Conflicting data and opinions abound; new threats to the hegemony of search, real or just perceived, continue to spring up; and, there are times when all of us may wonder just how much the numbers are actually changing and whether we should be concerned.

Case in point: comScore recently showed U.S. core searches increasing 11% year-over-year to hit an all-time monthly high in March;  however, as recently as December, it was showing declines in the same metric.

Even as its count of Google searches has trended up in Q1, comScore shows total visits to Google sites declining 0.3% Y/Y.  Meanwhile, Google released its own numbers for Q1 a couple of weeks ago showing paid clicks were up 20% Y/Y, or about double comScore’s growth rate for search. What gives?

While the cannibalization of organic search by paid is a real possibility, if not likelihood, the biggest factor in these discrepancies appears to be the growth of tablet and smartphone search. As the share of mobile increases, reports that only cover desktop and laptop searches, like comScore’s core search figures, will provide a less and less representative view of the full picture.

How Big is the Impact of Mobile on Total Search Growth?

In March 2012, RKG found that tablets and smartphones generated approximately 13% of paid search clicks. With higher click-through rates, mobile accounted for close to 8% of search ad impressions. Fast forward to March of 2013, and we saw mobile click share double while mobile impression share reached 20%.

If we take comScore’s March figures for desktop and laptop searches and add in those mobile impressions, we get a total Y/Y growth rate for searches of close to 28%.

That figure sounds a bit high, but it provides a sense of just how much mobile growth is driving total search growth. At RKG, we are seeing desktop and laptop paid clicks running roughly flat against last year, but mobile click growth drove total clicks to a 15% gain in Q1:

While advertisers face additional challenges with generating revenues on smartphones and assessing their full impact in a potentially cross-device, online-to-off buying cycle, the shift to mobile overall is not necessarily as fundamental as it appears. Tablets, which generate roughly half of mobile clicks or more, are largely supplementing or replacing laptops, and user behavior on tablets looks a lot more like that for PCs than it does for smartphones.

But Paid Is Just Stealing From Organic, Right?

So, we are finding our paid click growth rates trending closer to Google’s numbers than comScore’s over the past few quarters, but if paid is largely stealing from organic, that growth is somewhat illusory and likely unsustainable. Google can’t fill the entire SERP with ads, can they?  (Maybe don’t answer that one….)

A quick glance at the data in our analytics package might seem to confirm our suspicions of cannibalism, but we need to dig a little deeper. With the release of iOS 6 in September of last year, a growing amount of Google organic search traffic is being classified as direct due to Safari toolbar searches defaulting to secure search.

Without taking that factor into account, organic visits for many sites do appear to have declined or at least not grown nearly as much as paid clicks. In Q1, RKG estimated that 13% of total Google organic searches appeared to be direct site visits because Safari did not pass a referrer on an iOS 6 device.

In March of this year, this issue had the effect of reducing an estimated 17% lift in Google organic searches for RKG clients to just a 4% lift, as recorded by analytics packages.

Admittedly, these are rough calculations that only superficially address the question of cannibalization, but they suggest there is significantly more growth in the organic sphere than it may appear at first blush.

Other Threats On The Horizon

Even if we are convinced that search growth remains strong on both the paid and organic sides, there are certainly threats out there that hold the potential to disrupt the search industry. But, the threats that seem to get the most attention aren’t all that new: shoppers will go directly to Amazon, eBay, etc; consumers will eschew search engines in favor of social recommendations or searches on Facebook; everyone will use apps instead of search.

Each of these notions easily warrants a separate in-depth analysis, but I will just note a few quick data points, in case they are causing you to lose sleep at night.  With respect to Amazon, there seems to be the impression that they are devouring the entire retail industry at an incredible rate — but their net sales growth in Q4 2012 was 22%, which was the same growth rate Google reported for its gross revenue, excluding Motorola.

Much of Amazon’s growth is driven by third-party sales, which were up 40% Y/Y in Q4, and Amazon has long run search ads for third-parties on its site.  Should consumers make a sudden shift to Amazon, search advertisers will likely still be able to reach them one way or another.

On the social side, RKG data suggests Facebook and others are not generating site visits at rates not anywhere near search, even for those that have a strong presence in the space. We also found that commercial Web search volume on Facebook appeared to be very small, given the number of clicks on the Bing search ads running alongside those searches. Again, should user behavior change on that front, search advertisers will be well-positioned to generate traffic — it just might be from Facebook itself rather than Bing or Google.

