You may have have read how at SMX Matt Cutts reiterated that Google does not look at   bounce rate as part Google  algorithm.  Not so fast!  Matt actually dodged Matt's question .... and it seemed to slip by most SEO reports of the conference. 

Danny Sullivan certainly does not shy away from asking Matt pointed questions at SMX.    I enjoy the questions  and despite Matt's sometimes vague answers --- I feel I there is a lot to learn from what Matt answers, how he answers and what he avoids.  

I am specifically interested in user metrics collected on the SERP (as I discussed at SMX) as parts of Google's algorithm, particularly Panda.   Danny actually drilled in on this ... and Matt danced around by answering a different question.   I have read every account of the session I could find and not a single post  that I have found understood the difference between Danny's question and Matt's answer (except for Danny).  This is very significant.    If I hadn't been there and had only read about it, I would have a totally different view of the conversation.
To roughly paraphrase the conversation:

Danny:  “I know you have said that Google does not use Google Analytics bounce rate. What about bounce back to SERP and behavior on the return to SERP?” 
 Matt: “GA bounce rate is a very noise signal. We never use it". (He spends about 5 minutes not answering the original question - the question was about bounce to SERP - not Google Analytics bounce rate). 
Danny: What about bounce to the SERP?"   
 Matt: We do not use GA data, it very noisy signal. 
Danny: No, I am talking about the SERP and bounce to SERP. 
 Matt: Well, ummmm…. we don’t like to rule certain things out for the future ….ummm…..ummm.

Matt simply refused to answer this question until Danny gave up.    This is extremely revealing!  There is no reason not to answer if they are not using this factor.   However, if they are using it, they might want to hide it if they believe it that knowledge can be leveraged by SEOs.   So why not lie outright?  It has been pretty clear to me from seeing these exchanges that Matt likes to be as forthright as possible.  So this was a flat-out yes to me.  SInce,  I haven't seen any other blog or account of the Q&A pick up on this,  - it  appears he successfully dodged the question and confused the majority of the SEO community. 

In fact,  we know absolutely 100%  that Googles does look at bounce to SERP as part of PANDA!   The "Block Site" link only shows up on bounce back to SERP.  So there is no question about this.  What exactly they are looking at besides block is not clear.  I have speculated that the alternate click (telling Google that the quest for an answer to their question is not over) as well as very quick bounces.   Previously I had assumed that bounce back to SERP was not a valuable metric in itself (rather it is the other actions).  Now I am not so sure.     

Many of the recent changes in Google's algorithm that are demoting and promoting certain results are based on thresholds. Recent major algorithm changes which seem to be threshold based include Panda (user metrics), PageLayout/Above the Fold Penalty (page content analysis), Over Optimization Penalty (Linking analysis).

Google, in their typical complete lack of openness in regards to their organic algorithms, does little to let you know where you are in regards to these algorithms). Although there have been some sites getting notices of links that look like paid links, they still don't tell you where you are in regards to the threshold.  And there has been no communications on Panda or Page Layout.

Think about this ... you could be knocking on the door of Panda, Page Layout Alog,  or Link Over Optimization, meaning you could be one content initiative, page redesign, link building initiative or other site change away from triggering a loss of 30%-80% of your organic revenue (which of course is the best type of revenue.)   How many businesses can survive that? Say the passing score for Panda is 65. You have site A with a score of 66 , fat, dumb and happy as a clam even though  they are teetering on the verge of a specific hit.

And then you have Site B, maybe they failed with a score of 64  lost 50 % of their traffic and they are firing people or shutting their doors. Yet they might be hanging right below the threshold, not much different from Site A. Doesn't see fair, does it? Site C might be well beyond repair and a truly crappy site but the impact on site B and C might be indistinguishable. 

So what is the poor SEO to do? We need to practice defensive SEO. We don't wait to get hit.  SEO is no longer about spewing out reems of content, at least without thought to site relevance, content quality and usability. In fact, good SEOs are going to be learning quickly to integrate search usability into their methodology. Bad SEOs are still going to be pushing out content and outsourcing link building with no regards to quality of links or spaminess of the techniques. As I mentioned when I first started this blog, good SEO has changed and should no longer be at odds with usability - rather it needs to be tightly integrated. We rarely get rewarded for saving traffic - but we need to do it anyway.

