Things change so quickly on the web. When I started working at software firm Urchin in 1996, web analytics was a niche product, important to (and understood by) perhaps a handful of people at an organisation.
When Google bought Urchin in June 2005 and launched Google Analytics later that autumn, the industry was moving away from expensive, complex tools in favour of free, easy-to-use ones, and these are quickly becoming more capable than those of the previous generation.
Analytics has also expanded to encompass a larger tool set and conversion process. It includes offline data sources and multivariate testing solutions. Perhaps the largest shift is that virtually everyone is now aware of and able to afford a quality web analytics and multivariate testing platform.
You can sign up for Google Analytics at google.com/analytics. From there, you can create a new, free account. If you advertise on Google's AdWords system, you can also sign up for Analytics from within your AdWords account. Google Analytics has been closely integrated into your AdWords account under the new Analytics tab: click the tab and follow the instructions to start tracking AdWords campaign data and ROI information.
The Tracking Instructions page that appears contains the tracking code that you'll need to paste into each page of your site. Before doing so, you might want to complete a couple more steps to make sure Google Analytics collects the most relevant data for your site. You can configure your profile by clicking Edit in the Website Profile table.
Set the default (or index) page of your site. This will allow Google Analytics to reconcile log entries for www.example.com and www.example.com/index.html, for instance. These are in fact the same page, but are reported as two distinct pages until the Default Page setting has been configured.
Does your site use dynamic session or user identifiers? You can tell Google Analytics to ignore these variables and not count them as unique pages. Enter any query parameters you wish to be excluded, separated with commas.
To enable ecommerce reporting and the Ecommerce Analysis report set, select Yes. If your website is designed to drive visitors to a particular page, such as a purchase or email sign-up page, you can track the number of successful conversions using 'goals' and 'funnels' in Google Analytics. A goal is a website page a visitor reaches once they have made a purchase or completed another desired action, such as a registration or download. A funnel represents the path that you expect visitors to take in order to reach their goal.
Defining these pages enables you to see how frequently visitors abandon goals (and where they go instead) and the value of the goal. Each profile can have up to four goals, with a defined funnel for each. You can begin defining goals and funnels by selecting a profile and clicking Edit from the Analytics Settings page.
Hearts and minds
Google Analytics is not only a way to learn about customer behaviour, it can reveal something that's even more important: customer intent. Analytics data provides a peek into the hearts and minds of your customers and their needs and wants.
How did your visitors arrive at your site? For those who came through a search-engine query, a list of search terms will tell you what they were seeking. For example, shoe shoppers may be looking for Nikes but not Uggs; Clarks but not Donna Karan. You may therefore discover that part of your site is on target for visitors' needs, while the other half isn't. Similarly, Google Analytics can give you a list of 'referring URLs', websites that send you traffic. You can infer, for instance, that visitors coming from happycoupons.co.uk will have a different goal (making savings) than those coming from a manufacturer's retail link page (finding out about product features).
What are visitors looking for? Your checkout page tells you only where you've succeeded – not where you've missed an opportunity. Consider Best Buy's experience stocking a portable refrigerator to chill beer in the US. At first, they stocked it only during the Super Bowl. But search keywords reports revealed that shoppers still looked for the product code well into the baseball season – yet they couldn't find anything on the site.
Other key data that goes beyond conversions is the 'cart abandonment' rate: the percentage of customers who put items in the cart but leave your site before checking out. This crucial metric of your site's ability to close sales may indicate that something's amiss with your checkout process.
Where are your visitors landing, bouncing, and viewing? It's often assumed user experience begins on the homepage, and this misconception drives many an ecommerce site to waste hours of design work in the wrong place. Search engines dig deeper into ecommerce sites, bringing visitors to not just 'electronics', but also televisions, MP3 players or sat navs. Analytics data will tell you where your real 'homepages' reside, so you can focus your design work there. Curious? Just take a look at the 'Top Landing Pages' or 'Top Entry Pages' report in your Web Analytics tool.