“If apps were going to destroy the search industry as we know it, it would have happened by now” isn’t the most compelling argument, but the huge growth of mobile searches in the face of an ever-expanding app ecosystem suggests the overall threat isn’t dire. Certain industries could see a greater impact than others; however, travel, in particular, comes to mind.  For other verticals, apps are more likely to supplant direct site visits than competitive searches.

The Near Future Of Search

In a few months, AdWords advertisers which have yet to do so will be forced to make the transition to Enhanced Campaigns. The new model received a mixed response at best, but Google has already shown a willingness to tweak the details and make improvements as the transition date approaches.

Advertisers are still set to lose control over the segmentation of computer and tablet traffic, and there are concerns that the changes will drive up mobile CPCs considerably; but, sophisticated advertisers will handle this well, and the early impact will likely be smaller than most think.

That said, not all tablets convert alike; and, if poorer performing models continue to devalue the entire tablet segment (and, by association in AdWords, the PC segment), we may have some issues down the road. On the positive side, the new options around modifying bids by geography in Enhanced Campaigns should be a boon for advertisers and, ultimately, Google.

Finally, it doesn’t get as much attention as its much larger rival, but Bing’s performance in paid search has been trending very strongly, and that seems likely to continue in the near future.  A number of search agencies, including RKG, have seen Bing’s paid-click growth outshine Google’s by a healthy margin over the last two quarters or so. Official numbers are hard to come by, but given the behind-the-scenes changes being made there, this phenomenon looks to be very real.

Bing stands to get another shot in the arm later this year with its version of product ads, a format that has been a major traffic driver for retailers on Google.

At least in one arena, we can try to put some specific numbers together to shed some light on the potential here.

A Brief History Of Bing & Facebook

First, some background: Since 2008, Microsoft has been providing Web search results for Facebook along with search ads to monetize them.

That seems to be news to a lot of people, going by many of the articles written about Graph Search, but this deal goes all the way back to the days when Microsoft’s core search product was still known as Live Search, and its ad platform was adCenter.

Over the ensuing years, Live Search became Bing, adCenter was renamed Bing Ads, and Microsoft and Facebook took some early steps toward integrating social content into their Web results.

In 2010, Bing and Facebook partnered to bring People Search and Liked by your Facebook Friends content into both Bing.com and Facebook Web search results. But prior to Graph Search, the SERP for most Web searches conducted on Facebook basically looked like a standard Bing.com result pasted into a Facebook frame:

Bing Web Search Results on Facebook Prior to Graph Search

If anything, searches on Bing.com have been more likely to deliver social content via Facebook than Web searches on Facebook itself.

With the introduction of the Social Sidebar last year, Bing integrated feed content from Facebook (for those users willing to make the connection) along with a few other capabilities like the Ask Friends feature.

Current Bing.com Result with Social Sidebar

Even if Facebook hadn’t been investing a great deal of resources to the Web search on its site, it’s had a nearly unrivaled user base for some time now, and some of those users must be using it, right?

Facebook As A Bing Search Ad Partner

In an interview just a few months ago, CEO Mark Zuckerberg noted that Facebook was already handling “basically” a billion queries a day. For comparison, Google recently revealed it answered 100 billion queries a month, so there’s not as much space between the two as we might expect. However, based upon RKG research into Facebook’s traffic levels as a Bing search partner, it appears that Web searches make up just a tiny fraction of Facebook queries.

Analyzing publisher reports for Bing search campaigns targeted to the U.S., we find Facebook currently generating just 0.13% of clicks on Bing Ads, which also powers Yahoo.

Given that Bing Ads represent just 17% of total paid clicks for our clients, Facebook’s share of overall paid search clicks is roughly 0.02%. For another comparison, our figures suggest that Amazon, Ask and Aol all have search ad click shares in the single digits.  Facebook is about on par with ad traffic we see from Blekko.

Will Graph Search Drive More Bing Searches?

The figures above may sound surprising (and there’s a chance Bing’s own data understates Facebook traffic), but Web search has been far from prominent, if not flat out difficult to find on Facebook. With the changes rolling out now with Graph Search, we could very well see a significant increase in Web searches and search ad clicks originating from Facebook.

Here’s an example of a query using Facebook’s old search bar:

Facebook’s old search results

This is a query with decent commercial intent, but the user is presented with ten Facebook pages to Like (some of which appear just to be squatting on a potentially popular term) before the user sees the option to search the Web.

