Good Behaviour

I’ve always worked on the basis that doing calculations using mixed metrics ( by which in this case I mean page views, visits/sessions & uniques) whilst doing an analysis is generally not good. There are some exceptions but when looking at conversion I have tended to avoid calculating it based on page views as a percentage of visits/sessions.

There have been many cases when, in my work with clients, I have faced situations in which their conversion figures have consistantly been a shade higher or lower than my own over a given time period. These have in almost all cases been due to their figures being based on actual sales as measured in their sales management systems whilst I have generally only had access to the analytics data. Not an ideal situation as I would always rather work with actual sales data but it hasn’t been a major concern since the data is generally trended over several months and we are able to see where the changes have occurred – crucial insight has still been available.

In one instance I broke my own rule and took it a step further by looking at the conversion rate using page views to the Order Complete page instead of visits and naturally the figure went up. Of course there are several reasons why the Order Complete page may be viewed more than once in a single visit and not all relating to multiple purchases.

In a follow up session I looked in more depth at regional visitor profiling using weighted segmentation described here. It showed a significant difference in conversion (based on visits only) between the UK and “Rest of World” – not surprising since in this particualr instance overseas delivery was not catered for. However there was a small market oversees that was using the site for purchases to be delivered in the UK.

I got back to thinking about the original difference in conversion between their figures and mine and decided to run a comparative analysis of the Order Complete page based on UK segment and “Rest of World”, this time looking at page views per visit on the Order Complete page and trending over the Christmas period. The pattern became much clearer. Whilst visitors from the UK viewed the Order Complete page on average once per session throughout the period, page views per visit on the Order Complete page for the “Rest of World” segment spiked at 2 around Christmas time.

Rationale? The site didn’t have multiple addresses functionality in the checkout process and so overseas visitors looking to do their Christmas shopping had to repeat the process for each purchase.

I would still be wary of doing this kind of thing but in some circumstance it can yield some surprising and interesting results.

Bounce Rate (a.k.a. Single Access Ratio) is much used and well known in web analytics. Described simply as the ability of an entry page to retain a visitor after arrival and drive him/her onto another page or goal within the site, it is easy to understand its importance. As the Bounce Rate goes up the apparent ability of an entry page to retain a visitor and solicit another action leading to another page or goal goes down.

It is so important that in Google’s recent upgrade of its analytics tool, the overall profile was considerably raised throughout the user interface. Given the huge uptake of GA this is only likely to bolster its overall profile as a benchmark metric.

Google’s own notes on the Bounce Rate read as follows:
“Bounce Rate is the percentage of single-page visits (i.e. visits in which the person left your site from the entrance page). Bounce Rate is a measure of visit quality and a high Bounce Rate generally indicates that site entrance (landing) pages are not relevant to your visitors. You can minimise Bounce Rates by tailoring landing pages to each keyword and ad that you run. Landing pages should provide the information and services that were promised in the ad copy”

If the resultant action from an observed high Bounce Rate is to make changes to an entry page then it is probably worth setting some additional context.

Without wanting to reduce the status of Bounce Rate in the performance analysis of any site it is worth asking the question of whether a high Bounce Rate is always such a bad thing. It is precisely because of the growing popularity of the metric that it is worth considering its status.

Below are six points which add a bit of context to the Bounce Rate metric:

1. Networked sites with separate analytics accounts (same tool)
Larger organisations may run a suite of sites or micro-sites which complement each other but are not fully integrated. As a result each site may have its own analytics account. There may also be a roll up account sitting across all sites. In circumstance like this a visitor arriving on one micro-site from PPC search may then move straight onto another if there is relevance and an available link. This may be an entirely desirable outcome but the Bounce Rate report in the analytics account for the original site will show the visit as a single page visit and therefore contribute to a higher Bounce Rate. In this case it is worth looking at the same page in the roll up account to get a better idea of the overall Bounce Rate.

2a.Variations: Home Page, List page, product page
As with many reporting metrics in web analytics, averages can sometime be misleading. The same goes for Bounce Rate. The Bounce Rate will vary considerably between different types of page. Generally it is likely to be lower on the home page and increase the deeper into the site the entry page is. So a category page will have a higher Bounce Rate and a product page Bounce Rate will most likely be higher still. As the pages become more focused and less generalised there is less to encourage a visitor entering to look around. If they don’t see what they want immediately it is easier to go back than forward.

2b.Weight of entry visits
The previous point has greater relevance when linked with the weighting of entry visits. In an extreme example, if 90% of visits enter on a home page which has a Bounce Rate of 10% and the other 10% of visits enter on product pages which have a Bounce Rate of 90% there is clearly little point in diverting resource to address the Bounce Rate issue on the product pages. The overall site Bounce Rate will be low anyway and until the landing strategy changes to make more use of product pages, or better still landing pages, there is little point in diverting resource.