Conversely, analytics will tell you which landing pages have the highest bounce rate – ie on which did people land, look around and quickly leave? This data tells you which of your pages are letting your customers down, and can also help with your redesign, since you can infer intent through the list of keywords and referring sites.
Web analytics can also show you the top viewed pages: information that's often overlooked. While you may consider yourself in the business of selling products, most of your hits could represent people reading customer reviews. Or perhaps you're selling 10 lines of products, but two of them show especially high traffic. Knowing what interests your customers will help you design a site that better meets expectations.
Analytics can help you understand what drives performance up or down. Earning $15,000 in the last 24 hours is good; knowing what drove that surge is even better. Ecommerce tracking shows the number of orders placed, the value of those orders, and more, by hour, day, week and month. By segmenting your data over different time lines, you can see both seasonal trends and more subtle buying habits that could otherwise go unnoticed.
Putting this information into practice takes not just creativity, but also a willingness to experiment. This spirit of adventure comes through on some of the most successful retail sites. Crutchfield (crutchfield.com) has taken the unorthodox approach of putting its checkout cart on the left side of the screen. Is that a good idea? Analytics reports have confirmed that it is – at least for them. Wal-Mart commonly puts products on its website that are unavailable in its stores. When those zebra- patterned bed sheets prove they have a following, the company understands the demographics enough to place them in targeted store locations.
Earlier this year we redesigned the Google Analytics site. The goal was to make it easier to discover information relationships via navigation and visualisations. We created a customisable dashboard and introduced sparklines. We developed new graphing tools and a new type of date slider, which make it easy to see spikes and dips in traffic as you set date ranges. And one of the most popular new features is one of the simplest: the ability to email reports and schedule these emails so that information can be easily shared with key stakeholders in your organisation.
New visualisations have been among the most popular features. You can now view many reports by hour or day and graph data by day, week or month. And we've made it easier to compare visits to conversions. You can also graph two metrics against each other over time so, for example, you can compare the number of visitors and the bounce rate for a certain week, or see if visitors who come to a site through AdWords spend more or less time on your site than visitors overall.
We've also added a new ga.js tracking code. This pagetag allows for more flexibility and customisation. It's just as easy to install as the old code, but enables more seasoned users to track ecommerce transactions in a more readable way and make use of advanced tracking features. We've also added a Google Analytics codesite (code.google.com/apis/analytics/docs/) to help you take advantage of documented customisations that have been made to the tracking code.
If you have content behind a security firewall, an intranet or an internal network that prevents you from using Google Analytics, you may also want to consider Urchin software. You can configure it to fit your own requirements and process/reprocess log files as frequently as you wish. Urchin is also great for intranets, since it allows the analysis of firewall-protected content, such as corporate intranets, without any outside internet connection. You can even track your site with Urchin and Google Analytics combined.
One of the coolest things we've recently added is industry benchmarking. Still in beta, it enables you to see how your site's data compares to others'. For example, if you have a travel website and you get a spike in traffic on Mondays, you may want to know whether other travel sites get that same spike. We don't share individual data with competitors, but bucket data into industry verticals and then anonymise and aggregate it.
Finally, Website Optimizer is a free tool that complements the functionality of Google Analytics to hone your site further. You can create different versions of your web pages, and Google then splits your traffic automatically, so your visitors tell you which version they like best. Things you might consider changing include images (bigger, smaller, colour, black and white, with models, no models); headlines (questions, shorter ones, emphasising different points); calls to action (different button designs and copy) and layouts (three columns, two columns, one long scrolling column).
Test a few big changes, not several small ones. A good rule of thumb is one page variation for every 100 conversions. So if you get 300 conversions per month, test up to three variations, including your current page. Run your test for at least two weeks. Don't jump to conclusions: make sure the data has an opportunity to normalise.
Make testing an ongoing process. Once you find a winner, keep trying to beat it. Remember that not finding a winner can be helpful, too, since you learn what doesn't work and you protect yourself from making changes that could have permanently harmed your site's performance.
First published in .Net Magazine, Issue 181