It’s no secret that Google likes web pages that load quickly. In fact they offer free page speed tools to help you optimise your site and make the bold claim that “Fast and optimized pages lead to higher visitor engagement, retention, and conversions.”
Bounce Rate Increased 15%
Google Analytics was recently updated to change how it calculated when a visitor’s session ended. We were told this should have only a small effect, around 1% on average, on how our visitors were being counted. The change went live on Thursday August 11th. If you look at the graph above you can see on Friday the 12th our reported visits increased by over 40%. This trend continued for nearly a week.
Now a 40% increase in traffic would obviously be welcome, but our unique visitors report told a very different story. Nothing had changed much at all!
So, what was going on? It turns out there were some bugs in the Analytics update that created new sessions for users when they should have had only one. Full details are in the update to the announcement of the original change. It would seem that in our case visitors who clicked the back button in their browser to go back to the landing page they arrived on were being counted multiple times.
Google pushed a fix to this problem on Tuesday the 16th of August and everything seems to be back to normal now, but I’m sure we’re not alone in having spent some time trying to work out what was going on and what changes we’d need to make to be able to compare reports from before and after the change. Thankfully it looks like that’s not an issue anymore. Still, it seems like a pretty big bug to slip through the net for such an important product.
Google’s Instant Preview, which lets you see a preview of the page you’re going to click through to, has been in testing for some time now and was finally rolled out to the world on November 9th. For those of you who haven’t seen it in action before, here’s Google’s own video demonstrating it.
I played around with Instant Preview a little over the last couple of days, and while it looks nice on the page I haven’t found it particularly useful yet. I’ve a feeling it will work well for certain types of search, obviously anything specifically visual related, maybe shopping for product you know already. The level of detail offered in the Instant Preview certainly isn’t sufficent to decide whether or not the content of a given page is useful, so it seems to be trying to promote “good looking” web pages.
Time will tell I guess, but what effect is Instant Preview having on your website right now? For us, you can see from the Analytics screenshot below that over the last couple of days, as more and more people hear about the product, there has been a small spike in what I thought was “weird us traffic”.
The traffic can be identified as having a source of (direct) and a medium of (none) in your own Analytics. It’s normal to expect a certain amount of this traffic from all over the globe that Analytics essentially can’t identify, but only our US (direct) (none) traffic has experienced this particular spike.
Normally a little extra traffic would be more than welcome, but in this case I think Google is doing something wrong. It seems like every person who uses Instant Preview is being counted as a visitor to our site, whether they click through or not. Judging from the 91.92% bounce rate of the traffic in question yesterday, this is not a good thing!
The bounce rates above are only for this (direct) (none) traffic from the US, which wouldn’t be a problem if it were just the usual handful of visitors a day, not now that it’s heading towards several hundred with a bounce rate heading towards 100% it could start to have negative effect on the site overall. We work hard to keep our pages’ load times and bounce rates down, knowing that Google takes both into account when ranking pages.
Hopefully this is just an oversight on Google’s part which will be rectified soon, but in the meantime watch out for “weird us traffic” to your site with an abnormally high bounce rate. Let us know in the comments if you’ve experienced it too.
UPDATE 15 NOVEMBER
The UA we are seeing in our web server logs is 5.0+(en-us)+AppleWebKit/525.13+(KHTML,+like+Gecko;+Google+Web+Preview)+Version/3.1+Safari/525.13
UPDATE 23 NOVEMBER
It looks like Google have gotten around to fixing this problem now. We’ve seen the strange traffic fall off today on our Analytics, and Google have just posted this update to their Analytics blog: http://analytics.blogspot.com/2010/11/instant-preview-issue-resolved.html
We recently ran a test that improved our email enquiry conversion rate by 25%, but we decided not to make the change. Picture this. You spend months or years even building a website, and it works like a charm. You’re getting plenty of visitors and they are taking the actions you designed for them to take. Everything is great, right? Well, probably, but are you sure?
One thing we’ve discovered again and again here in RevaHealth.com is that it’s never worth assuming that you understand what your website’s visitors want when they arrive. In the past when we added new features we looked at a couple of headline metrics to make sure we hadn’t broken everything, and hopefully improved a key metric like conversion rate or reduced another like bounce rate.
Google Analytics & Website Optimiser
Nowadays we try and go into a bit more depth about things, making more subtle improvements to parts of the site that some visitors see and others don’t, and this requires more fine grain measurement than before. A key tool that we’ve been using for years now is Google Analytics. If you’ve not been using it (or something similar) on your site up to now, I can recommend it wholeheartedly. Right now we’re using it primarily for page tracking, but we’re hoping to use it for goal tracking and even event tracking in the near future.
Another free tool from Google we’ve just started using is their Website Optimiser. On a site like ours where we get enough traffic and enough conversions in a day or two to our key pages to give meaningful results it’s a boon. We don’t have to do any of the maths or work out confidence levels or margins or error. It does it all for us. Best of all, we can test multiple changes to a page all at once, and see which combination our visitors use the most.
Testing Two Types Of Conversion
Recently we were discussing the two main actions that people take on our site: contacting a clinic by email and looking up a clinic’s phone number. The former is far more valuable to us a company. Our customers, the clinics, can see the end result of a visitor using the site to email them. When a visitor looks up a phone number it is very unlikely that the clinic will ever know that the visitor found it on our site. The clinics can see the value of one action, and not the other. Even though we tell them how many people find their phone number on our site, it doesn’t really resonate with them.
Knowing how much more valuable an email enquiry was, we decided to test whether taking away the ability to look up a phone number would cause a significant increase in the numbers who contacted a clinic by email. The results were pretty extreme.
25% Up, 100% Down
We tested a over a two day period with only half of our traffic having the ability to find a phone number. The half who didn’t have the ability to find a phone number converted to email enquiries 25% better than before, which would seem like a good result, until you look at the numbers involved.
Over the course of the two days the 50% who couldn’t see phone numbers created an extra 109 email enquiries. In the same period the people who could look up phone numbers did so 1,418 times. To put it another way, only one in thirteen of our visitors who wanted to look up a phone number created an email enquiry instead when that option wasn’t available to them.
In the end we decided that the cost of frustrating so many of our visitors far outweighed the gains we could make from having 25% more email enquiries. Phone number look ups are here to stay because they are what a large portion of our visitors want.
We already knew that a lot of people were looking up clinics’ phone numbers on our site, but testing it in this way gave us a very clear picture that those visitors are very different to the visitors who create an email enquiry, and they need to be treated differently.
Have you discovered unusual or unexpected behaviour from your site’s visitors? Are you giving them all the actions they are looking for? Are you tracking and testing them? Let us know your experiences in the comments below.
In Google Analytics you can only view 500 rows of results from a query at any time. When you try to export the results of the query, unhelpfully it only exports the results that are on your screen at the time, so if you need more than the 500 results on the screen, you have to page through them and export them by hand one page at a time.
Or so I thought.
A little bit of digging around has revealed that if you add the string &limit=50000 to the end of the URL of the page you are looking at, and then export as CSV or TSV you will get up to
50,000 20,000 records in your downloaded report file. [Update: So many people were using this trick that Google limited the output to 20,000 rows.]
Being able to export large volumes of segmented data is of great value to us here in RevaHealth.com, especially with our focus on long tail SEO. Hopefully this tip will save you some time too.
P.S. If you’re using Analytics and haven’t checked out advanced segments yet, you owe it to yourself to do so. They are the answer to innumerable problems that existed in Analytics reporting up to now.