On August 12, Google announced that their high-quality sites algorithm, otherwise known as "Panda", had been rolled out for all languages save for Korean, Chinese, and Japanese. The change is said to impact 6-9% of users' queries, down from the 12% seen in the initial Panda update back in February. Though the official announcement post doesn't make mention of this update's effect on English queries, Vanessa Fox at Search Engine Land reported that a few minor changes were made, but there shouldn't be a substantial impact.
Only 9 hours earlier, the Google Analytics blog announced a change to the way visits are to be calculated, effective immediately. We'll get into just how this changed in a bit, but according to the announcement post, "most users will see less than a 1% change".
So, with all of of my clients being US based, I wasn't expecting to see much of a change from Panda, and I can deal with a 1% change in Google Analytics. However, apparently two insubstantial impacts make a big one, because upon checking Analytics on Monday night, I was surprised to see this:
Organic traffic is up 30% week over week
This particular client saw a 20% drop from the initial Panda update back in February, and we've been working to get back to previous levels ever since. Was this the recovery we'd been hoping for?! After all, the site in question hardly fit the mold of the typical 'Panda-lized' site. Though we were told not to expect much change to English SERPs, I was hopeful.
Google's decision to push Panda 2.4 and the Google Analytics update on the same day wreaked havoc on my ability to see what was really going on. I can only imagine that some of the first-time Panda sites using Google Analytics are reeling right now.
Google pretty frequently points out that many of their teams do not share information intentionally. As an example, the search team has stated time and again that sites that run AdSense advertisements do not receive preferential treatment in the SERPs, despite the fact that this would positively affect Google's bottom line. Similarly, Google's other web properties like Maps, Places and Knol (purportedly) aren't given any special treatment, either. Perhaps this is a similar case, but it's borderline irresponsible for Google to have pushed these two updates simultaneously. I believe the onus falls more on the Analytics team, but it's hard to know really.
Seeing Through Pandalytics 1.0
In trying to get to the bottom of this issue, it's important to understand how visit calculation in Google Analytics had changed. Straight from the announcement blog post:
What’s changing?
Currently, Google Analytics ends a session when:
- More than 30 minutes have elapsed between pageviews for a single visitor.
- At the end of a day.
- When a visitor closes their browser.
If any of these events occur, then the next pageview from the visitor will start a new session.
In the new model, Google Analytics will end a session when:
- More than 30 minutes have elapsed between pageviews for a single visitor.
- At the end of a day.
- When any traffic source value for the user changes. Traffic source information includes: utm_source, utm_medium, utm_term, utm_content, utm_id, utm_campaign, and gelid.
Ultimately, this change is about assigning proper attribution for conversions and engagement. As Michael Whitaker points out in his blog post, previous to August 11, it was possible to find plenty of keywords with pageviews and unique visitor counts but 0 visits. Grab the custom report from his post to see for yourself, or take a look here (with filter applied so that visits = 0):
Now, each new keyword is going to count as a visit, which is really the right move. See this same report set to a date after the change:
Some Examples
So what is actually going on here? Well, here are a few real-world scenarios where the way that visits is calculated is changing.
Scenario 1:
- User searches Google for "Product Name" and clicks on your AdWords advertisement.
- User leaves site and searches a few more times, click on competition and comparing prices and features.
- User ultimately decides to with your product, Googles "Your Brand + Product Name", clicks your organic listing, and buys the product. This whole process takes less than 30 minutes.
Previously, the second visit to your site would still count towards the original query, "Product Name". The conversion is attributed to the most recent non-direct source, so "Your Brand + Product Name" gets the credit, but would not appear in your organic keyword report (or would with 0 visits attached). Now, this counts as 2 separate sessions, and "Your Brand + Product Name" will appear in your organic keyword report with 1 visit.
Scenario 2:
- User searches "Product Name" and lands on your site.
- User exits and visits a few other sites.
- User searches "Slightly Refined Product Name", lands on your site, and buys.
Again, now this counts as two visits, where it used to be one. In fact, for this particular client, I believe this scenario was pretty common, as average query length increased significantly, suggesting users were refining their queries.
So Is My Traffic Up Or What!?
Still, in my example above organic visits were up over 30%. This is quite a bit more than the expected 1% change from Google Analytics, and the "insubstantial impact" from Panda 2.4. How can I know if there is any Panda recovery at play? If I want to compare apples to apples, the answer is going to have to come from my visit-agnostic numbers: Pageviews and Unique Visitors.
