Google Instant rolled out September 8th to much fanfare and ballyhoo. The search marketing industry has been abuzz ever since with speculation about the impacts on both paid and natural search.
Our firm has taken a pretty close look at the initial impact on paid search performance and we want to share our findings with you folks.
We studied Google AdWords data from the period prior to the launch of Instant and compared it to the first week-plus following the launch for a wide range of clients. Our client base is heavily retail, so those in other sectors may have different findings.
We looked at the impacts both in aggregate and by advertiser to see if averages hid meaningful shifts. We looked exclusively at data from competitive, non-brand search terms.
We tried to answer the following questions:
- What impact has Google Instant had on impressions and clicks on paid search ads overall?
- Does Instant create a greater emphasis on ads served at the top of the page and diminish the traffic on ads served closer to the bottom of the page? Some speculated that Instant would bias users against scrolling and effectively increase the incentive for higher positioning.
- Does Instant help or hurt the long tail of paid search? Some have wondered if watching the results change as you type would encourage users to keep typing as results get more and more targeted, or on the contrary encourage more to stop early and click on the first mildly relevant link.
- Does Instant help or hurt conversion rates? The stated goal of the product is to get people where they want to go faster. Does it also help them find more relevant ads?
- Does Instant impact some types of keywords more than others?
Question #1: Overall traffic. Initially following the Google Instant launch we did see a small spike in Google impressions and clicks, both from week to week and relative to our traffic levels on Bing and Yahoo. It should be noted that part of the week to week increase benefited from favorable comps to a slow Labor Day weekend, while, even at it’s peak following the launch, Google’s traffic share just slightly cracked two percentage points above its 30 day average.
We’re not sure whether the media coverage around the Instant launch delivered a genuine lift in impressions and even click traffic or whether this blip is more related to roll-out glitches. Clearly whether Instant will garner Google even greater market share, or will ultimately be little more than an update to the suggest function remains to be seen.
Question #2: Top vs. bottom.
We haven’t seen a large shift in traffic composition between ads at the top of the page and those at the bottom for most advertisers.
We looked at a handful of our larger accounts and measured the baseline fraction of traffic that comes in from ads in the top 3 positions to see if that fraction has materially increased.
As the chart shows, the median shifts aren’t zero, but they aren’t huge either. The median advertiser saw a slight dip in the number of ads in the top 3 positions—likely unrelated to Google Instant—but a slight increase in the fraction of click traffic those high positioned ads represent. Very interestingly, the fraction of orders increased slightly more than the fraction of clicks, possibly indicating improved conversion; it is not crazy to suggest that the people who stop typing early to select an ad do so precisely because they’ve read the ad more carefully than the average searcher has done historically.
On the flip side, have we seen a relative decrease in the importance of ads in position 8 through 12? It bears mentioning that since these already account for a fairly small portion of the traffic for most advertisers, the percentage changes are easier to influence materially. For example, if these low positioned ads usually account for 5% of traffic and that drops to 4%, that shows up as a 20% decline in the table below.
Understanding the spikiness of this data set (particularly the order counts), this data isn’t alarming either.
Question #3: Head vs. tail. This is a bit more interesting. Here we picked an arbitrary number of clicks as defining a “head” term, and looked to see whether these head terms had become more or less important as a percentage of the whole.
The first observation is that the number of keywords meeting the definition of a head term increased for most advertisers. That could simply be a seasonal increase in traffic, but in two-thirds of the cases studied the increase in head ads was greater than could be explained by sheer traffic volume.
These ads did drive a larger fraction of the total clicks and sales for most advertisers in the week after Google Instant launched, but given the increase in the number of ads, this by itself doesn’t mean the tail has been significantly diminished.
Question #4: Conversion rates. Early, early indications are that conversion rates for head and tail may have improved as a result. Not ready to proclaim victory on this front just yet, and we’re not talking about a large change, but we’re seeing some indications that this could create a bit of a virtuous cycle.
Question #5: Keyword-level effects. While there may not be a sea-change in overall performance, there are some interesting keyword-level effects that can be tremendously important for some advertisers.
For example, we’ve noticed a huge shift with respect to treatment of singulars and plurals, with the more popular of the two seeming to become the default. For advertisers with tremendous volumes of business tied to a handful of terms, this can be a very big deal.
For several advertisers, ads running on competitor’s trademarks and domain names seem to have dropped off the map. This may be an Instant effect, or possibly a change to Quality Score algorithms. We haven’t studied this comprehensively enough to guess whether this is universal.
We’ve also noticed some odd effects on keywords that have other completely unrelated meanings. As an example: if you type in “toothpaste” all the results are geared towards “toothpaste for dinner” with no ads showing. This is true even after you’ve typed the whole word with a space after it. When you hit “enter” the ads for toothpaste appear.
The reverse of this is what Glenn Edelman of Wine Enthusiast refers to as “short typing.” You’re looking for “Wine Enthusiast” and you get to “Wine En” and see the results you want. You reflexively hit enter and Google now brings back results for “wine en” which turn out to be different. We’re not sure this is a large enough effect to worry about, but if it is, please refer to it as “short typing” and credit Glenn!
We strongly encourage everyone to do a keyword level report from the period before Google Instant launched and compare it to a similar report for the period after. Use a vlookup to match the traffic volume on the keyword after, to the traffic volume on the top keywords before and vice-versa. The results are fascinating!
At this point, we don’t see cause for massive alarm for most advertisers in paid search. I’d love to see more data from the SEO community to see if the impact is more pronounced there.
As always, your mileage may vary, so dig under the hood and see what you see!
Many thanks to Mark Ballard, RKG’s Senior Research Analyst who helped me research and write this post.
Hope to see you folks at SMX East! Drop me a line at firstname.lastname@example.org if you want to grab a cup of coffee or a beer at some point during the show.
Opinions expressed in the article are those of the guest author and not necessarily Search Engine Land.