3. News & media sites
Bounce Rate may be viewed very differently when looking at news and other media sites. Peaks in traffic on these sites often occur on week days at around 9am and again at around lunchtime. Visitors may often come in, take a quick look at the home page headlines and leave again. This would be enough to get a quick fix until there is time to read more. As lifestyles become busier and more time pressured this kind of activity is less surprising. On the web this would results in a high Bounce Rate. The solution might not be to encourage the visitor to delve deeper (thereby lowering the Bounce Rate) but in fact to make sure visitors can get exactly what they want on that fleeting visit thereby ensuring they continue to return for their fix. Bounce Rate has reduced status here.

4. Brochureware sites / offline calls to action (web-centric telephone numbers)
Some sites may not have a business goal that can be easily measured online. The desired action may require a telephone call. In this case it is possible that a landing page could supply all the required information for a visitor to take the decision to act and pick up the phone – hopefully to dial a telephone number that is unique to the website or landing page. High Bounce Rate + desired outcome.

5. Traffic source: Banners Vs PPC entering on a landing page
Just as different entry points have different Bounce Rates the same entry point may have different Bounce Rates depending on the sources of referral traffic. One landing page may be acting as a receptor for an advertising campaign which uses both online display (banners etc) as well as pay per click search (PPC) advertising. It is quite possible that banners may direct the greater volume of traffic and have a lower Bounce Rate whilst PPC may drive a lower volume with a higher Bounce Rate yet despite this the overall volume of good outcomes may still be low. In this instance it would be worth looking at overall volume of good outcomes for both referral sources and see which has the better conversion rate – sometimes the effect of a higher Bounce Rate can actually act as a quality filter to weed out the wheat from the chaff.

If you use or are thinking of using a tag based analytics tool pay close attention to your page naming convention. Get this wrong and you’re toast. Get it right and life will be sweet.

The page naming convention is the piece of the jigsaw that allows you to find and identify content groups and pages in your analytics GUI. Some tools like Google Analytics don’t make use of page naming conventions, they just use the page URLs and / or possibly page titles. Others like WebSideStory’s HBX have specific variables in the tag that need to be filled.

The page naming convention allows the user to split areas of a site into content groups and identify pages uniquely within those content groups. It might be easier to think of it as a kind of Russian Doll. You have the site as the main content group, within that you may have a number of other content groups representing your main areas and within each of those main areas you may have another sub-level(s) of content and so on. Establishing unique content groups and page is important when creating funnels and segments which are crucial to doing in depth analysis.

It’s wise to create a naming convention before adding the tags. Within WebSideStory’s HBX the content group names and the page names are represented in two main variable within the javascript tag, these are identified as PN=… and MLC=…

JS tags for other tools use similarly identifiable variables.

Stuff to bear in mind before creating a decent page naming convention:
1. Before anything else, always think about the (business) objectives for the site and what questions your analytics tool will need to answer . Gather together the key stakeholders and ask them what their requirements are. It’s a nightmare to go through all the trouble of setting up an analytics tool only to find you can’t answers questions from a key stakeholder.

2. Do you want a “horizontal” or “vertical” naming convention?
“Horizontal” naming conventions generally identify pages by type right the way across the site i.e. Category, sub-category, product pages etc.

“Vertical” naming conventions identify pages by section i.e. ‘News>UK>article title’, ‘Sport>football>article title’.
Generally speaking media sites like online news providers tend to use vertical page naming conventions while sites selling stuff tend to use horizontal page naming conventions.

3. Don’t use illegal characters. These may be things like “()-*&? etc. They may also vary from tool to tool. Always check with the analytics provider first. To be absolutely safe stick to upper and lower case letters and “/” to separate content groups.
There are other issues with tag based tools around ensuring that you are tracking links properly, streaming video if you use it, internal search etc but these are technical issues that should be covered off as part of the essential house keeping of tagging.
If you use a tag based system that doesn’t make use of page naming conventions consider how you can improve things by using search engine friendly page URLs. There are two good reasons to make the effort here:
1. It will help categorise pages in the analytics tool user interfac
2. It will help with your SEO
Example: (you can click on the image to enlarge it)

If you are looking at setting up SEO friendly URLs you can still consider them in the context of horizontal of vertical although they will work better for SEO if they are based on a vertical solution. The trade off is a business decision you will have to take.

The objective of this post has simply been to point out that the level of ease and therefore the depth of insight you get from your analytics tool can be much enhanced if you take the time at the outset to establish a decent road map in you naming convention. I have saved myself a lot of agro in past doing this. Any thoughts or comments most welcome.

Solutions implemented to improve site performance will not always appeal to everybody.

The average conversion rate may rise (and fall) but really there’s no such thing as an average conversion rate. Performance and conversion rates become more useful when they are segmented based on audience type.

Audience types can be defined and segmented in many different ways and people who buy advertising space spend their lives thinking about this. In web analytics segmentation is based on actual observed behaviour and leaves less to guess work. Segmentation ranges from how a visitor found your site to what they did on site. Depending on the level of sophistication of the analytics tool, cross segmentation can be used to obtain even better insight. Segmentation can then be used to represent groups of visitors with different objectives and levels of motivation.
Developing some understanding of visitor motivation helps minimise the risk of losing customers and maximise the potential for retaining new and existing customers.