Visits up 30%
Unique Visitors only slightly up
.png)
Pageviews also only slightly up
So it is pretty obvious from the images above that while traffic is on the rise, it's not quite up 30%. It does remain pretty difficult to tell if there was any sort of Panda impact at all, or just a natural growth from some recent link building successes.
Google Pushes an Update to Analytics
Another factor at play in some of this data is that Google acknowledged a bug in the original rollout of this change, and updated their announcement blog post on the following Tuesday:
We identified an issue responsible for unexpected traffic changes following our recent update to how sessions are defined in Google Analytics. A fix was released at 2pm PST Tuesday August 16th. The issue affected some sites using the following configurations:
- If a user comes to a customer’s site with a space in some part of their traffic source data, then revisit the same landing page during that session by refreshing the page or later pressing the back button, a new session will be created for every hit to that page. (Clicking a link elsewhere on the site that leads back to the page should not matter.)
- Google Analytics implementations using multiple trackers (an unsupported configuration) are also affected when a space is included in the traffic source data. These sites will see fewer visits from new visitors, and more visits from returning visitors (with some variation due to different implementations).
Taking a look again at the visits report above, this bug obviously affected the site in question, as visits after Tuesday dropped considerably. Still, the overall effect here is a change significantly higher than a 1% increase.
Again, non-English sites using Google Analytics that are seeing Panda for the first time may be in for a bit of a headache. I'm hoping I was able to shed some light on this problem.
Hi Mike, thanks for the post! This combination of changes along with the bug with analytics had caused "a bit" of headache in a lot of SEOs :/.
There is not a lot of information on how Google Analytics really works and the impact of the updated they made, so I wanted to leave here a post I found last week that explains with a lot of detail the impact of the changes: https://www.analytics-ninja.com/blog/2011/08/how-google-analytics-calculates-visits.html. Hope it's useful!
I think that relying on Google Analytics alone is extremely dangerous for any SEO. Piwik and Open Web Analytics are both excellent, and free, and also have a few very useful features which GA doesn't.
Totally agree. In fact, relying on one tool only is dangeours, so everybody should be using two tools so, in cases like this, can compare results and don't go crazy :D
Hi Mike!
Very good post indeed, that I pair with "Beware: Google Analytics May Be Lying To You" by Brett Snyder in the SEER Interactive blog and this "La nueva medición de sesiones de Google Analytics, a fondo" (A deep review of the new sessions' measurement by Google Analytics) by Xavier Colomer (use Google Translate, as it is in Spanish).
As an European (more: as an Italian and living in Spain), Google couldn't choose a worst moment to launch Panda in the not english Googles: simply it obliged us SEOs to say bye bye to our summer holidays!
Apart this anecdote, the fact that August has normally a very low volume of searches plus the change in counting the sessions in Google Analytics (with several bugs, especially if you were still using the old interface), made very hard to understand if a site was hit by Panda or not. Surely, on big scales that was something very easy to understand (see the cases of Wikio, Kelkoo, Ciao and other websites, especially price comparison ones), but for niche sites that was harder to understand just checking the Analytics data.
That is why, IMO, is always better to rely at least with another analytical tool, as it could be Clicky, in order to double check - as much as it is possible - the data.
"That is why, IMO, is always better to rely at least with another analytical tool, as it could be Clicky, in order to double check - as much as it is possible - the data."
And it is even better no to rely on the metric 'visits' while computing outcomes, conversion rates etc. Not every visit leads to conversion. Not every visit leads to sales. Then why we still calculate Goal conversion rate as conversions/visits and not conversions/visitors. Why we still calculate ecommerce conversion rate as no. of ecommerce transactions/visits and not as ecommerce transactions/visitors. After this recent update a single visitor can generate even more visits by accessing your website via different traffic sources. Which means if your conversions calculations revolves around 'visits' then expect a drop in conversion rate of your macro and micro goals and overall website. Google analytics is already flawed in the way it does outcome analysis and this new update will just make it worse. Now is the perfect time to calculate conversion rate the right way.
Totally right :D
Well since Google messed with our seo, my shop has disappeared from the internet landscape. I have put in all my keywords and I don't come up. I don't even show up in any images. This is bad. I have not had a sale since it was launched. It is as if I don't exist. I may as well close up. Google did us no favors. In all their algorithyms... they messed up the small business persons. We are lost in the wasteland of the internet. This is terrible.