For the same query, Graph Search auto-populates just five Facebook pages before giving an option to conduct a Web search:

In addition to being more prominent, the link to the Bing search is more seamlessly integrated into the other results. For specific searches and misspellings, often all of the Graph Search results point to a Web search link.

When a user does conduct a Web search through Graph Search, the new SERP does a better job of incorporating Facebook content such as Like and Share counts, as well as related Facebook pages. The layout is also more visually appealing and looks far less like an afterthought:

Web Search Results on Graph Search

What Does This Mean For Search Ad Clicks?

All together, these changes should lead to more Bing searches on Facebook and, in turn, more search ad clicks. It is still very early, but we do see indications that this is occurring.

Facebook’s share of Bing Ads clicks still appears small, but it is now 2-3X higher than it was during the fourth quarter of 2012. The timing of that growth fits with the rollout of Graph Search, but it’s important to note that Facebook has stated that Graph Search is only rolled out “to the order of tens or hundreds of thousands of people.”  That’s a pretty small fraction of their users, so if the growth above is a direct result of Graph Search, it represents only a small amount of the potential.

Still, advertisers would have to see Facebook Web search volume grow by orders of magnitude before it will have appreciable impact on their paid search programs. While that could occur, it’s hard to imagine that Facebook sees driving up traditional search ad revenues as the primary monetization opportunity from Graph Search.

If Not Bing Search Ads, Then What?

Obviously, no one can say for sure how the monetization of Graph Search will play out, not even Facebook itself, but there have been some compelling ideas put out there. Certainly Sponsored Results as part of the search suggestions will continue to be a natural fit.

Less directly, as a number of analysts have noted, the ability of Graph Search to answer certain types of questions with commercial intent could bring a resurgence in the value of Likes. In order to ensure prominent placement in the Graph Search results, marketers of all stripes will need to make sure they have plenty of Likes for their page. One way to achieve that is by running the existing Facebook ad formats.

For his part, Zuckerberg has made a point to “temper near-term expectations” for Graph Search revenue, including on Facebook’s Q4 earnings call. At the same time though, Facebook continues to trumpet the new tool as a third pillar of its ecosystem along with the News Feed and Timeline, so they clearly have lofty ambitions in the long run.

But, revenue volume isn’t the only metric that shifts significantly during the holiday season, and as we’ll see below, the assumptions and strategies that serve us well for the rest of the year just won’t cut it as we navigate through the holiday peak.

Here are six of the most critical ways historical paid-search trends shift as we hit the holidays:

Shift #1: Click To Order Latency Swings Dramatically

For most of the year, the average latency between a paid-search click and order is pretty stable at around 3.5 days. As we move into early to mid-November, though, we see that latency rises and peaks at roughly 4.5 days.

This reflects the growing number of shoppers who have begun the research process, but will ultimately make their purchases over the course of the next few weeks — many of them timed to the heavy promotions on Black Friday and Cyber Monday. 

Note that here and elsewhere in this analysis, unless otherwise specified, we are associating orders with the day of the clicks that generated them, not the day of the orders themselves.  This is an important distinction and one we’ll examine in detail below.

Thanksgiving highlighted orange; Cyber Monday red

As we hit Thanksgiving and subsequently those two big promo days, average click-to-order latency begins to fall precipitously.  Consumers no longer have an incentive to wait for promotions to kick in and are facing an approaching deadline.

Latency eventually reaches a nadir about a week before Christmas, which is roughly when we hit ground shipping cutoffs. In 2011, the day with the shortest click-to-order window was December 20th. At that point, Christmas shoppers could still place orders with 2-day or overnight shipping and not have to fight the in-store crowds or disappoint their loved ones with hastily prepared homemade gifts.

Shift #2: Average Order Value Falls Off

As holiday shoppers get down to the wire, they also begin to spend less and less per order. In 2011, the largest holiday season orders occurred the week of Thanksgiving, with an average-order-value between $115 and $120. By mid-December, AOVs fell to between $90-95, with a low of $86 on December 24th.

These AOV trends likely reflect a combination of factors:

• First, the mix of non-gift and gift orders should shift from the former to the latter as we move beyond Thanksgiving, showing that shoppers are willing to spend more on themselves than others.
• Gift shoppers that purchase prior to the big promotions around Thanksgiving are likely less price-sensitive.
• The promotions themselves, often heavy price markdowns or a percentage off, can directly drive down average order size.
• Gifts bought at the last minute are more likely to be for people the buyer is not as close to, and therefore the buyer may be inclined to spend less.