I have mentioned briefly before what a flawed metric bounce rate is.  Michael Martinez wrote a great post in far more detail on the same subject which got me pondering this subject even more.   While I agree that it is a flawed metric, I disagree in that it is still a number I want to have and analyze. To be sure:

• As an absolute number, bounce rate is completely meaningless.   Who knows how long a user spent reading a page even if they bounced?   The amount of time required to satisfy a user might vary depending on the page.  A 15 second visit might be more than sufficient if all a person was looking for was a phone number!  I have seen sites with great bounce rates that did get hit by Panda (in one there was no content above the fold, just links - so they got a lot of clicks - and a lot of frustrated users I imagine).   Conversely I have seen sites with very high bounce rates spared by Panda. 
• It is very easy to degrade user experience and improve bounce rate.  For example, consider a hotel page where users really want to find a phone number. If that was one of the drivers to traffic, and if it was moved off the main SEO page (perhaps to a contact hotel page) navigation to the contact page would go up (and bounce rate down).   However,  a critical piece would now be missed by most users.   More people would return to Google to continue their quest (and this gives negative metrics about your site). 
• Implementation can impact these numbers even though the experience might be identical.  (Think about the use of Ajax vs. rendering new pages).
• Why would Google ever risk using a product which is strategically important to them to potentially inflict damage on the customer using the product?   Even if it was a good metric  - Google could not be that stupid.  Plus, not everyone has Google Analytics and it would not make sense for them to use a metric that is not available for all sites.

As Michael points out in his post, sometimes a bounce is an indication of a good visit (user got what they want). 

Let's look what happens with a good visit (keep in mind that I am using the Google analytic definition of 'bounce' - that is, a single page visit). 

1. Users enter a query
2. Google returns your page and user clicks over.  This should be your most relevant page.   This means that if Google did it's job on relevance and you did your job with delivering the relevant message on the page - the users has no need to navigate to another page on your site.  
3. Assuming you have a good page design, the user finds the content he was looking for.   He then either goes on his merry way ( a bounce which I will call action A), or finds your site so engaging that he continues to browse, register, buy some products or something (positive engagement - call it action B).  So therefore, Action A = Good Bounce,  Action B = good engagement.  

Now, how about a negative visit!

1. User enters a query
2. Lands on your site and does not find what he is looking for
3. Some users bounce (call that action C) and some users may click around to find their answer (call that action D).  

So in this case,  Action C = Bad Bounce and Action D = Bad Engagement. 
So how do you know what percentages of your bounces are good vs. bad?  And what percentages of your engagement is good vs. bad.  You can't tell by bounce rate - you just don't know.   When the ratios change you don't know what is driving it.  

Suppose  you make changes to your site and your bounce rate goes up.   Is that a bad thing?  You don't know changes in good bounces (A) vs.  bad bounces (B).     By looking at the numbers, you can't really tell.  So trying to draw conclusions from bounce rate alone is a dangerous things.   But that does not mean you should not look at bounce rates! 

So should SEOs look at bounce rate (or CEOs or anybody?).    I still say yes, but not as an absolute metric as it is not meaningful by itself.   Really, you need to look at all the factors impacting bounce rate, conversions, knowledge of your users, site changes and more  to understand whether your site has gotten better or worse to reach some sort of interpretation of what your  bounce rate means.       Also, remember to segment your visits down to the smallest segments that gives you statistical significance.  Your goal is to understand if users are being satisfied and the bounce rate is another piece of information in your quest to find the answer to that question. 
You have to look carefully at the change to truly understand the shift in bounce rate.   Again, if you remove critical content,  bounce rate may go down and user dissatisfaction go up.   

So bounce rate does matter to me, but only as an additional piece of information to help provide insight into user experience from SERP.   As an isolated number, it is meaningless.  Meanwhile we still have a huge gap with what we know about our site and with what Google knows about our site.   Addressing that gap is key to good SEO in an AfterPanda world.   I'll be writing more about that in upcoming posts.  

