I switched to the new Google Analytics interface and almost immediately ran into that old problem of wanting to export more than 500 rows of data without having to resort to using API calls. The old “limit=50000″ trick doesn’t work with the new format, but thankfully there is a work around which I came across on the Convonix blog.
If you choose to show more than the standard 10 rows using the drop down at the bottom of the page, a new “rowcount” variable is added to your URL. For example, I changed a page to display 25 rows and the variable looks like this:
By changing the 25 you can change how many rows get displayed and then export them, up to a 50,000 row limit apparently. I’d caution against relying on this as a long term solution though. The previous 50,000 row limit trick got reduced to 20,000 after so many people started using it, and I imagine the same will happen with this trick once its use catches on. In the meantime though, enjoy!
We all like to keep an eye on the usual metrics when looking at our Google Analytics accounts. Visits, unique visitors, bounce rates, time on site, and conversion rates all get a look in. These are all great pieces of information for making sure that things are working the way you expect them to on your website, but what if you want to look a little deeper?
Unfortunately Analytics can’t answer every question you might have about your site, in which case it’s time to dust off your Excel For Dummies book and get stuck into manipulating the data yourself. For those of you looking for a good guide to some of the most useful Excel functions for SEO analysis I can recommend the Microsoft Excel for SEO guide from Distilled.net.
Digging deeper often requires large amounts of data to give meaningful answers, so you’re going to want to get familiar with adding the “&limit=50000” to your GA URLs, or better still start using the Google Analytics Data Export API or the Excellent Analytics Excel plug-in.
Keyword Lengths and Conversion Rates
I’m a firm believer that the more you know about your visitors and their behaviour the better you can tailor your product to suit their needs. So, from time to time we go and look at some metrics that are slightly off the beaten track. Have a look at the graph below for instance:
It charts the traffic and email enquiry conversion rate of traffic over a recent two week period. The first thing that struck me was the more keywords people use to find WhatClinic.com the more likely they are to convert. The second thing was that just over 50% of our email conversions come from people who use 4 or 5 keywords to find the site.
All well and good you say, but what use is information like this? Well, for a website like ours with a long tail focus it shows us how long the keywords in the long tail are. We typically optimise pages for one or two keywords, usually two or three words in length. The data above suggests that maybe some pages should be optimised for slightly longer keywords, or perhaps even two longer keywords.
Thanks to other curious SEOs like SharkSEO we also know that you can write two completely different meta descriptions for the same page and the search engines will pick the description that best matches the keyword being searched for. This opens up some new possibilities about how to organise our data and our site structure. Using the keyword length and conversion data above we can make more informed decisions about how to optimise the resulting pages.
Are Keywords Getting Longer?
Just over a year ago I wrote about how people were using longer keywords to find WhatClinic.com. Seeing as we’re talking about keyword lengths again I thought I’d take a quick peek at some data from this year. I was in for a surprise.
If my data was to be believed keyword lengths were almost exactly the same as they were a year ago. The answer seemed too neat to me, so I decided to do a little segmentation. My suspicion was that by looking at our traffic as a whole I was missing some underlying trends, and it turns out I was right.
Traffic from Ireland accounts for around 19% of our total visits, but as you can see from the chart above it accounts for over 30% of our one and two word keyword traffic. Again the question is how is this information useful or actionable? The simple answer again is to do with the messaging – the page title and the meta description in particular.
In Google.ie we now rank quite well for certain one word keywords like “braces” or “dentist”. While this is great for us in terms of traffic, the pages are really optimised for people looking for our page about braces in Ireland, or dentists in Ireland. This means that as the keywords used to find these pages get more generic / head / short tail that maybe we should look at changing the messaging on them to better reflect more closely what the user is looking for. For the cases above, I think that the messaging might be OK, but we’ll test some alternatives and see how they affect CTR and conversion rates.
The Importance Of Segmentation
The Irish traffic above really skewed the keyword length data above. Seeing as our website deals with so many geographies and our keyword rankings quite a lot across them, any decisions about site structures and one page optimisation should only be made once the overall site figures have been sliced enough to have confidence in them.
Excluding the Irish traffic, keywords have gotten slightly longer since 2010, but not massively so. It is the relative shortening of Irish keywords that is much more significant to us on this occasion.
We have previously observed similar big differences in user behaviour based on whether the landing takes place on a brochure / listing page or on one of our search results pages. We’ve even observed that the nearer the top of the tree structure a user lands the more likely they are to convert.
It’s often worth digging deeper than the reports or segments in Google Analytics can offer by themselves because the information that comes out can offer you a clearer picture of some of the bigger underlying trends affecting your site and give you the information you need to not only stay ahead of your competitors in the SERPs, but ultimately make your site better for your users.