It also helps to give some kind of weighting to your segments. This makes it easier to know where to concentrate resources.

A 3 dimensional bubble chart can help visualise this (you can click on the chart to enlarge).

In this example the Y axis shows the conversion rate within each segment based on a specific KPI; the X axis shows, as a percentage, the contribution by each segment to the total volume of good outcomes; the size of the bubble for each segment effectively acts as the weighting metric – the larger the bubble the bigger the weighting in context of all site visits.

By creating this chart it’s possible to see which segments are the heavy hitters and which have the most potential. It also gives more meaning to scenarios when the segments are overlaid on them.
Segments and scenarios
Scenarios are created to understand visitor behaviour by looking at conversion through the scenario. If several different segmented visitor types are overlaid on a scenario it’s possible to get a much richer insight into levels of motivation.

In this simple two step example scenario, it’s possible to see how the various segmented sources of traffic convert on a comparative bases. (Click to enlarge)

As I’ve already mentioned, once you start looking at segments overlaid on to scenarios, it may help to cross reference some of them to gain deeper insight. For example, say 40% of your traffic entered on the home page and 40% entered on a product page and you’ve created segments for each to see how they convert; then assume 50% of your traffic comes from organic search and this source generally converts well. You may then want to know how does traffic from organic search that enters on the home page compare with traffic from the same source entering on a product page? More importantly even if one converts better than the other does that mean the poorer converting group has no value? Not necessarily – they may just be looking for something else - check the search terms they came in on, look at the bounce rate and so on.

As more on and off-line marketing is tracked by campaign tracking and bespoke landing pages segmentation will surely become increasingly important in assessing performance and value for money. I have already experienced this with several clients.

Any thoughts / comments please do post. Many thanks.

A butterfly beats its wings in the Mojave desert and there is a tornado in China

I’m not suggesting that web analytics is like Chaos theory but I am suggesting that seemingly small changes to a site can have big impacts.

It’s a happy but rare occurrence when one fix is so clear that by its implementation a site will be transformed from having poor performance to stellar performance. A cost / benefit analysis will help demonstrate expected return.

Grand solutions like fixing internal search, reducing basket leakage or improving retention through the checkout process are all quite obvious; alternatively driving more visits to a product page or improving conversion from product pages to the shopping basket are equally so.

These are important areas which everybody should cover off in the process of analysing a site and getting them right should yield good returns. But it is because they are so important and demand so much attention that we may be overlooking a number of other smaller, simpler and cheaper fixes that could have an equal effect on performance.

Example:
The site analytics data shows a relatively low conversion from product page to basket page and this has an impact on overall site conversion. When a customer adds an item to the shopping basket they should then automatically be taken to the shopping basket. “A bird in the hand is worth two in the bush”. Many sites assume that the customer will want to continue shopping or that he/she can be persuaded to continue shopping and so the customer is left on the product page. A look at the average number of items purchased per order or per visit will resolve that question.

I have seen conversion rates double by making this simple navigational change. In this case the decision process about whether to progress to the shopping basket or not was taken out of the customers hands. If a customer wants to continue shopping there is nothing to prevent it after arrival on the basket page.

We spend a good deal of time trying to effectively second guess what our customers want and then providing them with the options. On some occasions it may be preferable for both the customer and the e-tailer to have a decision taken by default.

There are more of these simple and extremely cost efficient fixes out there. Analysing site data gets us closer to them. Making sure we already have our analytics configured by default to measure performance on the big stuff gives us more time to explore the data for these little gems.

I used to work in online media so I’m used to blank expressions and a polite change of subject when I answered the question, “what do you do?”

After travelling for a year in 2003 I came back and took a job in the new and emerging area of web analytics.

Friends, family and people I met began asking The Question again (some more than once!) and as I gave my long winded answer I detected a greater level of interest than I had before. There seemed to be an opportunity, but I needed to get it across in a more digestible way. An analogy was required - supermarket shelf stacking.

Web analytics is like product placement on supermarket shelves. Supermarkets use every trick in the book from puffing bakery air onto customers making them feel more hungry as they enter the shop to placing frequent buy products like milk and eggs at the back of the store so customers have to walk past other products on their way to pick up the two things they came in for. Supermarkets also place more expensive items at eye level and slightly to the right because we read and browse from left to right. The eye then comes to rest on them. Related products are grouped together in continuously different ways to improve cross sell. The list is endless…

Web analysis is essentially part of the same process but taken online. Data is studied so we can understand what our customers are like, why they have come to the site, what they are interested in and what their journey through the site is like. When we’ve reached the limits of the data we turn to usability in its various guises and to online surveys. I think this is like the mortar that fills in the gaps as we build our impression of how customers are behaving online and from there how we can best meets their needs.

The result is that we go and tinker with the web pages to make sure that we are putting products and the most tempting messages in the best places to help the potential customer find what they came to buy.

The principle isn’t much more complicated than that, the techniques are. The object here and in the work that we do with our clients is to illustrate the idea, point out that it works in the same way it work for supermarkets (very well), and suggest that it’s worth doing and that if it’s done well it can pay for itself.