There’s also a bit of a chicken and egg situation here as retailers, recognizing some of these factors, choose to highlight items they believe are more likely to be bought as gifts, and these items often have a lower-than-average price point.

Shift #3: Revenue Per Click Skyrockets

Despite decreasing order sizes, average revenue-per-click (RPC) increases sharply over the holiday buying season on the strength of higher conversion rates. On Thanksgiving last year, RPC jumped about 30% over the day before, which had been a pretty typical mid-November day in terms of RPC.

A few days later, we saw a higher peak on Cyber Monday, followed by a drop-off to just above pre-Thanksgiving levels. We then saw steady growth to another peak in mid-December.

While for much of the calendar year, the average performance of the previous 30 days will give us a reasonably close estimate of our actual revenue per click today, the dramatic increases in RPC during the holidays nullifies this connection.

For example, revenue-per-click on Cyber Monday in 2011 was nearly 50% higher than the average over the previous 30 days. Had a retailer been determining their bids by a 30-day average, or even one based on a shorter time window, they would have missed out on a great deal of profitable traffic.

On the flip-side, once we hit shipping cut-offs and revenue-per-click begins to subside, the 30-day average view of revenue-per-click will be overly optimistic and suggest bids that are too high. This scenario will carry over right through the early days of the new year.

Shift #4: Divergence Of Revenue Tied To The Click Day vs. Order Day

Similarly, how we tie clicks and revenue together in our bidding assumptions becomes more critical as we enter the holiday season. Because of the latency effects mentioned earlier, many of the orders that occur on any single day were actually generated by a click that occurred on an earlier date.

If we want to know what our bid should be on a given date, we need to predict the full value of the clicks on that date over the long-term — we can call this the click-day revenue per click.

It’s more common, however, to calculate revenue-per-click for any given day based on all of the clicks and orders that occurred on the same day, even if we know some of those orders were generated by earlier clicks — we can call this our order-day revenue per click.

Unfortunately, as latency shifts significantly, as it does during the holidays, the simpler order-day calculation can become quite inaccurate as a proxy for click-day RPC and lead to poor bids that undervalue early season clicks and overvalue late season clicks.

Shift #5: Normal Day-Of-Week Trends Fall Apart

Given what we’ve seen so far, it is little surprise that historical intra-week performance patterns become less predictive during the holiday season. While there are atypical days throughout the year, often tied to other holidays and/or promotions, the holiday season has a number of incongruous days within a short period.

At the same time, revenue-per-click is either rising or falling sharply in general, depending on where we are in the season.

Comparing day of week revenue-per-click differentials for the holiday season to the rest of the year may not look like there are dramatic differences — Mondays during the holidays look particularly strong compared to the rest of week, but otherwise, the trends appear fairly similar. But, applying weekparting bid adjustments during the holidays in the same manner we might throughout the year can lead to big mistakes.

On Cyber Monday, for example, our annual day-of-week trends in 2011 would suggest revenue-per-click would be down about 2% from the day before. Our holiday-season averages would suggest a lift of 3% from Sunday. In fact, revenue-per-click jumped 23% from Sunday to Cyber Monday.

A couple weeks later, on Tuesday, December 13th, as RPC approached its peak for the season, revenue-per-click was up 8% from the day before, even though our average result for the season would suggest a 12% decline was due. These are not cherry-picked anomalies, but common and predictable results within the larger holiday picture.

Shift #6: Mobile Traffic Patterns Fluctuate More Erratically

It’s no secret that mobile search has grown rapidly over the last couple of years, but that growth has come in fits and starts, often around major device releases, but particularly during the holidays when a lot of devices are being purchased and shoppers are out and about.

Last year, we saw a surge in mobile’s share of paid search traffic in late November, right around the key Black Friday weekend. Mobile generated nearly 12% of clicks at that time, up from 9% just a few weeks earlier.

This year, when shoppers are again hitting up brick and mortar stores in droves, we should expect a repeat of last year’s mobile surge, only with higher stakes. As of early November, RKG pegs mobile share at about 16%. If we see results similar to last year, we could see spikes to a quarter to a third of traffic.

This marks both a challenge and an opportunity for online retailers. Assessing the full value of smartphone traffic remains a significant hurdle, but brick and mortar retailers will need to steer the mobile audience to both their online and offline locations, while pure-plays will have to convince that same audience to go home and buy online, or better yet, buy right then and there on their devices.

But, the issues with tracking all of this activity could make paid search performance appear worse than it really is, especially on those mobile-heavy days. Advertisers should keep an eye on the relative shares of mobile and desktop and not overreact to issues of quantification rather than performance.