The GA tracking update (which included the bug) was on the 11th and the bug fix happened on the 16th, so any Visitors data from the 11th-16th is essentially useless for comparisons.
Our Visitors data jumped by significantly more than 1% on those days and returned to normal soon after. Our Visits data was normal before, during, and after the time period, but Google's metrics such as Average Time on Site, Bounce Rate, Average Pageviews per Visit, etc, are calculated based on the Visitors, not Visits, and our statistics for last week were essentially ruined for historical comparison.
It is a free service, so I suppose I can't complain too much, but it's still disappointing coming from Google. If your site was significantly impacted by the five-day bug I recommend adding annotations to the start and end dates so you can remind yourself a month or a year from now when you're comparing strange stats.
Absolutely agree.
Google Analytics has been making serious strides toward offering some of the functionality previously only available in paid packages. This, however, is definitely a setback.
Also agree with making annotations. 6 months down the line this might be end up slipping the web analyst's mind.
This article makes me a bit itchy. Panda has been rolled out here in Germany only last week and so far, my sites actually went up in performance. Seeing that Google US is running 2.4 and that Rand mentioned in an ealier WBF that the criteria are likely to become stricter.
Well, let's see what the future holds and react accordingly. For now, I have told my people not to relax simply because things are going well now. The next update is already in the making and who knows what's going to happen then...
And now I keep on itching.
Very informative article. Brings some excellent points to light. Thank you.
Awesome Post and some great comments being listed!!!
Hey Mike, thanks for this post.
We've seen 13% increase in visits and 1% increase in pageviews, so it looks like we are seeing a definate skew from this change... great to understand why it's happening!
Unfortunately Google will not correct data with retrospective effect.
My advice is to add a note in Google Analytics saying “corrupt data from 11th august and 7 days ahead”.
One of my sites increased 40% in organic traffic. This site has adsense on it, and I soon realized something was wrong when there was no increase in earnings :(
I searches many times for their actual features what about but today I got from seomoz. Reallly I am fully confident to clear all the doubt about panda 1.0
Thanks a lot..:)
Hey Mike Thanks for this post!
I am almost clear wit this issue but I am not really happy with that… I would say I am following a very good approach of not relying only on Google Analytics and using other software as well
This helps me double checking the issues that Google Analytics is going through and also any significant change (negative or positive) in Google Analytics can be tracked easily!
is it possible that Google rolling them out simultaneously is not really a "coincidence"?
The last update (English ones - US & intl') did get complaints from many folks on almost every major webmaster forum.
That solves the mystery of my Google Analytics data gyrations over the last few days. Thanks for the posting.
Excellent Post! Where did the statistics about the 6-9% change come from?
Edit from MikeCP: Accidentally edited. Original comment still above.
That number came direct from Google's announcement post:
https://googlewebmastercentral.blogspot.com/2011/08/high-quality-sites-algorithm-launched.html
I think this change in attribution is a good thing from GA, I must say I'm rather excited about when they sort out Conversion Paths so I can say farewell to last-click attribution...
I agree with gfiorelli1 it is hard to rely 100% on any data from GA, most clients I work with do not even use GA they are on Omniture.
But in the sense to see a real value from this test I feel you need to see a bit more data too for a longer period of time. Also did the client run any thing on the day where the roll out of panda took place, could have been a sneeky EDM or some display activity.
There's not just a space bug. If a client has their analytics set up wrong (using utm_source on internal site links) then those will add extra visits too. I explain it here: https://www.dotcult.com/google-analytics-changes-session-definitions
I am curious to know if this has rolled out into the UK. My opinion is that it has since I have conducted backlinking campaign for specific keywords but have not seen a climb for these keywords at all but have seen an increase in traffic.
I will confirm this now to see if there has been an increase in conversion.
Thanks for the post
that is some serious detail, thank you. hopefully the minor change will improve a couple of my sites.
Thanks for this Great post Mike.
Is this means that if a user comes to a site by searching 2 different keywords, within the 30 minutes, it will be counted as 2 visits.
If the user comes from 2 different traffic sources, would it still count as 2 visits ?
Indeed, that seems to be case. Keywords are stored in the utm_term parameter and source is stored in utm_source. As the announcement said:
When any traffic source value for the user changes. Traffic source information includes: utm_source, utm_medium, utm_term, utm_content, utm_id, utm_campaign, and gelid.
This would be two separate sessions.
mmmmm, thought something was a miss alright. I saw a big increase in traffic (apparently) , but no increase in conversions, also webmaster tools showed no increases/ decreases