Tag: google analytic tips

Let Your Customers Tell You What They Want: Paths of Desire
Listen to your customers. Let them tell you what they want… Let them show you how to make your site better.

Navigation is one of the biggest problems we all face when shopping online. It crops up again and again as a pain point in customer surveys. So why not let your customers help you make your navigation better?

This series of articles explains how to use on-site search reports to improve your site and help shoppers find what they want.

You’ll also discover how you can use this information to gain market insight which will be valuable offline as well.

Traditional market researchers would love to get their hands on this kind of data. And it’s already sitting there waiting for you.

The step-by-step instructions are based on Google Analytics, which has some excellent reports for this purpose. But the principles can be applied in any analytics system which allows you to see the search phrases people use and the pages where they use them.

The title of this piece and the hints above are probably enough for you to get going. This technique is so simple and the results are often so clear that you’ll probably learn something if you dive into those reports right now.

But if you’d like to learn a bit more, please read on. This post contains Part One
1. Why Site Search is So Important
2. Part Two: How To Use Site Search Reports in Google Analytics
3. Part Three:
  - Bonus Google Analytics Site Search Tips
  - On-site Search Resources
Why Site Search is So Important

The simplest way of finding out what your customers want is to ask them. And then pay attention to what they tell you. But that’s easier said than done even in a real ‘bricks and mortar’ store.

We’re all familiar with this dead-end retail dialogue:

“Can I help you?”
“No, I’m just looking.”

Online the situation is even more extreme. We don’t get to speak to our customers.

So you should grab any chance you get to let the visitors to your site tell you what they want using their own words.

I’m a great fan of using embedded surveys or feedback systems to listen to the voice of the customer. Give visitors a chance to tell you things and they will be very generous with their suggestions and advice. Any form of usability testing, whether it’s a formal lab-based study or a quick bit of “do it yourself” testing, is another great way of hearing visitors describe what they want to do on your site.

Clickstream reporting tools, like Google Analytics and Yahoo!Web Analytics or Coremetrics, Omniture, Unica etc. are not normally considered any good for analysing visitor intent. They’re great for recording ‘what’ visitors did, but not ‘why’ they did it. But that’s not completely true, as I will explain.

Search is the Language of Intent

There are two places where analytics tools allow us a rare and valuable glimpse into the intentions of the visitor — the elusive ‘why’:
- the external search keywords reports
- the onsite search terms reports
In the case of the of the external paid search you need to be sure that the report you are looking at is one which is showing the actual phrase typed by the visitor, and not just the keywords which were matched. But provided that is the case, the words you see there are ones your visitors chose.

It’s as if that bricks and mortar dialogue had been:

“Can I help you?”
“Yes, please: I’m looking for…”

That’s much better. Now we’re talking. We can do a lot with that information. For example, plenty of work has been done on ways of matching intentions, as expressed in external search phrases, with the content on the site’s landing pages in order to reduce bounce rates and smooth the path to the checkout. But that’s not what this article is about.

Internal Search is the Voice of Frustrated Intent

We worry a lot about getting people to our sites and perhaps not enough about helping them once they’ve arrived. So the internal search terms reports can easily be overlooked.

There’s no pay per click budget at stake to focus attention on internal search. Meanwhile the volume of data is also smaller: typically less than 20% of visitors will use the internal site search. Most of us tend to use search to get to sites, but want to point and click when we get there.

So navigation and information architecture seem vital matters to site-owners, while internal search is left on the sidelines. As a result, I suspect that more attention is paid to the path-analysis and site overlay reports in our web analytics tools compared to the internal site search reports. This is a mistake.

Path-analysis is a time-bandit which can consume large chunks of your life for little gain. Often the more you learn, the less you know. On the other hand, internal search reports can often provide instant insights and even indicate what action to take. There are not many analytics reports which give such specific details about what you need to do.

Internal Search Terms Tell You What Visitors Want — and Where to Put It

Two wise men have encouraged me to treat internal search reports as a gold mine: Jared Spool and my Father-in-Law.

