Seeing Page URLs and Page Titles in the same report is like the gift that keeps on giving. Not only can you easily track 404 error pages in Google Analytics without having to change the tracking code on your site, but it also makes it very easy to optimize your Title Tags. The Title Tag is probably the most important on-page SEO element, so it’s in your control to change it.

Pull up your Top Content report and use John’s tip and insert the parameter &segkey=request_uri|page_title into the URL of your browser. The report should look similar to this one, with Page Title in the secondary dimension dropdown:

Next, we can use the Advanced Filter in powerful ways:

1) Find missing Title Tags

Look for Page Titles containing “not set“:

2) Find short Title Tags

A Google search result will display about 70 characters of the Title Tag. You can use a regular expression (regex) in the Advanced Filter to find title tags based on number of characters. If your title tag is very short you are wasting a valuable opportunity to describe your page better.

Here is a Regex you can use:

^.{0,69}$

This will return results that are fewer than 70 characters. An example of fewer than 30 would be:

^.{0,30}$

3) Find Title tags that are too long

Conversely, you may want to shorten title tags that are too long. Here is the regex:

^.{70,}$

This means more than 70 characters. If you have more than seventy characters in your title, it will be truncated in the Google search results.

Of course you’ll want to prioritize what pages to fix, but these three filters should give you plenty of ideas to write better title tags.

Great tip by John at Lunametrics on how to show Page URL and Page Title side-by-side in a Google Analytics Contents report.

Here is a little background. Out of the box in Google Analytics, you can see Page URLs in the Top Content report and Page Titles in the Content by Title report:

but you can’t directly see both the Page Title and the Page URL:

Again, hop on over to John’s post on how to do this, but here is one practical example why this is so useful:

Tracking 404 error pages on your site, but without having to change the tracking code on your site (Occam’s Razor anyone?).

Typically – and this is the case of a Yahoo! Store if it has been set up properly to track 404 pages – the Title of the 404 page stays the same “Page not Found…”, but you don’t know the actual Page URL requested. In other words, it’s easy to find those 404 error pages via the Content by Title report, but not the Page URL needed to fix them. That’s why the recommended way is to send a virtual pageview on the 404 error page to make it easier to find in your reports.

But now that you can see Page Title and Page URL at the same time, it’s easy to do and no code changes required:

Just use the Advanced Filter at the bottom of the table and search for “Page no Found” or whatever Title you are using:

Voila! Now you know the actual URLs and you can now of course click on those results and do further segmentation to find and hopefully fix the source of these 404 error pages.

Bonus tip

Too tedious to build this report every time? Just have it emailed to you on a schedule. GA will remember the filters and send you the right data.

Hope this helps and that you don’t have too many 404 error pages on your site!

I still like the Visits to Purchase report in Google Analytics even though it may be sub-optimal.

At first glance you could be led to believe that 77% of transactions occur in a single visit. If  almost 80% of your transactions come from single visits then why bother with attribution management? There are many other things you could be doing before worrying about those 20% multi-visit conversions.

So what’s the issue? If you segment this report by Returning Visitors a different picture emerges:

How can you be a returning visitor with just one visit? Surely you’d have to have visited the site at least twice to be counted as a returning visitor. This great question was addressed by Avinash Kaushik a while ago, and the answer lies in the label in the column header (highlighted in yellow): from the start of last campaign.

A campaign in Google Analytics is basically traffic from any source other than direct type-ins or bookmarks. So mainly traffic from search engines and other sites. But the key point here is last campaign. It means that Google Analytics doesn’t count the visits before the last campaign.

What does this look like in practice? Consider these two visitors:

One visit: Yahoo! Organic (conversion)

5 visits: 1st visit: Google CPC > 2nd visit: Yahoo! Organic > 3rd visit: Bing CPC > 4th visit: twitter.com referral > 5th visit: Yahoo! Organic (conversion)

Since the last campaign for both visitors is Yahoo! Organic, both scenarios would be attributed to 1 Visits in the report, even though only the first visitor actually converted in one visit. There could have been one or twenty prior visits in the second scenario and it would still be in the 1 Visits section. Therefore it’s not really a true count of Visits to Purchase report, nor do we know what the sources of any prior visits are.

Fortunately, it is pretty easy to get a true count of visits report using advanced segments. There is a handy dimension appropriately called Count of Visits that you can use:

This one is for single visits. I also created segments for other visit counts:

If you’d like to grab them directly and apply to your Google Analytics profile, here are the segments 1-timer, 2-timer and 3-timer+

Now apply these segments to the Visits to Purchase report and a more interesting picture emerges:

What we see now is that in fact closer to 50% of transactions truly occur in one visit. Maybe we do have to pay attention to attribution?

But the true power of using these count of visits segments is in other reports, such as your keywords:

What is interesting to see is the distribution of the visit counts for branded and non-branded keywords. As might be expected, branded keywords drive a lot of traffic for repeat visits and have a relatively small percentage in the one visit count segment; for branded keyword 1 it’s only 306/821 = 37%. Contrast that to the non-branded keyword: here 169/205 = 82% of visits are in the 1-timer segment. Very few people would use the non-branded keyword to get back to your site on subsequent visits. Another way to look at this is that non-branded keywords tend to be earlier in the sales cycle than branded keywords.