Conclusion

With so much riding on such a compressed time-table, our margin for error during the holidays shrinks considerably, and we cannot assume that what works for us during the rest of the year will work for us in November and December. Consumer behavior during the holiday season is unlike that of any other period, and it can change on a dime around critical events like Black Friday and ground shipping cutoffs.

The more accurately advertisers can predict holiday shopping behavior, the better we can prepare and anticipate the changes we will need to make to our paid search programs. Most of the points addressed here are of particular importance to bidding, and retailers should be cautious about being overly reliant on a black box algorithm unless they are certain it is accurately taking into account the insights that savvy marketers can bring to the table.

Following the March release of the new iPad, the tablet market enjoyed a few months of relative calm before the heavy hitters began rolling out a new wave of devices that they hope will finally make serious gains against Apple’s dominance in the space.

In mid-June, Microsoft announced it would co-opt its own Surface moniker for two new tablets due in October, one of which will utilize the touch-friendly Windows 8 OS. Two weeks later, Google announced its first self-branded tablet in the Nexus 7, which hit the market shortly thereafter.

Not to be outdone, Amazon, which has already had the most tablet success outside of Cupertino, announced and released a new line of Kindle Fire tablets earlier this month, including a larger model to compete more directly with the 9.7″ iPad.

Meanwhile, the speculation is that Apple will soon be moving in the other direction with an iPad mini, which would give it a foothold in the lower price-point market where the original Kindle Fire found its niche.

Also, despite its devastating legal loss to Apple, Samsung can’t be written off entirely either, as it moves to play off the relative success it had with the Galaxy Note, but in a more traditional tablet form. Now, even Toys ‘R’ Us is reportedly getting into the game with a low price tablet option.

Tablets Already A Significant Segment Of The Paid Search Audience

From the beginning off September through the end of 2011, tablets roughly doubled their share of total paid search traffic. All of the companies above would love to see that feat repeated this year − so long as that growth is driven by the success of their own devices.

As of mid-September of this year, tablets are generating nearly 10% of all paid search clicks, up from between 7-8% immediately following last year’s holiday surge. As a segment of the average paid search program, that is approaching about 2/3 the combined share of Bing and Yahoo, which should move it up our list of priorities, particularly for well-built, established programs.

Trouble Is, Strong iPad Conversion Rates Are The Exception

A couple months after the original iPad premiered, RKG published results showing shoppers using the tablet converted about on par with desktop shoppers. That conventional wisdom for tablets has been unchanged in the more than two years since. But while the tablet market was the iPad market for all intents and purposes until late last year, that has changed, and so should the conventional wisdom.

In a follow-up study earlier this year, RKG found significant variance in the conversion metrics across different tablets, and those differences could weigh heavily on search marketers if the latest crop of iPad challengers can gain appreciable share over the next few months as they did last holiday season:

Google’s Nexus 7 hasn’t made a big impact like the Kindle Fire, but the holiday season could change that.

While the iPad continues to deliver revenue per click (RPC) on par with desktop search, the growth of its competitors is driving down the click value of the overall tablet segment:

Looking at the latest Q3 results across our client base, RKG found the iPad generating an RPC that is 98% that of desktop search on average, while the rest of the tablet field generates an RPC that is just 40% that of desktop traffic.

If we’re segmenting and bidding tablets as a unified group, the arithmetic suggests we are undervaluing the iPad by about 6% and overvaluing the other tablets by over 100%.

To date, the facts that this segment was relatively small (non-iPad tablets still generate just over 1% of paid search clicks) and the process for bidding segmentation is cumbersome (requiring duplicating campaigns rather than our simply being able to apply an adjustment to a core bid), have made these margins of error acceptable to most.

But, what happens if the iPad competitors generate 2, 3, or 4% of traffic by the end of the year?  For those who haven’t already, it may finally be time to address this performance discrepancy with finer segmentation.

Search Engine Segmentation Options

Although the manner by which we must segment different audiences isn’t ideal, the targeting options we have with the engines are commendably robust, particularly on Google. Already allowing advertisers to target users by their mobile OS − where Bing Ads is today − and even specific OS versions, AdWords added the ability to target specific devices in June.

For most programs, this is probably a little bit of overkill at the moment, but if individual tablet competitors do begin to make headway, it could be a valuable option.