My Father-in-Law’s contribution was a story from the days when he was doing his UK National Service in the RAF. He told me about the RAF approach to planning a new camp. He explained that they would put up all the buildings first, but lay down very few paths between them. They would then wait to see what happened.

As people went about their daily routines in the camp, natural paths would start to emerge. As the paths wore down the grass, it would become clear which were the important routes. At that stage it was then possible to provide paved paths which matched the needs of the people using the place.

You’ll hear different versions of this story from all over the world. Another example involves the building of a university campus. In fact there are books devoted to the subject:

The phrase “desire paths” was coined by Gaston Bachelard in his book ‘The Poetics of Space’ to describe the paths people wear into the ground along the routes we actually want to use. We can see them all around us. There’s even a Desire Path Flickr group full of examples.

There are desire paths on our web sites too. But how do we spot them? Not with ‘path analysis’, despite the name.

This is where Jared Spool comes to the rescue. One of his standard pieces of advice is to use your internal search reports to look for clues as to where people wanted to go. In particular, look at the page where the search began. This is your visitor saying to you: “I got to this page, looking for ‘keyword’, and at this point I gave up using the navigation and tried search instead.”

Once enough data has built up on your site you can often spot some patterns, like the emerging paths in those newly-built RAF camps.

Seize those search terms. Treat them as the golden nuggets they are. This is the voice of your customers.

Examine your Site Search Terms in Detail
- Look closely at the pages where the search began and possibly the page most often preceding it (OK: path analysis has its uses, I concede)
- See if the keyword deserves a place on that page, or in the navigation. Your visitors think it does
- Put it there if you can
- Help visitors turn into customers
Coming Up: So How Do You Tap Into These Golden Insights?

In the next part of the series you’ll learn how to put Google Analytics Site Search Reports to work for you. Let those reports show you what your visitors want, the language they use to describe it and where they’re having problems finding it. I’ll explain how to use to this information to improve the performance of your site, increase customer satisfaction and gain valuable market insight.

After that I’ll give you a bonus tip which will allow you to look at those preceding pages without using the path analysis reports!
May 25, 2011
Why you need at least 2 views in Google Analytics

The first thing any GA implementation should include is two views: a working view and a ‘raw’ unaltered backup view. One day that raw view will save you from a data disaster.

The ‘raw’ unprocessed view is one in which the data has not been modified in any way when configuring the view settings or by adding filters.

You should always have such a set of data available so that you can sanity check anything else you do against a pure source. And you will also have a backup in case anything you do in your working views messes up the data.

Updated: February 2014 Google have changed their terminology. What used to be called ‘profiles’ are now called ‘views’
Updated: September 2012 Make that 3: why you need a spare GA view
June 2012 How to use ‘Copy View’ in Google Analytics v5

How to Set Up a Raw View

When setting up a raw view you should enable ecommerce and configure search where relevant (but do not strip the search parameters). Doing this will just enable more reporting, no data will be changed.

You should not configure the view to ignore any parameters and you should not configure the default page. Doing either of those would modify the data.

Off the top of my head I cannot think of any filters that you would dare to apply to this view. You don’t want to filter anything out, and you don’t want to use filters to change data.

So definitely NO excluding of IP addresses, or changing all the URIs to lowercase. This raw view is precisely the one you would want to cross-check against when debugging such filters.

What to Do if You Only Have One View

If you only have one view set up for the site at the moment, check whether any filters have been applied, or if the default page or any parameters to ignore have been configured. If none of the above have been done, then this view should become your raw view. It will contain historical data in the raw form, so it makes sense to keep this valuable resource intact and create new working views instead.

If this is the case, I recommend editing the name of the view to make it clear what it is. It’s probably called something like ‘www.mydomain.com’, which will sort nicely to the bottom of most lists. So I would just add a function and warning to the end: ‘www.mydomain.com (raw data – do NOT change)’

If you haven’t got a view like this already, you need to set one up as soon as possible. I recommend sticking the date of the next day at the end of the name when configuring this view so that you can see at a glance when the data begins.

