Earlier this month when the adCenter Blog confirmed that they were making changes to Bing's broad and phrase match behavior, I was optimistic about the prospects. In a few recent posts, I've touched on some issues that may be contributing to the relatively weak financials Microsoft and Yahoo have reported for their respective paid search efforts. One of the larger issues appeared to be adCenter's overly restrictive ad serving.
In the broad match announcement, and through other communications, advertisers were advised to be "prepared to take on additional traffic" and adjust our campaign budgets accordingly. Given that our performance for Bing-powered ads was more efficient than necessary (despite our best efforts) and more efficient than Google, this sounded like great news. In other words, we would be happy to bring in more traffic, even if the quality wasn't quite as good as the existing traffic.
Unfortunately, this doesn't appear to have been the big change I was looking for, as our results haven't shifted much, particularly not in the ways I had hoped.
No Gains for Search Alliance Traffic Share
First and foremost, we have not seen a significant increase in the Bing/Yahoo Alliance's traffic share since the adCenter ad matching change on August 5th.
Before getting to some specific data on how Bing's ad matching has changed over time, let's first consider another major contributor to traffic share trends.
Search Partner Traffic
Declining search partner traffic had been a significant factor in the Alliance's lost share, but that effect appears to be fairly stable in the last few months:
This view gives the share of paid clicks from search partners for the Alliance and Google as a ratio. If that ratio were 1, both the Alliance and Google would be receiving an equal percentage of their respective traffic from search partners. Currently, search partners make up a smaller share of the Alliance's traffic, coming in at about 80% of the rate Google receives.
Going into the Alliance we were taking steps to rein in Yahoo's poor quality partners and the impacts of our efforts were accelerated when ad serving was taken over by Bing. In September of 2010, partner traffic contributed about 30% of the Alliance's clicks. By December that had fallen to between 20-24%, where it has remained ever since.
With partner traffic appearing stable and most reports suggesting the Alliance is either gaining or maintaining its share of natural search, it would appear that the ad matching change was not significant. That is in fact what we see when we dig into those numbers.
As we've done previously, we took a look at the rates at which adCenter matched ads to queries using the three main matchtypes. Here we define exact match as when the keyword and the search query are identical, phrase match as when the keyword in its entirety can be found in the search query and broad match as everything else.
Again we are comparing the Alliance to Google and presenting their results as a ratio where 1 would represent both generating clicks from a particular matchtype at the same share of their respective traffic totals. The Alliance is currently generating non-brand clicks from broad matched ads at roughly 70% of Google's rate.
We have not seen a spike in broad matching by adCenter since their algorithm change on August 5th. For the week of 8/14/11 the Alliance broad matched at 68% of Google's rate, which is what we saw the week of 7/24/11. Way back in April, the rate was higher, although not significantly, at 71%.
The Alliance's exact matching rate has declined a bit compared to Google over time, but at 85% of Google's rate for 8/14/11, it was similar to the 86% we saw on 7/10/11. As we've pointed out before, adCenter continues to rely on phrase matching at nearly double Google's rate.
Brand Term Ad Matching
Note that the matchtype figures above are from non-brand keywords (those that do not contain our clients' brand names or variations thereof). For brand terms, we have seen a significant change in ad matching, although the timing of it doesn't fit with the 8/5 change.
I had heard some anecdotal accounts from our analysts and seen data for individual programs suggesting Bing was sharply increasing broad matching for brand keywords in a manner that wasn't necessarily advantageous to advertisers. Looking across a larger data pool suggests that these weren't isolated cases.
As a result, many have seen CPCs for brand terms spike and ROI suffer and we have had to redouble our efforts to keep traffic quality up by applying negative keywords and more restrictive ad matchtypes appropriately.
An Aversion to the Long Tail?
Since the ad matching change didn't have the impact we were looking for, is there anything advertisers can test in order to boost their Alliance traffic efficiently? It runs counter to much of what we appreciate about paid search advertising and the internet in general, but maybe the long tail is the problem (or more specifically how adCenter handles it).
Last week, the adCenter blog offered a run-down of adCenter "idiosyncrasies" that "you may not have known". One in particular jumped out at some of us here at RKG:
It’s a bad idea to add low volume, low cost keywords to your campaigns; low quality keywords risk decelerating the entire ad group.
At the risk of overanalyzing this statement, which may not hold much authority and may just be vague or poorly worded, this raises some serious concerns. It seems as though adCenter is conflating terms that are low volume and cost (i.e. the tail) with terms that are low quality. Now, it's easy to understand a term with a low click-through rate being considered low quality by a search engine, but low volume and cost alone implying low quality is a non sequitor to those of us that have seen tail terms receive strong CTRs and convert at great rates when they are eventually searched by customers.
It's also concerning that tail terms may "decelerate" the entire ad group which contains them. While it's generally a good practice to separate out top traffic terms into their own ad groups in order to tailor ad copy specifically for them, should we be forced to segregate out other terms based on their expected traffic levels, even if they are relevant to each other and their shared ad copy?
This apparent aversion to the tail manifests itself clearly in our data. Here is a view of head vs. tail for Google and the Alliance, where a head term is defined as one that generates at least 0.1% of a program's traffic on that engine.
That definition puts Google's tail share at a neat 50%, while only a third of adCenter traffic is generated by the tail. We do our best to run Bing paid search with as much care and attention to detail as we give our Google campaigns, so these stark discrepancies in performance should reflect an inherent difference between how the engines handle what we provide them. User demographic differences may contribute as well, but they are likely minor in comparison.
In the end, maybe we need to challenge our assumptions. Our figures on the Alliance's troubles in paid search have generally been in the context of Google, which is clearly doing quite well. Our expectations for Bing and Yahoo paid search traffic levels are largely predicated on their combined share of total search traffic as reported by the comScores and Hitwises of the world and those figures could simply be wrong.
As I've said before, we wish the Alliance the best in monetizing the traffic they do have as it should benefit both them and advertisers, while hopefully helping consumers find what they're looking for along the way. In the meantime, advertisers may want to test their strategy around the tail, even if it defies their views of best practices.