Google AdWords Device Targeting Settings

Taking the simple case of an advertiser only concerned with bidding segmentation (i.e., they are not promoting an app and they wouldn’t benefit from different copy, landing pages, or geo-targets for different devices), a good bet for now would be to group the iPad with desktop and allow the other tablets to fall into a single mobile bucket with smartphones, which convert more similarly.

As the goals of an advertiser take on more complicating factors, such as the aforementioned app promotion or supporting brick and mortar locations, the advertiser will need to layer in additional tiers of device segmentation.

A Shift In The Months Ahead?

One of the bigger tablet questions that remains without a satisfactory answer is the extent to which the observed conversion differences are driven by the differences in the devices themselves or the demographics of their users.

Fortunately, we should get a much better perspective on this in the next few months, especially if Apple does release a smaller version of the iPad.

When RKG originally found the Kindle Fire converting at relatively poor rates, we speculated that, among other factors, the smaller screen size and the comparative sluggishness of the device may have impeded shoppers from using it to complete purchases online. We also wondered whether Fire users were just more frugal than iPad users or more likely to buy from Amazon over the competition. While the new line of Kindle Fires hasn’t been available for long, we have seen early results that suggest the Kindle’s performance for paid search advertisers is improving.

Unfortunately, if Apple puts out a massively successful iPad mini, but it doesn’t convert as well because of its smaller form factor, we’ll have to revisit this issue all over again.

Despite careful studies showing, including one by RKG in 2006, and confirmation by Google in 2009 that conversion performance doesn’t vary much by position, you will still hear pronouncements, even among enterprise level practitioners, that some specific position on the page generates exceptional performance. This piece is not meant to rehash that argument, per se.

Instead, we’ll examine how attributes of the current ad auction including: a Quality Score calculated at auction, broad matching, dayparting, increasingly personalized results, local competition and a changing competitive landscape, all conspire to scatter an ad across the listings regardless of anyone’s best efforts to “own” a particular position.

The Ambiguity Of Average Position

Up until about a year ago, when Google introduced a ValueTrack parameter to pass along position for each ad click, this analysis wasn’t easy to do at scale.

Those using Google Analytics have had a limited view of click position for some time, but for others, short of scraping the SERP, which may or may not have been above board, the best view of ad position was previously Google’s impression-based Average Position, usually viewed at the daily level.

While the name itself clearly highlights it is just an average, and thus our ad may appear across multiple positions, it offers no insight on the extent to which that is occurring.

For example, if an ad has an Average Position of 3.5, did it appear 50% of the time in position 3 and 50% of the time in position 4? Or, did it appear in positions 2 through 5 25% of the time each.

Furthermore, Google’s Average Position tells us nothing about which positions actually drove traffic to our site. For that, we’ll need to assess the click position data we have via the ValueTrack parameter.

Actual Click Position vs. Daily Average (Impression) Position

Our sample here consists of over 120K ad clicks that took place over the course of a single day on about 50K unique keyword instances, i.e., identical keyword phrases in different ad groups or with otherwise different settings are viewed as separate from one another.

To keep the chart below readable, we’ve rounded the daily average impression position to the nearest integer. We’ve also grouped anything over position 10 in one bucket for both the click position and the average position.  This includes any clicks from ads appearing on page 2 or higher.

The size of each circle here represents the percentage of all clicks for that daily average position that occurred at the given click position. Or, the same data in table form:

So, for ads with an average position of 1, 95% of clicks happened when the ad actually appeared in position 1, 3.6% happened when the ad appeared in position 2 and a small percentage of clicks occurred in lower positions.

A number of interesting elements jump out here, including:

• First and foremost, it’s clear that ad clicks really are taking place across a wide array of positions regardless of the average daily impression position.  The spread escalates the lower the average position is, but is apparent throughout.
• Except for in position 1, most clicks do not take place at the same position as the average impression position.
• Click position is shifted upward from the average impression position. This makes intuitive sense as click-through rates increase as we move up the listings.
• We see a relative peak in clicks at position 3 and a relative drop off at position 9. The former likely has to do with ads moving from the side listings to the top of the page at that point.  The latter may be a function of appearing below the fold.
• It isn’t until we get to an average position of 4, that the top position doesn’t generate a plurality of clicks.

Limiting Variables – Google.com & Exact Match Only

To get a sense of the extent to which broad match and the Search Network may be skewing these results, let’s take them out of the equation by looking only at keywords that are restricted to exact match and to showing on Google.com.

We can see that the degree of dispersion has been reduced, but it is still evident and some interesting nuances remain. Clicks are more likely to occur in positions closer to the average impression position, but the position 3 peak is even more prominent.