Update, June 2012: You can now use ‘Copy View’ to quickly create a clone of an existing view and then edit the settings.

Setting up Goals in a Raw View

Although you should not have any filters applied to your raw view, it’s possible to configure goals. In the days when GA was limited to four goals per view I used to use this view for housekeeping and debugging goals, such as the goal for the 404 page, so as not to us up slots in the main working views.

I’d recommend having at least that goal configured here. The ‘Reverse Goal Path’ report will then have a full, un-modified set of data to work with when you need it to go looking for internal sources of bad links. But I would also configure such a goal in one of the main working views these days and configure an Intelligence alert for its conversion rate. This is data is something you need to have in front of you, not hidden in a view you seldom visit.

How the Raw View Can Help You or Save The Day

For most purpose the raw view is one that sits in the background collecting all the data. It’s not a view most people would want to see.

You’ll need the raw when configuring filters which do things like include or exclude visits of some kind. By comparing the new filtered figures with the un-filtered version you will be able to sanity check whether the difference is credible or whether you need to re-check your filters.

You’ll also find the raw view useful if you find a situation in which a parameter you have stripped out by using the ‘ignore parameter’ setting suddenly turns out to have a use. For example, you would normally strip session ids out of a working view. But it’s conceivable that at a later date you might want to be able to find all the sessions which encountered some error page and create a segment in order to track what other aspects those visits have in common.

Finally, you’ll also be very grateful for the raw view if you ever make a mistake with a filter and discover days later that it has trashed some relevant data. You won’t be able to undo the damage in the working view, but at least you’ll have the raw data to fill in the gaps in an external report…

Updates

September 2012: Why you should also have a spare view in Google Analytics
Two views is the minimum you should have for each GA web property. But it’s much better to have a third view available for the future.

I’ve just been reminded of this by an experience with a client who has just moved their ecommerce site to a new platform. They have built up many years of historical data in the existing views. So there is a big benefit in continuing to use the same view in order to preserve the continuity and make easy comparisons possible.

BUT… the new site has a very different structure and uses different URIs. The existing goals need to be replaced by new equivalents.

Most of the existing goal slots have been used up in the main working views. We need a new views with a fresh set of 20 slots. But a new view will have no historical data.

Fortunately the client already has another spare view which has been storing data for several years. It has a limited set of data-hygiene filters in place to create a ‘clean’ set of data so this is not a ‘raw’ view. But all 20 goal slots are empty.

This approach means that although historical comparisons of things like checkout abandon rates will need to be done externally, most of the key comparisons of ecommerce conversion rates and so on will all be fine.

Of course, as soon as this view becomes the new working view, we must remember to set up another spare for next time….

June 2012: How to copy and duplicate views in Google Analytics
The easiest way to create a new view in GA is to copy an existing view.
1. Go to the ‘view settings’ tab in the Admin area:

2. There’s a ‘Copy view’ button down at the bottom of the screen, to the right of the ‘Apply’ button:

3. Choose a name for your new view and click the ‘Copy profile’ button:

4. Then click back over to the ‘View Settings’ tab and make your changes.
In the case of a ‘raw’ view, remove any ‘default page’, ‘Exclude URL Query Parameters’, and uncheck ‘Strip category parameters out of URL’ if you have it configured in the site search section. Also check to make sure that no filters are being applied to the raw profile.

The view’s goals will also be copied into the new view. This may be exactly what you want. But if you will need to use those goal slots for something else, you will need to try to stop these recording any data. At the moment there does not seem to be any way of deleting the goals. So the only solution I know of at the moment is to set the goal to ‘inactive’:

August 6, 2010
Setting up more profiles in Google Analytics
Why would you want to set up more than one profile for each website in Google Analytics?
1. There are things which GA advanced segments cannot do
2. You should always have two sets of data
One of the classic uses of profile in the early days of Google Analytics was to look at chunks of traffic. Segmentation. So when Advanced Custom Segment appeared the days of using profiles to look at the behaviour of different types of visitor seemed over.