What you are hoping for is that you make a good enough first impression on the non-branded first visit to drive subsequent visits through branded keywords. Easier said than done, but getting a better understanding of the underlying behavior of your site visitors through web analytics is a step in the right direction. I highly recommend segmenting your data using Count of Visits and you’ll be pleasantly surprised at the number of valuable nuggets you’ll uncover.

Although I tend to be skeptical of best practices – they might be a starting point, but I prefer doing testing to find out what works or not – there are some things that are always better than others. A fast-loading site always beats a slow-loading one,  working links are always better than broken links, and reducing errors in general makes obvious sense.

In the same vein I would say that a short checkout form is better than a longer one. Of course you need a certain amount of information to be able to process and fulfill an order, but you should only ask for the information you absolutely need. Reduce friction and remove any obstacles that make it more difficult for your customers to buy from you.

But can we back up this assertion with web analytics?

I am a big fan of tactical web analytics, such as tracking 404 pages, zero results searches and checkout errors.

This report shows the number and type of checkout errors seen by visitors:

The number one error message customers see is the missing phone number.

In a default Yahoo! Store checkout, the phone number is required:

But do you really need the phone number? If there is a problem with the order, can’t you just email the customer? I think the data shows that your customers don’t think you need their phone number either! And if you do require the phone number, you should at least tell your customers why you need it.

If you do not absolutely need to have the phone number, then how about making the phone number field optional?

Fortunately, this is easy to do in the checkout manager in your Yahoo! Store. Just edit the phone field on the shipping page and uncheck the required field:

Because this is just the data from one store I don’t know if making the phone field optional in the checkout qualifies as a best practice , but at least the data gives you a clear picture of the problem. And in this case it would be easy to fix it.

Now on to the other error messages!

I have had quite a few requests to confirm that you can read custom variables so I thought I’d just show a quick demo. I also recommend you take a look at the video from the Google Analytics team that describes this feature (fast forward to around minute 29:00).

As the name implies _getVisitorCustomVar() only reads custom variables that have a visitor-level scope. You can’t use this approach for reading page-level or session-level custom variables, so the scope should be set to 1 when you set a custom variable. The argument in _getVisitorCustomVar(1); is the slot you use for setting the visitor-level custom variable.

Anyway, here is a simple demo where we set a custom variable on another page. Once the custom variable is set, it can be read and be used to execute code on other pages.

Please click here first to set the custom variable. Then come back to this page and see what happens.

It’s always good to pause and ask yourself if the data you are looking at makes sense. Particularly with the awesome powers of advanced segmentation in Yahoo! Web Analytics and Google Analytics you are likely to come across cases where the data looks weird when you set up custom reports. The likely explanation is that you are mixing together dimensions (in GA parlance) or groups (YWA lingo) that don’t really go together. Or perhaps the data model used by the web analytics tool can explain what you are seeing.

Case in point is the seemingly strange issue that Google Analytics can report more visits and than unique visitors for any given page. If you’d like to follow along, you can get the Unique visitors custom report and apply to your GA account.

and the resulting report would look something like this:

The same custom report in YWA gives this:

How can is be that there are more unique visitors than visits to a page?

Well, it turns out that the answer lies in the data model used by Google Analytics.  The visit is assigned to the first page but not subsequent pages, whereas unique visitors are assigned to each page:

If that is the case I hear you ask, then shouldn’t there be pages that have 0 visits attributed to them? There sure are.  You can just segment this custom report by landing page for example:

and you will get something looking like this:

All these pages have 0 visits because they came after the landing page during the same visit.

The same visit attribution model applies to traffic sources such as keywords as well. Consider the case of a visitor coming to your site via two different sources during the same visit.

For example, the visitor might search on both Bing and Google and comes to your site:

As you can see, bing gets the credit for the sale even though it has 0 visits attributed to it. And Google gets the visit, but not the credit for the sale, because it initiated the visit. How do I know that this is the same visitor you might ask? Well, in this case I was able to track the conversion back to the city, which rather conveniently had only one visit:

Of course in practice, this is not a very scalable way of looking for assist keywords or doing attribution management because you wouldn’t normally be able to tell that this data belongs to the same visitor. One thing you can do though is get a general sense for how often there are assists during the same visit:

Pull up your main Keywords report and use the Advanced Filter at the bottom of the table and enter these conditions:

and look at the results:

What this means is that 2.32% of transactions had an assist during the same visit.

Interesting, even though it is not terribly actionable since we don’t know what those assists are. But at least we can now make sense of  “zero visits”!

A logical progression from segmenting your data in web analytics is to then be able to target those segments. For example, you might want to show a special offer only to loyal repeat customers and not to others.

In a very informative webinar last week by the Google Analytics team on custom variables, one particular example caught my attention: the ability to not only set a custom variable, but also to read its value.