We also see exactly zero clicks from ads showing in position 9 or 10.

Search Network Position vs. Google.com Position

So, some combination of Search Network impressions and broad matching appear to introduce a great deal of variability in the positions that generate clicks.

Comparing average click position at each average impression position for clicks on the Search Network to those on Google.com, we can see how the first factor affects these figures:

For virtually every average impression position, a click from the Search Network is likely to take place at a lower position on the page than a click on the Google.com domain.

This gap becomes more prominent for average impression positions between 3 and 6, helping to explain the differences between the first two charts above.

Because Search Network traffic generally performs more poorly than that from Google.com, how clicks are split between the two at any given average position, may impact overall conversion rates for a keyword. This effect likely contributes to the notions that certain positions are considerably more valuable than others.

Here we see that the percentage of clicks from the Search Network actually runs lower at top positions on the page and increases consistently to about position 5 before the trend diffuses. This may seem counterintuitive since we expect to be more likely to get picked up on the Network the higher our position.

While the keyword mix could be a factor here, it’s likely that the click-through rate bump we see for higher positions is stronger on Google.com than the Network. This is consistent with the data above showing that Search Network clicks are likely to take place farther down the page at any given average impression position.

Match Type Effects Or Lack Thereof

Somewhat surprisingly, the average click positions of broad matches and exact matches at any average impression position tend to track very closely to one another:

Like Search Network clicks, broad matched clicks tend to perform worse than average. This is because our keyword is not as closely tied to the user’s search and intent as an exactly matched query.

With broad match, we are qualifying for many additional auctions and we might expect more of a bias in how that manifests itself here.

Takeaways

Hopefully this piece has made it clear that search ad position is in a state of flux, even when we account for the larger factors within our control.

SERP personalization and an ever-changing competitive landscape make it impossible to appear in a specific, “optimal” position (even if one existed) anywhere close to 100% of the time. That is, unless you’re willing to pay huge bucks for position 1, in which case Google will gladly take your cash and you may have better luck.

The result of all of this is that the bid to click curve is continuous, just as it appears in any Google Bid Simulator graph. When we increase our bid, we’re not jumping from one position to the next, we’re really changing the distribution where our ads may appear.

Lastly, and maybe this should go without saying, but this data is by no means an encouragement to bid up to position 1 because that’s where so many clicks happen. If you’ve gotten this far and still have that impression, then I have failed.

One recent bright spot for Yahoo was their first year over year growth in search revenues after traffic acquisition costs (ex-TAC)  since their alliance with Microsoft in the fourth quarter of 2010. In fact, before last quarter, Yahoo hadn’t shown Y/Y growth in search revenue ex-TAC since they began reporting it in 2009.

Not too shabby then, and Yahoo’s results were generally well received by the financial industry. The price of Yahoo stock jumped 3% immediately after their Q1 earnings announcement, and it has remained elevated through mid-June even as Google stock and the larger Nasdaq index had fallen 8% and 6% respectively.

But not all signs look positive for Yahoo, nor for advertisers looking to achieve volume increases there. While Yahoo’s search revenue was up 8% Y/Y in Q1, search pageviews fell by 8%, their largest decline since at least 2009. This suggests that Yahoo costs-per-click (CPCs) have increased, ad click-through rates have increased, or some combination of the two.

Yahoo doesn’t break out those metrics, but on their earnings call, CFO Timothy Morse offered that click-through rate growth outpaced CPC growth, but both contributed to their revenue gains. This fits with RKG’s take on Q1 stats, which found Microsoft adCenter-powered ads across Bing and Yahoo delivering 11% higher CPCs and a 29% higher CTR.  But, due to a substantial decline in impressions, we saw click volumes fall 4%.

Things start to get interesting when we dig into what’s really behind all of these figures.

Why Your adCenter CPCs Are Rising

One of the biggest issues Yahoo should have foreseen with the Search Alliance was that Microsoft’s ad serving logic in adCenter was far more restrictive than Yahoo’s own system was at the time, as well as Google’s AdWords.

When RKG looked at this last year, we found adCenter was broad matching our ads to queries just 40% of the time for Yahoo, compared to Yahoo’s 60% broad match rate prior to the Alliance. This surely contributed to Yahoo’s paid click declines since late 2010.

In August 2011, announcing changes to their matching behavior, the adCenter blog advised advertisers to be “prepared to take on additional traffic.” Well, we didn’t see the types of changes we were hoping for, namely, additional quality non-brand traffic. Instead, what we noticed was a dramatic increase in broad matching for our clients’ own brand, or trademark, terms. This trend only continued into 2012.

