Click Bias and the Modern SERP
If there is one thing we know in the world of SEO; a #1 ranking is everything! But why is that? Is that because the top ranked documents are always the best? No, not really. But that’s because the end user implicitly trusts the search engine (or are lazy?) and they tend to click there more often than any other position. And I thought we could look at some user behaviour some more today and see what it can teach us. ( we looked at a white paper the other day on search engine user behaviour as well for those interested).
When we start to discuss this end user behaviour we need to first address what is known as ‘click bias‘. And it must be known, there certainly is a prevalent bias. Search folks have done a lot of testing (as have search engines) where they will mix up the results and even put the less relevant listings in the top positions, the more relevant further down. Guess what? Yup, people still tend to click the top listings, regardless.
Making a mess of implicit feedback
This is one very important consideration for search engineers when trying to glean signals in areas such as implicit user feedback, including click data. SEOs are often quick to state they believe search engines use this kind of data. It is an erroneous assumption. You see it is this type of click bias by the end user that creates noise and makes the signals less trust-able.
Once more, if we looked at people’s query and click history to see which potential listings are of value, we’d also need to assess the problems relating to click bias’. Because people will click top listings habitually, we can’t assume that the top documents were necessarily the right ones being returned.
As you can see here, people (in this study) were clicking on the top 2 listings with a staggering 70% rate of interaction. As noted already, other studies they have mixed 10 to 1 in various configurations and the data remains the same. It is NOT about the relevance as much as it is bias.
What is also interesting is that people also seem to ‘trust’ the results more as far as time spent as well. While the time spent does not degrade proportionately with the click bias, it is also quite telling in the first 3 listings of the SERP. Here’s another study;
In this one we can see that most certainly the first position was getting the lion’s share of the clicks, but the time on page degradation is once again prominent in the first 3 listings of the SERP. This does seem to imply that much of the click bias is either strong listings returned or an implicit trust of the results.
Next we can have a look at some heat maps for more reinforcement;
In these examples we can once again see that the gaze fixation is steadily above the fold and this time, influenced by the inclusion of an image in the SERP. This speaks to the new world of SEO where we simply must be aware of any universal search elements in the query spaces we’re targeting.
The bias tends to have an even narrower focus when they come across images, videos and other universal results presentation. This would further degrade and value to rankings below position 2-3 in the SERP, (depending where the universal element lies).
Personalization Moves the Needle Further
And of course we can’t stop there. In this modern age and the rule of Google, we must consider the world of personalization. Why? Because this further influences the end users SERP click bias as well. Here’s a little chart from another recent study;
This time they gave users two sets of results over two subsequent search sessions. First asking them to perform a task on the regular results and then the next day, manually adjusting the results with sites/listings that the end user had looked at in the first session.
As you can see, the study group had a serious bias towards the listings that were tailored to them from the original session. This should be obvious, we must also consider personalization as having the ability to further bias the end users behaviour.
Over the last 6 months, post-Caffeine, Google’s SERPs have also undergone a number of changes which can also have a huge impact. In particular;
Real time / social ““ these listings started taking up some real estate and can potentially syphon clicks from the regular listings.
Local Shake up ““ while I’ve not seen any data on the mass localized listings, but I’d have to assume it’s much like the video/image verticals as far as various pak sizes getting more clicked.
Google Instant ““ this one isn’t likely as much of an issue as most studies have shown there is little change in user behaviour.
Google Instant Preview ““ here we haven’t seen much data either, but we can generally assume that uglier sites will do worse. This might actually have an effect on the ‘first link’ click bias, but I doubt it will be huge.
While we’re witnessing a TON of changes over the last 6 months, much of it has lost prominence as seems to be the case with other universal elements. What is important to remember is that placement and visibility (does the presentation encourage clicks as we’ve seen with Video/Images) are going to be the main factors in how CTRs are affected.
Beyond the 10 Blue Links
So what can we take from all this? Well, that part seems rather straight forward to this wandering Gypsy;
- The land of 10 blue links is dead
- People are biased via trust or need for efficiency
- Users like shiny objects (images, videos, maps)
- Personalization makes a HUGE difference
We really do have to be paying attention to the evolution of the SERP. If you are still grasping tightly to the tactics of years past, you will fail. A full featured SEO campaign must focus on far more than just a bit of on-page and massive amounts of links. It must now, more than ever, consider a strategy that makes the most of all the relevant verticals for your market.
Of these, the least obvious, but certainly important, is personalization. How does one even do that? Make a site that is sticky. Give people a reason to come back time and time again. Enable social and other sharing mechanisms (add to favourites, email to a friend). Consider time on page (dwell time), page depth, bounce rates and other implicit feedback.
The next time someone says, “SEO is dead“ or related ignorance, feel free to straighten them out. Search is evolving more than over over the last 10-12 months. This means SEO is also evolving. If YOUR strategies and tactics aren’t evolving with it, you will find yourself (and your clients) out in the cold real soon. Adapt or die my friends… adapt or die.
If you missed it be sure to see Monday’s post on search engine user behaviour. Here’s some goodies related to this post worth checking out;
- Eye Tracking and Online Search: Lessons Learned and Challenges Ahead (PDF)
- Google on presentation bias in search
- The Value of a Google Result Position
- Google Organic SEO Click-through Rates
- Value of a Google Ranking
- Distribution of clicks on SERPs and eye tracking
- SERPs user behaviour eye tracking study
- Eye-Tracking Analysis of User Behavior in WWW Search (PDF)
- Eye tracking on personalized and universal search
Dave, solid stuff, as usual.
I do think that search engines might use CTR as a minor factor though. Since they know about click bias, and they have more accurate clickthrough data than we do, they can discount the click bias.
Say a site in 5th position is getting way more clicks than their average 5th-ranked result for that type of query. Google might think, “hey, for some reason users prefer this site. let’s put it 3d and see if it outperforms our average CTR for that spot”. Makes sense to me.
I just plain loved this part:
“Of these, the least obvious, but certainly important, is personalization. How does one even do that? Make a site that is sticky. Give people a reason to come back time and time again. Enable social and other sharing mechanisms (add to favourites, email to a friend). Consider time on page (dwell time), page depth, bounce rates and other implicit feedback.”
Looks like Google is pushing the lazy website owner to do what he’s supposed to do. Shouldn’t we want to make our sites stickier, more interesting and relevant and helpful? Well, now we have another incentive to do so – rankings.
Thanks for the post, really enjoyed it.
Excellent research and writeup, Dave!
“Adapt or die my friends… adapt or die.”
And can you imagine an SEO who doesn’t watch the SERPs? Oh, I’ve met quite a few. Your post explains another important caveat – don’t rely on ranking tools! Know thy query space!
hehe…. aye, tis a foreign concept at times that’s for sure. As for ranking tools, interestingly I posted some tidbits on the Dojo boards of interest. Drop by real quick if ye get a chance. But most certainly I’d be trying to get localized data to bloster any analytics.
Comments are closed.