Advanced segments have many advantages over profiles for this purpose.

You can see the different segments alongside each other in the same report in order to make comparisons. This is particularly useful when looking at charts.

You’ll find this a great way of establishing which type of traffic was responsible for a change which you have noticed in the overall numbers.

Techniques like this allow you to see things at a glance in an illuminating way. Try setting up a segment for ‘medium contains email’ and then comparing it with the default ‘search’ segment (or, better still, a ‘medium=cpc|ppc|organic and keywords=your brand terms’ one). You may well find that surges in email traffic are also matched by simultaneous rises in navigational search visits. This would suggest that your emails are reminding people to visit the site, via search, even if they’re not clicking on the email itself. Those emails are worth more to you than the basic clicks suggest.

Custom Advanced Segments are also much better than the old ‘profiles’ approach, because you can think up new ones on the spur of the moment and they will be applied to all your historic data.

I used to love the old Clicktracks system because it allowed me to do something similar. If I suddenly thought “I wonder if…” and wanted to delve into the data, I could set up a set of rules-based conditions and use them to segment all my data. Except in those days you could go for a coffee, or maybe lunch, while the data was processed. GA does it so fast some people even use the ‘test’ feature in segment configuration to extract a set of numbers and never bother to save the segment.

But…

There are some things which GA advanced segments can’t do

The most limiting of these is that you cannot segment funnel reports. So you cannot use segments to see how different types of visitor behave in your checkout or your other micro-conversion funnels.

This is a significant loss.

Common sense says that an existing customer returning because of an email about a very strong offer is going to have a very different reaction to your shopping cart page from someone landing from a generic search and who has never heard of you, let alone already been through your checkout. So how would you set about optimising the shopping cart page for that last group when their data is mixed in with all the others?

Segmented profiles also mean that you could give different people access to only the data which is relevant to their work. For example only the email traffic, or only the unpaid search traffic. Even if the ‘different people’ is actually just you with a different hat on, it can be useful to have this task-specific view of the data without all the other noise.

So, although Advanced Segments are one of the greatest features in Google Analytics, I think every site should still set up a series of profiles, rather than just rely on the basic two.

Two?

Why you need at least two Google Analytics profiles

Yes, two. You should always have one set of raw, unmodified data being captured alongside your main working profile so that you have something to cross-check against. Or fall back to.

But I’ll go into more detail about what you should and shouldn’t do with your raw profile next time in a post explaining how to set up a raw profile in Google Analytics.
July 19, 2010
Simple way to spot landing pages and keywords to fix
Knowing where to start making improvements to your site can be difficult. But here’s a very simple technique which will make the trouble-spots stand out. You can use this trick in loads of Google Analytics reports. It will work for keywords, for landing pages, for referring sites. The technique is quick, easy and universal.

Updates
1. Using Google Analytics Weighted Sort [August 2010]
2. Screenshot illustration of v5 Weighted Sort [September 2011]
The golden metric is the bounce rate, of course. But the trick is to filter the data appearing in each report to focus on the trouble spots and the hot spots. Then draw your lessons and make your changes.

The secret is the “advanced filter” link which appears next to the inline filter at the bottom of your reports.

This simple text link opens up a powerful (almost hidden) form for filtering reports in multiple ways. It’s the kind of tiny feature which can be every bit as useful in real life as some of the big new ‘bells and whistles’ which Google make more fuss about.

You can probably figure out uses for this without me having to explain. This tip makes it much easier to narrow down a list of pages (or keywords, or whatever) in order to spot the problem areas and take action. Once you start using this you’ll find it will come in handy again and again.

Please give it a try, it’s really simple and it often makes the insight almost jump out at you. The more I use this feature, the more I like it.

If you’d like an example, please read on.