Let’s assume that I set a custom variable called CustomerType on the transaction page that has the possible values New or Repeat, then it sounds like we will be able to do something like this:

<script>var customertype = pageTracker._getVisitorCustomVar(1);
if (customertype = 'Repeat') {
display special offer on the page
}
</script>

The method _getVisitorCustomVar() is described in the GA code site. Cool stuff!

Back in April 2009 Google announced a change to search referrals coming from google.com. The old familiar format when you search on google.com

http://www.google.com/search?hl=en&q=flowers&btnG=Google+Search

would be replaced with:

http://www.google.com/webhp#hl=en&q=flowers

Initially, some people were concerned that the use of the hashtag # in the URL would cause issues with web analytics tracking, but Google found a way to make sure that the referrer data would still be valid (albeit different) when visitors click on a results link. This is what the new referrer string looks like:

http://www.google.com/url?sa=t&source=web&ct=res&cd=1&ved=0CBYQFjAA&url=http%3A%2F%2Fwww.1800flowers.com%2F&rct=j&
q=flowers&ei=9EuxS5XsCI3-tQOJlo2PAQ&usg=AFQjCNHgC4BdrCNZRz6VZnHa2XGbvsBKmA

In December 2009 I estimated the proportion of these new Google organic referrers to be 10%. It looks like that this proportion has now been bumped up to 20% since early February 2010:

So, how is this information actionable? I will admit that I primarily just wanted to make this observation about the referrer change rollout, but one thing you can do with the new referrer format is find out the exact search results ranking.

First of all, what is attribution management? If I may quote John Lovett from his paper “A Framework For Multicampaign Attribution Measurement“:

Multi-campaign attribution is the practice of attributing credit to all marketing exposures that led to a Web site and subsequently resulted in a conversion event, rather than attributing all credit to the exposure immediately preceding the conversion.

In other words, many of your customers are deliberatively making your life harder by visiting your site multiple times from multiple sources before converting and, to make matters even more complicated, it’s really hard to tell that this is actually happening!

Here is an example:

Standard Traffic Sources report, where one transaction is credited to Google Organic and one to Google Adwords.

I can also see the actual transaction:

Fair enough…Google gets the credit, end of story. Or so one might think… Bear in mind that this is just the last, converting visit. We can’t really see what or if the customer came to our site before. That’s where attribution analysis comes in, and here is the salient visit history from our own attribution management tool:

Turns out that the referral from flyertalk.com in spot 7 in my referral list played an important role in the conversion process and should thus get at least some if not all of the credit:

Useful to know, right? And from John’s same paper, “According to 275 Web site decision-makers surveyed in 2008, a full 52% agree that attribution would enable them to spend marketing dollars more effectively.”

OK, but back to my original question: Is attribution management right for smaller online retailers?

It’s quite a step from “useful to know” to actionable advice. The example I showed is not actionable per se. All I know is that there is more to the transaction than first meets the eye. It’s just one multi-visit trail leading to one transaction.

In order for attribution analysis to become interesting, we have to collect far more data to come up with aggregate purchase paths that exhibit distinct differences – kind of like segmenting in web analytics. For example, attribution analysis can tell you that many branded search terms are preceded by keywords that look like they are converting badly in your web analytics, but which provide a crucial assist.

The need for a lot of data could however rule out many smaller online retailers who may not have enough traffic for those patterns to emerge. Furthermore, doing any kind of decent analysis takes time and resources, possibly at the expense of other worthy initiatives, such as web analytics in general, AB and Multivariate testing, email marketing, landing page optimization, SEO and SEM, fulfilling orders, etc…

Has anyone looked into at what point attribution management becomes important? Is it in terms of revenue, traffic, marketing spend, or number of different channels? Could there be a simple metric that a busy retailer could use to determine whether she “needs” attribution management?

Is attribution management just for the big boys?

In April 2009 Google announced a change to the referrer information coming from Google organic searches. The interesting bit to many people was the addition of the cd parameter, which is the actual search results position. There are a few ways you can see the keyword position in your web analytics, such as with filters, custom variables or event tracking. I personally like event tracking in Google Analytics, and before the change in referrer information we were at least able to see the page a keyword was on.

I thought it would be interesting to revisit this topic from a web analytics/optimization perspective.

1) I estimate that about 10% of Google organic searches have the new referrer information. I have not detected a significant change over time.

2) Number #1 ranking gets by far the most clicks, followed by #2 and #3. Positions #4 through #8 seem to get similar clickthroughs, and from position #9 on down traffic drops significantly. Nothing too surprising here and mirrors what we see on e-commerce section pages: visitors tend to click from top to bottom.

A number #1 ranking however doesn’t mean much if it is unqualified traffic that doesn’t convert. From an optimization perspective, how about segmenting your data to show only traffic from top ranking keywords that doesn’t convert?

Then check keyword landing page combos to find such underperforming keywords and their respective landing pages.

Another way to look at it is to say that you have done all you can from an SEO perspective, can’t get higher than #1 ranking. Now it’s up to your site to convert those visitors. Or you are targeting the wrong keywords in your SEO efforts.

Happy Holidays!