Google Freshness update hit the search engine result pages (SERPs) a couple of weeks ago and like any major update (especially in the wake of Panda) created stress and anxiety for websites everywhere as fear of another search apocalypse started flooding the SEO forums. The particular angst inducing statement from Google was specifically:

“today we’re making a significant improvement to our ranking algorithm that impacts roughly 35 percent of searches and better determines when to give you more up-to-date relevant results for these varying degrees of freshness”

Nothing makes website owners more nervous then when Google announces improvements!  

35% certainly sounded like this would have result in Panda-like tremors – but as the traffic reports and discussion started hitting the Internet, it seemed more like a ripple as opposed to another quake. So what is going on here?

When analyzing Google algorithm changes  (and what Google will do in the future), I like to put on my product management hat to try and understand exactly what problem they are trying to solve as it pertains to their product - the organic search results.

Problem number 1: When users are searching for trending/hot topics, content relating to that hot topic would get very little or no coverage in the standard organic results on the SERP. Most likely it would show up in Universal Search of the SERP in the form of news, but the other 10 results would most likely have no content related to the search – essentially giving irrelevant results. Strong, high PageRank pages would dominant over the newer, fresher, more relevant content which yet to garnish any links or reputation. So for instance, if you searched for “NBA Talks” you would see organic results that looked like this:

· NBA.com – would probably head the list because of high page rank and strength for the keyword “NBA”. Very poor relevance.

• Articles talking about the pending NBA lockout from many months ago – this would seem relevant based on the keyword search and likely had strong PageRank and inbound links since they had been around a while. However, this is not what the user is looking for and a poor result for Google.
• Very little new content relevant to the breaking story.

So this part of the update seems to be all about news and trending search terms. By looking at new content being indexed, news content along with dramatic jumps in search volumes for specific phrase, Google can conclude that a certain phrase needs fresh content. In this case, I believe fresh means newly, indexed pages relevant to hot topic. So now if you search for NBA Talks now, you see this:

Wow – forget about the old, high PageRank authoritative content. The results are dominated by new content. In fact 100% of the first page result are fresh and in this case news content. This is a far superior result for than old algorithm which mostly showed older content.

However, search for “NBA”, you see none of that fresh content in the results. Clearly, this part of the Algorithm impacts very specific searches related to trending news, but only for search terms clearly relevant to the trending term.

Problem #2: For certain type of queries, a good result is dependent on freshness. Google specifically gave the example of reviews. If you are searching for reviews, you want to see recent reviews which pertain to the latest version of a product or represent the current state of a business. However, Google’s algorithm returned results based on the authority of the pages which would surface the most relevant content from the strongest pages. The content was relevant to the explicit search term (for example ‘iPad Review’). However, most users probably want to see iPad 2 reviews, even though they didn’t explicitly type that in the search box.

In this case, fresh content can certainly mean long-standing pages that reflected the latest information via updates as opposed to the first case I discussed which appears to be news specific. In analyzing the search results, it still seems that strong, authoritative, content dominates, but with perhaps a tweaking based on freshness.

With a little help from SEMRush, I am able to compare the changes in the search result for ‘iPad Reviews’, pre and post Freshness change.

• Before the change – the top ranking page was an Engadget review for the original iPad 1 from nearly 2 years ago. Probably not the review most users were looking for.
• In fact, the entire first page of results was dominated by old reviews.

Now, after the change, things look quite a bit different.

• This number 1 slot has been replaced with an iPad 2 review, also from Engadget, indeed a much better result.
• The number 2 slot has been replaced with an iPad summary page from CNN with a link to an iPad 2 review (way down below the fold). This is not really a good result (it would have been better to list an iPad 2 review first). It shows freshness in action, but also that high PageRank, authoritative pages can still outrank the more relevant page (with pinch of freshness mixed in).
• Number 3 is still an iPad 1 review (very authoritative with 396 linking root domains, so it is trumping freshness).
•  Number 4 is actually a Kindle Fire review which mentions iPad – very fresh (48 minutes old as I write this) – and is an example of how trending stories can make it into the result even if the search term is not the hot-topic.
•  Slot 6 is another iPad 2 review.
• However, slots 5 and 7 are really poor results, search result and tag pages, yes, the type of pages everybody has been told are bad and should be de-indexed in the wake of Panda.

