, at least when an organic link reaches the first page.

Google did not separate brand searches from non-brand to my consternation. Obviously the strong likelihood of an ad being in the first position when the organic link is also first is heavily impacted by brand traffic.

With respect to cannibalization the findings were quite interesting and not surprising. The higher on the page the organic link the more likely it is the ad cannibalizes some traffic, but even with a position 1 organic link 50% of the ad traffic is incremental.

Now, it kills me that we can't separate brand from non-brand here. We know there is a navigational component to brand ad traffic that isn't incremental, and separating non-brand position 1 from brand position 1 is crucial.

While they're at it, they should really take apart the brand ad component by the type of brand. There are at least three types of brand names with all kinds of grey in between.

  1. Brands whose products and services are available widely. Someone searching for Sony TV could by from Sony who might have the top organic listing and top paid ad. But every other ad on the page will be a reseller of Sony products -- very attractive to a consumer. Same idea with "jetBlue" -- to a degree, they compete with their own distribution chain.
  2. Brands whose names are similar to the category of product or service. "Free credit report," "drugstore," "envelopes". It's less clear whether these searches reflect users who are open to "shopping" anywhere, or whether they're looking for freecreditreport.com, drugstore.com, or envelopes.com specifically. Undoubtedly there is a mixture of both groups which would impact the degree of incremental traffic.
  3. Brands whose names are unique, and they do not compete with their own distributors. "Talbot's", "Walmart", "Walgreens", etc: it's pretty clear that these users are trying to navigate to the site.

Common sense suggests that a significantly larger share of traffic is incremental for brands that fall into the category of #1 and #2 than is likely to be the case with brands in category #3. Our own initial research into this seems to confirm this intuition.

We’d like to see a follow up study in which Google disentangles these three types of brand searches, in addition to separating all brand searches from instances in which an advertiser’s organic link ranks first for a non-brand search phrase.

Last, there may be some “helpful” cannibalism involved as well as the hurtful cannibalism of paying for traffic you would have gotten for free. Paid ads siphon some traffic away from first page organic listings, but particularly on brand searches, they also siphon traffic away from organic affiliate links. Pushing the “10% off Acme at couponaffiliate.com” below the fold or off of the page may save a brand money when commissions and discounts are factored in. It’s not easy to tease this out, but our attribution tracking gives us a pretty good look at this phenomena.

Some might question (indeed comments did question) whether Google’s data can be trusted. Is Google just telling us that the vast majority of ad clicks are incremental to hide the truth about the extent of cannibalism?

I don’t think so for a number of reasons:

  1. It’s hard to fake data in a carefully done study. There are plenty of poorly done studies out there that produce meaningless ‘findings,’ but when the methodology is solid – which this is, brand non-brand nuances notwithstanding – it’s difficult to make up numbers. It would also be legally dangerous for them to publish a study that has been intentionally cooked. There have been times when they haven’t published research findings that are hurtful to their interests, but it’s hard to fault them for that.
  2. These findings don’t help their cause much. For years and years agencies and others have claimed that ads and organic listings actually reinforce each other. That is to say: having an ad present increases traffic on the organic listing and vice versa! This study TOTALLY refutes that old saw. Google acknowledges that half of ad traffic is cannibalized from position 1 organic listings, and 4 – 10% of ad traffic is cannibalized from organic listings further down the page.
  3. These findings are consistent with what we’ve seen. Your mileage will vary and testing is wise, but these results make logic sense and jibe with what we see.

Google also suggests conducting tests and helpfully provides a nice methodology for testing using geo-targeting. The proof is in the data, and you should test yourself, but these benchmarks may be useful for refining efficiency targets.


11 Responses to "Do Paid Ads Cannibalize Organic Traffic?"
Jim Novo says:
At last, we can finally put the "ads and organic listings always reinforce each other" idea to bed and accept that cannibalism exists. Bravo to Google for going ahead and publishing this, since for all practical purposes they are the only ones who can really examine these effects in depth (though some of us have done pretty convincing studies). Probably won't hurt their business much, since the "but we still get Brand exposure" crowd is quite large. In the meantime, 1 big myth down, probably 20 or so to go, eh George?
Jim, thanks for your always insightful comments. There is a whole crowd that won't believe any study Google puts forward even when the findings run counter to their own interests. No one wanted to believe Hal Varian's research that conversion rates were independent of position, despite the fact that it would be in G's interest to say position 1 converts better than position 10. Given that, this will likely be ignored as well, but the folks who get it will benefit from paying heed to this one. It's been a while since I riffed on Search Myths. Maybe it's time to do so again?
Nag Patta says:
Hi George - I agree that testing is the way to do. We work with a lot of customers who ask the very same questions and do extensive A/B testing to determine how they interact with each other. They see varied results depending on the type of keyword and target demographic, among others. You can see one possible outcome in this article - http://techcrunch.com/2012/04/02/brightedge-hebs-digital-case-study/. Happy to share the results of more such testing!
Great analysis as always George. I'm glad that they've backed down from the first claim, that 89% of clicks were incremental. Ultimately, I think they just shouldn't bother with these studies, because the only result that will be accepted as legitimate is one which suggests that advertisers should spend less on paid search as a result of the strength of organic results. Otherwise, it just looks like Google is creating reports that justify its bottom line. Sort of reminds me of those Kent cigarette commercials from the 40s: "7 out of 10 doctors smoke Kent!"
Nag, David, thank you both for your comments. I'm doubtful that enough folks will care to manage programs with this degree of precision. We wished more managers cared about really getting it RIGHT. Too often their bonuses are tied to "growth" that create disincentives to trimming the sails (sails...sales....get it? :-))
Nice analysis, George, and great 2nd chart. DRod - who knew you were alive in the 40's!
Thanks Chris, you're right DRod is remarkably well-preserved for a man his age!
Don't hate me for my breadth of historical knowledge!


Check out what others are saying...
[...] Do Paid Ads Cannibalize Organic Traffic?, Rimm Kaufman [...]
[...] In part perhaps due to placement and a lack of user knowledge between ads vs content, paid ads do significantly cannibalize organic traffic. [...]
[...] What is the overall traffic and revenue impact upon a group of targeted keywords ranking organically in the top 4-10 position when we kick off PPC bidding on those terms that weren’t previously bid upon?  Is there a correlation between ad position and organic ranking? [...]

Leave A Comment