[Update: May 10th - This warning originally read "This report is based on sampled data". Google have now explained that the warning appears any time you report on more than one dimension in Analytics on a date range containing more than 500,000 visits. See the Google Analytics Fast Access Mode help page for more detail.]
Everyone loves Google Analytics. It’s free, full of great features, and most of the time it works like a charm. We use GA data regularly in WhatClinic.com to make important decisions about everything from internal linking structures to on page keyword priorities, and it really comes into its own when something on the site is broken and we need to track the problem down.
However, like a lot of great free web services GA suffers from being too popular. The amount of data it has to crunch for our website alone is staggering, so imagine the load when you add in the tens if not hundreds of millions of other websites that use it too. Unfortunately the strain starts to show as soon as you really dig into your data.
The Dreaded Fast Access Mode Message
One of my most used features on GA is the Advanced Segment, which lets me filter the traffic I’m reporting on in a myriad of different ways. I can report on only the people who landed on our search pages, or only the people who came from Canada, or pretty much any set of people with an analytics parameter in common.
Google have a problem with this great feature though: in order to give you an answer to your query in a reasonable length of time they often fall back on using sampled data, or “Fast Access Mode” as they now call it, if you are using a date range that contains over 500,000 visits.
Using sampled data is absolutely fine when your data points don’t suffer from a lot of volatility, but if they do then the results can be very unreliable. Take a look at the data set above. It uses sampled data because I’ve created an Advanced Segment with two parameters (user location and page URL) and I’m looking at a period of three and a half months which contains far more than 500,000 visits. Analytics is reporting a conversion rate in April of just 2.65% for the set of visitors I’m looking at.
Now look at the data set below. It is reporting a conversion rate in April of 4.36%. The difference is that it isn’t looking at sampled data because the period of time is much shorter and the visitors number is less than 500,000.
Our conversion data can be very volatile in places because of the niche or long tail nature of what we offer and what I need to report on from time to time. For instance. We might get 1,000 enquiries for dental clinics in Dublin in a week, but if I look at just the visitors who were looking for clinics in Rathmines that figure might drop to just 10, and those ten might have been made up of 3 on Monday, none on Tuesday, 1 on Wednesday, and so on. It’s the same when you look at American visitors to our Mexican dentists pages, or UK visitors to our Turkish cosmetic surgery pages.
Be Sure To Perform A Sanity Check
As with the case above, when something looks like a problem we all tend to look into it and see what’s going on, but when something looks good we tend to let it slide. My advice is any time you see the Fast Access Mode message in GA you should sanity check it by using shorter periods of time. This takes a little time but it will give you a far more reliable picture of what’s going on.
When Advanced Segments in Google Analytics were introduced just over a year ago I was delighted. Finally there was an easy way to slice our website data in more than one dimension at a time. By setting up an advanced segment for just “Organic Visitors from the United States” for instance I could really drill down into what a specific set of users were doing on the site and see how their behaviour differed from the norm.
Every now and again though I used to see some strange results come back for reports I was doing. At the start I put this down to the feature being in Beta, but the anomalies didn’t go away. Having come back to use advanced segments in recent days to dig into some very specific user behaviour I noticed the problem was still there so I decided to find out what was causing it.
As an example, when I did a search (filter) in my top content report for all the URLs containing “/dentists/” I got a figure of 313,000 page views for the period of time I was looking at. However, if I swapped out the filter and used an advanced segment I had set up to do the same thing (i.e. only include data for pages that contained “/dentists/” in the URL) I got a figure of 434,000 page views, and I could see pages that clearly didn’t match the segment I thought I had setup.
At this stage the difference was far too big to ignore, and I assumed that lots of other users would be aware of it too. I was right. The Google Analytics help forum was full of questions about why data sliced with filters wasn’t matching data sliced by advanced segments.
The answer came in one line from a guy called MikeOstrowskiASU. He said simply that “Advanced segments are based on visitor sessions.”
Suddenly it was all clear. The advanced segments when used on the top content report were bringing back all the pages visited by anyone who had visited a page with “/dentists/” in the URL during their session rather than just the pages with “/dentists/” in the URL.
Just to confuse matters, advanced segments on visitor reports do match up with filtering because they are based on user sessions. For instance, using an advanced segment for “Visitors from the United Kingdom” on the visitors report will match up with going into the Map Overlay and clicking on the UK.
So, the moral of the story is be careful how you use advanced segments and what you infer from the data you get back. They might not be doing what you think they are.
Have you had a similar problem with Google Analytics, and have you found any workaround for it?