Overall, there are better results for the user looking for up-to-date information with a couple of iffy pages mixed in where freshness is trumping quality. Google still has work to do on this.

So what are the implications and how do they affect our SEO strategies going forward? These are my initial thoughts, however since it is in the early days of this update – it will take some more time to reach final conclusions:

1. News and timely content will get more visibility. Now, besides showing up in Universal Search and on the News tab, the first page will be dominated by news content. Keep in mind though, that this will not be for general searches but for searches very specific to trending and hot topics.   Therefore, a shift to more of a news focus should result in additional traffic due to this increased visibility.   It still remains to be seen if, new, fresh content will be able to retain strong rankings over time or whether they will fall back.  
2.  Fresh content is getting more visibility for general searches too – this includes updated content as opposed to just new pages.  The impact is more subtle and does not seem to dominate the SERP like the first case. 

   How much content needs to change before it is viewed as fresh? Too soon to tell but you can bet that Google has put in checks to try and stop the obvious gaming that is probably already in motion (such as randomizing sidebar text snippets or swapping out lists of links). I am not saying these techniques won’t work or you shouldn’t try them … just that if they do work they might not be sustainable as opposed to legitimate fresh content.

3. Google is likely categorizing searches to determine whether freshness is important to the quality of the page. A website strategy should focus on freshening content on pages that need fresh content to improve the page (such as review pages).
4. I am sure Google will be tweaking this. The PageRank Algorithm did a decent job of making sure that worthy pages made it to the top, even if not timely. I have seen some poor results making it into the search results strictly on freshness.
5. In general, the impact of this update does not seem nearly as severe as Panda – however this one might whittle away at your traffic over time if the fresh content tends to slowly take up positions.

6. This does not mean that site authority and linking programs are not as important as ever. In most case, sites showing up all have strong linking profiles.

In all, a very interesting update form Google and one I am sure they are not finished with.  I'll continue to analyze this over time to see the longer term impact.

In my last post I pointed out what poor and misunderstood metrics bounce rates and time-on-site are, as traditionally reported by Google  Analytics, Omniture and other web analytic tools.   Post-Panda, here is the information  we need  - (to either prevent a Panda hit or as part of a Panda recovery strategy).

• True time on site (time spent on all pages, including bounces and last page).
• Bounce, as I mentioned previously and as currently defined is simply a single page visit.  While a useful metric, it will vary by site in value and also can be skewed by page and site design and technology (for instance, Ajax). 
•  Exit rates - This is far better than bounce.  I want to know what percentage of people exit:
• 15 second exits
• 30 second exits
• 45 second exits
• 60 second exits
  My assumption is that the longer users stay on the site, the more positive the search experience.   
• I'd love to know back buttons exiting users from the site.   A back-button to Google is what I am trying to prevent.  

To count exits, you need to set a timer that triggers a ping event every 15 seconds -  so if you get a visit with no pings, you know that the user left in less than 15 seconds.    This would also let you calculate true time on site.   Of course, the last pageview may never end if the user decided to take a lunch  break, so you would need to set some sort of max time to count for any one page (perhaps 120 seconds).  

This is what we need to understand and track over time our sites performance in response to search.   Let me add one very important thing when analyzing your metrics - you must create a segment that only looks at these statistics for visitors from search and specifically a Google only segment.   What matters is your SEO visits, users coming to your site via other means may have very different metrics. 
Once you get good metrics like these in place, you will have a much better understanding on what pages or parts of your sites might be causing problems  - as well as a better way to measure improvements to your site.

I am going to spend some time looking for off-the-shelf solutions  to help capture these metrics .  I'll let you know what I find in my next post! :)