The advanced inline filters interface is hidden behind a text link, so you might not notice it at first. But you may already be using the original inline filter to narrow down keywords, for example, to ones which contain or exclude brand terms. That technique is also good for grouping types of page, such as all the product pages or all the category pages on an ecommerce site. I would typically use that approach when looking at the product pages vs the category pages as landing pages.

In the case of a set of category landing pages you might then click on the bounce rate column in order to sort them in descending order to spot the worst examples. But the snag with this is that the worst cases may (we hope!) turn out to be odd pages which have attracted a tiny level of traffic which happened to bounce. They’re not a priority.

This is where the advanced inline filter comes in handy. You need to set a bottom threshold to the report. Just add an extra condition to your filter so that you only see the landing pages where a significant number of people entered the site:

Suddenly the report comes into focus and you can see the pages which matter: just the ones with significant traffic, with the ones with the highest bounce rate at the top. Details of the advanced filter can be seen at the top of the report, while the ‘edit’ and ‘clear’ filter links are at the bottom.

[Update: but please check the links in the resources section at the bottom of this post to see why Google Analytics ‘Weighted Sort’ fetaure, introduced after this post was written, provides a new and better way of doing this for most cases.]

When you’re doing this kind of analysis, it’s also good to switch to the comparison view of the data.

The comparison view makes it easy to judge the relative scale of the issues. This is one of the most useful views of data in Google Analytics. As with advanced inline filters, the interface does not draw much attention to a very powerful and simple tool. This is one of my favourite ways of letting GA help me spot where the problems lie. Just give that little icon up at the top right of the table a click and see for yourself!

In the example above you can see that most of the bounce rate bars are green and to the right = good. This shows that most of the category pages have a lower bounce rate than the site average.

In this case the data is from a blog, not an ecommerce site, but on a shopping site it is often the case that the category pages and the home page will both have lower than average bounce rates. Both are ‘routing’ pages: their function is to present options for further navigation and to route people to other pages. So ideally both should be better than average in earning at least one exploratory click.

But two of those bars are red and to the left. So why are people so much more likely to bounce from those category pages? The next step is to look at the pages themselves and the sources of the visitors landing on them to see if there’s an explanation.

In this example the really bad category page on this blog about skateboard history is a category called ‘skateboard competitions’. And the visitors are coming from search. So it’s not difficult to guess what is going on. These are almost certainly young skateboarders hoping to enter a skateboard competition of some kind. When they land on a page containing information about events which took place before they were born, they back out of there fast.

You might find something similar on an ecommerce site: visitors landing on a category page after searching for something which seems to match in terms of pure search phrases, but is actually completely irrelevant. For example searches for red ballet shoes landing on your red shoes category page. But you sell trainers. If those are paid search clicks, go running to your negative keywords list. If it’s organic search, take a close look at your Title field and Meta Description field: can you make it more obvious what kind of shoes you sell?

I’ve shown a landing page report in this example, but this approach works very well with keywords. In fact these functions are available throughout Google Analytics. The principle is the same:
- Narrow the rows of data down so that you’re only looking at the traffic which matters and not drowning in numbers
- Sort, or use the comparison view, so that you see what’s bad and what’s good
- Examine the pages and the traffic source data to see if you can work out what’s happening
- Then make changes
That last bit’s harder than clicking buttons in an analytics tool of course. But that’s where the fun and the satisfaction really is.

Acknowledgments: of course Avinash Kaushik was a strong influence on this! In this particular case I think it was a presentation at SES London. But if not there, somewhere else. And he’s certainly blogged about other great examples of using Advanced Filters.

Update: August 2010 — Google Analytics Weighted Sort Provides a Better Way

At the end of August 2010 Google updated the GA reporting interface to provide a built-in automated way of eliminating those insignificant visits and bringing what matters into prominence.
- Read the official post on Weighted Sort in Google Analytics for full details.
- Even better, read Avinash Kaushik’s very compelling exploration of weighted sort which explains how you could use it, why you should and even some insight into how it works!
March 6, 2010