We started migrating clients to Yahoo's new search management platform (still often called “Panama”) in early December. We’re now about two months in and were curious to see what impact the new platform was having in aggregate on our clients and their campaigns. Determining that impact isn’t straightforward, particular going through the holiday surge.
To come up with holiday-invariant measures, we indexed Yahoo monthly results against Google monthly results and looked for trends. Specifically, we pulled data since January 2006 for all of our paid search clients. (We excluded clients running Yahoo-only initiatives, or in the startup process.) We opted to only focus on Google and Yahoo, excluding smaller engines. For Google and Yahoo, we pulled statistics from our data warehouse across our client base: total clicks, cost, sales, impressions, and orders. We then computed derived statistics like advertising-to-sales ratio (A/S), cost-per-order (CPO), conversion rate, click-through-rate (CTR), cost per click (CPC), etc. For each month, we divided the Yahoo measure by the Google measure. This indexed Google to 100%, and cast each Yahoo stat relative to Google.
An essential aspect of this analysis is the fact that neither we nor our clients come to the table with the expectation of spending a certain amount of ad budget on Google and a certain amount of budget on Yahoo. We almost never set budgets by engine. Rather, the overwhelming majority of our clients instruct us to buy as much good advertising as we can. As long as that advertising meets the client’s success metric -- be that efficiency, cost per order, ROAS, ROI, whatever -- we will buy as many good clicks as we can. In short, the ratio of our Yahoo advertising efforts relative to our Google advertising efforts is a function of performance, not by an external constraint to spend a certain amount of money on one engine or the other.
This indexation allowed us to look at Yahoo results relative to Google results by month across our entire client base for 2006 and the early part of 2007. Our interest was focused on December 2006 and January 2007, as Yahoo rolled out the Panama platform.
The first graph, the single pink line, tells a dismal story of Yahoo performance relative to Google for most of 2006. In January of 2006, for every dollar our clients spent in aggregate on Google, they spent about 38 cents on Yahoo. Over the course of 2006, the graph shows Yahoo losing share in search advertising spend to Google. The graph reached a low in November 2006, where relative to each $1.00 our clients spent on Google, they spent $0.22 on Yahoo.
In December as our larger clients migrated to Panama, this share statistics number upticked slightly. And in January, as more of our clients migrated, the share statistic upticked again. While two months of improvement don't constitute a long-range trend, and while the modest improvements do nothing to staunch the tremendous loss of share Yahoo suffered from Google over 2006, it is interesting to see that the arrival of Panama coincided with two upward motions in the ad spend share in Yahoo's favor.
The second graph expands the picture to present not only total ad spend share but also clicks and cost per click. The pink line on the graph is the same pink line from the prior graph. (You can see that the first graph exaggerates the trend by compressing the y-axis --apologies to Edward Tufte.)
The second graph shows that the collapse in Yahoo's total ad spend share across 2006 was due to both falling click volume (relative to Google click volume) and falling CPCs (relative to Google CPCs). The second graph also suggests that what caused Yahoo’s ad spend share to uptick relative to Google in December and January was improved CPCs.
That is, our clients increased their Yahoo share of ad spend in December and January, not by buying more clicks, but by paying more for those clicks. Our bid management platform increased CPCs (relative to Google) and moved ads higher on the page because those Yahoo clicks performed better. Note that by December and January, the blue line had climbed back to 100%. As our bid management system is driven by customer profitability and efficiency goals, this CPC parity indicates quality parity between Google and Yahoo for these months. This might suggest migration-Panama was more effective at monetizing clicks than the ancient DTC-XML Overture platform.
Again, these effects are small. And they’re only two months of data. Nonetheless, these upticks over the last 60 days may indicate that Panama is helping Yahoo provide higher quality clicks to advertisers, commanding higher CPCs for those clicks, and helping Yahoo increase their share of total ad spend (a small amount) relative to Google.
Tomorrow we'll continue this analysis and look at the situation from the advertiser perspective. Across our client base, we’ll see that Yahoo provided higher converting clicks during the last two months (by a little bit). This in turn allowed our bid management algorithms to buy ads higher on the page, pay (a little bit) more per click, increase traffic to our client’s sites (a little bit), while simultaneously improving our clients’ Yahoo efficiency metrics (a little bit).
In a comment below, Jared correctly points out that Panama quality-based bidding didn't roll out until Feb 5th. So the effects here can't be attributed to Yahoo's new ad ranking method. I just checked in with our senior engineers on this -- they said they had asked Yahoo in November how ads would be ranked during these transition months, when some advertisers were on the Panama bid platform and other competing advertisers were on the DTC-XML platform. Our engineers said Yahoo's response was "unclear" and "ambiguous". So, while this uptick can't be attributed to Y quality score, it could relate to how Yahoo divided up traffic between Big Advertisers (early Panama migration retailers, few accounts, big spends) vs. Small Advertisers (later migrators, many accounts, smaller spends).
It will be interesting to watch these share graphs after February, once quality score kicks in.