Showing posts with label SERP. Show all posts
Showing posts with label SERP. Show all posts

Tuesday 26 July 2011

Mission ImposSERPble: Establishing Click-through Rates


Google and its user experience is ever changing. For a company that has more than 60% of the search market, it's common to hear the question, “How many visitors can we expect, if we rank [x]?” It’s a fair question. It's just impossible to predict. Which is a fair answer. But, as my father says, “If you want fair, go to the Puyallup.” So we inevitably hear, “Well, can you take a guess? Or give us an estimate? Anything?”
To answer that question, we turned to major studies about click-through rates, incuding Optify, Enquiro, and the studies released using the leaked AOL data of 2006. But these studies are old; this study is new. Ladies, Gentlemen, and Mozbot, it is our immense pleasure to present to you…
The Slingshot SEO Google CTR Study: Mission ImposSERPble
There have been a number of changes to the Google user experience since those studies/surveys were published years ago. There's a new algorithm, a new user interface, increased mobile search, and social signals. On top of that, the blended SERP is riddled with videos, news, places, images, and even shopping results.
We made this study super transparent. You can review our step-by-step process to see how we arived at our results. This study is an ongoing project that will be compared with future SERPs and other CTR studies. Share your thoughts on the study and the research process to help us include additional factors and methods in the future.
Our client databank is made up of more than 200 major retailers and enterprise groups, and our sample set was chosen from more than thousands of keywords based on very strict criteria to ensure the accuracy and quality of the study results.
The study qualification criteria is as follows:
  • A keyword phrase must rank in a position (1 to 10)
  • The position must be stable for 30 days
Each keyword that we track at Slingshot was considered and every keyword that matched our strict criteria was included. From this method, we generated a sample set of exactly 324 keywords, with at least 30 in each of the top 10 ranking positions.
We are confident in the validity of this CTR data as a baseline model, since the data was generated using more than 170,000 actual user visits across 324 keywords over a 6-month period.
Data-Gathering Process
Authority Labs: Finding Stable Keywords
We currently use Authority Labs to track 10,646 keywords' daily positions in SERPs. From this, we were able to identify which keywords had stable positions for 30 days. For example, for the keyword “cars,” we observed a stable rank at position 2 for June 2011.
Stable 30 day ranking - ImposSERPble
Google Adwords Keyword Tool: All Months Are Not Created Equal
We found the number of [Exact] and “Phrase” local monthly searches using the Google Adwords keyword tool. It is important to note that all keywords have different monthly trends. For example, a keyword like “LCD TV” would typically spike in November, just before the holiday season. If you’re looking at searches for that keyword in May, when the search volume is not as high, your monthly search average may be overstated. So we downloaded the .csv file from Adwords, which separates the search data by month for more accuracy.
Google keyword tool csv download - ImposSERPble
By doing this, we were able to calculate our long-tail searches for that keyword. “Phrase” – [Exact] = Long-tail.
Google Analytics: Exact and Long-Tail Visits
Under Keywords in Google Analytics, we quickly specified the date of our keywords’ stable positions. In this case, “cars” was stable in June 2011. We also needed to specify “non-paid” visits, so that we were only including organic results.
Google analytics non paid - ImposSERPble
Next, we needed to limit our filter to visits from Google in the United States only. This was important since we were using Local Monthly Searches in Adwords, which is specific to U.S. searches.
Google analytics phrase and exact - ImposSERPble
After applying the filter, we were given our exact visits for the word “cars” and phrase visits, which included the word “cars” and every long-tail variation. Again, to get the number of long-tail visits, we simply used subtraction: Phrase – Exact = Long-Tail visits.
Calculations
We were then able to calculate the Exact and Long-Tail Click-through rate for our keyword.
EXACT CTR = Exact Visits from Google Analytics / [Exact] Local Monthly Searches from Adwords
LONG-TAIL CTR = (Phrase Visits – Exact Visits from Google Analytics) / (“Phrase” – [Exact] Local Monthly Searches from Adwords)
Results
What was the observed CTR curve for organic U.S. results for positions #1-10 in the SERP?
Based on our sample set of 324 keywords, we observed the following curve for Exact CTR:
Google CTR curve - ImposSERPble
Our calculations revealed an 18.2% CTR for a No. 1 rank and 10.05% for No. 2. CTR for each position below the fold (Positions 5 and beyond) is below 4%. An interesting implication of our CTR curve is that for any given SERP, the percentage of users who click on an organic result in the top 10 is 52.32%. This makes sense and seems to be typical user behavior, as many Google users will window shop the SERP results and search again before clicking on a domain.
Degrees of Difference
CTR study comparisons - ImposSERPble
The first thing we noticed from the results of our study was that our observed CTR curve was significantly lower than these two previous studies. There are several fundamental differences between the studies. One should not blindly compare the CTR curves between these studies, but note their differences.
Optify’s insightful and thorough study was conducted during the holiday season of December 2010. There are significant changes in Google’s rankings during the holiday season that many believe have a substantial impact on user behavior, as well as the inherent change in user intent.
The study published by Enquiro Search Solutions was conducted in 2007 using survey data and eye-tracking research. That study was the result of a business-to-business focused survey of 1,084 pre-researched and pre-selected participants. It was an interesting study because it looked directly at user behavior through eye-tracking and how attention drops off as users scroll down the page.
Long-Tail CTR: Volatile and Unpredictable
For each keyword, we found the percentage of click-through for all long-tail terms over the same period. For example, if “cars” ranks at position 2 for June 2011, how much traffic could that domain expect to receive from the keyword phrases “new cars,” “used cars,” or “affordable cars?” The reasoning is, if you rank second for “cars,” you are likely to drive traffic for those other keywords as well, even if those positions are unstable. We were hoping to find an elegant long-tail pattern, but we could not prove that long-tail CTR is directly dependent on the exact term’s position in the SERP. We did observe an average long-tail range of 1.17% to 5.80% for each position.
Google CTR data table - ImposSERPble
Blended SERPs: The “Universal” Effect
Starting in May 2007, news, video, local, and book search engines were blended into Google SERPs, which have since included images, videos, shopping, places, real-time, and social results. But do blended SERPs have lower CTRs? Since these blended results often push high-ranking domains towards the bottom of the page, we predicted that CTR would indeed be lower for blended SERPs. However, a counter-intuitive hypothesis would suggest that because certain SERPs have these blended results inserted by Google, they are viewed as more credible results and that CTR should be higher for those blended SERPs. We analyzed our sample set and failed to show significant differences in user behavior regarding blended versus non-blended results. The effect of blended results on user behavior remains to be seen.
Google CTR blended data table - ImposSERPble
As previously mentioned, this study will be used in comparison to future SERPs as the Slingshot SEO Research & Development team continues to track and analyze more keywords and collect additional CTR data. It is our hope that these findings will assist organic SEOs in making performance projections and consider multiple factors when selecting keywords. We look forward to additional studies, both yours and ours, on CTRs and we encourage you to share your findings. With multiple prospective and recent social releases, our research team will be dedicated to examining the effects of social platforms and Click-through rates, and how the organic CTR curve changes over time.