By the beginning of Q2 2012, adCenter was broad matching brand terms at nearly twice the rate that Google was doing the same. The two engines had been at parity in this respect in mid-2011. This has led to much higher brand CPCs on Yahoo and Bing in two ways:

1. Competitor ads are being matched to our clients’ brand terms at greater rates, driving up our CPCs for auctions that were previously far less competitive.
2. Our own brand ads are being broad matched to the brand phrases of others, as well as more competitive non-branded auctions, at greater rates.

The wheels were set in motion for this in March of 2011 when adCenter loosened its trademark policies to bring them into alignment with Google’s. As adCenter’s broad matching on brand has nearly doubled, so too have their CPCs, even as Google brand CPCs have steadily declined.

While we have little control over the first behavior above, there are some tactics we can employ to alleviate the impact. On the second point, there’s a great deal we can do and it’s just a matter of due diligence for a segment that is often too easy to overlook because of its relatively strong performance.

Tips To Keep CPCs For Your Brand Terms In Check

1.  Watch your brand bids

Or better yet, watch your bidding logic. While it may sound simple, you can’t pay more than you bid, and you may be paying far more than necessary to maintain visibility on your brand terms. At a basic level, paid search bidding systems seek to maximize a conversion metric given a certain efficiency constraint.

If your efficiency constraint is, say, a simple sales to cost ratio that applies to all keywords in your program, you’re likely to end up with very high bids on your brand terms, due to their great ability to convert.

This leaves you at risk of actually having to pay those high bids, or close to them, if the engines find clever ways of making you. This could include the engines increasing broad matching for queries with the highest bids, employing top of page thresholds that act as artificial competitors and so on. The bottom line is we can’t assume that if we are in position 1, our bid doesn’t matter because of the supposed nearest-competitor-plus-a-penny auction mechanics. It certainly does on Google.

Simply put, just because we can afford a certain bid in aggregate, based on an overall efficiency target, doesn’t mean we should or need pay it. Instead, advertisers should be considering the marginal impacts of raising and lowering their bids and even the extent to which paid listings cannibalize organic traffic.

Because brand terms are fundamentally different from the rest of your paid search program, one approach to get at this question is to simply test applying a much stricter efficiency target for your brand segment. This will likely lower the CPCs you’re paying, without having much impact on the traffic you receive from queries on your brand.

2.  Be vigilant with negative keywords and match types

If we’re finding our brand terms being broad matched to distantly related queries, we might get upset with the search engines, but ultimately we only have ourselves to blame.

Researching negative keywords should be a routine part of any paid search program and, once again, we can’t overlook the brand segment just because it performs more efficiently than the program as a whole.

When negatives aren’t enough to stem the traffic from poorly matched queries, we may want to adopt or limit our keywords to more restrictive match types. At RKG, we generally choose to run individual keywords on multiple match types with bids appropriately set for each, based on expected performance differences.

Limiting a keyword to only exact match is the nuclear option, but sometimes it’s necessary. In most cases though, retaining phrase, exact and utilizing broad match modifiers will do the trick when normal broad match proves untenable.

3.  Properly funnel traffic to the appropriate keywords

The truth is, even very broadly matched traffic, whether it is from a generic phrase or even a competitor’s trademark, can provide a great deal of value to us. But just because it is valuable doesn’t mean the situation is ideal. We would much prefer to have a keyword that matches the query more closely, so we can better target our copy, choose our landing page, and set our bids.

Even small differences between a query and keyword are predictive of a lower revenue per click than an exact match, and that gap only grows as the match gets more broad. When the same keyword triggers for very close queries and very broad queries, its overall performance stats get diluted and we end up with a bid that is too low for some of its clicks and too high for others.

As with adding negatives, we should routinely be scouring our search query logs to look for valuable keywords to add to our accounts. There isn’t a guarantee that adding a new keyword will prevent an existing one from showing for that query, but it should help in conjunction with the judicious use of negatives and match types mentioned above.

Progress On Non-Brand?

To close on a positive note, we are seeing signs that the adCenter engineers are making progress at expanding broad matching for non-brand terms, while, importantly, preserving click quality.

This is the segment where there is an opportunity for Bing and Yahoo to deliver real incremental value to advertisers, while improving their own fortunes at the same time. And, although we should strive to have robust and extensive term lists, broad match will always serve an important role in paid search.