Segmenting According to Demographic Aspects
Web analytics can’t tell you much about who your users are without being linked to another data source (see the “Custom Variables” sidebar in Chapter 9, which touches on integrating other data sources with Google Analytics). You may be able to segment according to two demographic dimensions with the data you get with the standard setup that most people will encounter:
■ Where your persona lives (city, metro area, state, or country)
■ What language the user’s browser is set to
Many readers will not find these to be useful ways to differentiate users, but your situation may be different. AwesomePetToys.com is an e-commerce website and, as such, will sell awesome pet toys to pet owners no matter where they live. Geography and language simply isn’t a useful way of dif- ferentiating users since they found that cultural differences aren’t as impor- tant as other factors like pet ownership and disposable income. On the other hand, a company like Thomson Reuters Techstreet, a Pure Visibility client that sells technical standards documents in various languages, might be interested in understanding any differences between users with different
native languages—particularly with regard to what documents they search for and purchase.
Segmenting According to Behavior
Segmenting according to user behavior is what web analytics does best and, being familiar with your options, you may wish to adjust your user research agenda to collect more data that you can use for segmentation.
Some of the behaviors that you can use for segmentation and possibly learn about directly from users are:
■ When users visit (day of week or time of day)
■ How often they visit (have they visited before, and how often they visit)
■ How they get to your website
■ What they search for, either in search engines or on your website
■ What page they land on
■ Whether or not they visit a page or type of group of pages
■ Whether or not they complete a goal
Or, to put it another way, when they access your website, how they access your website, and what they do on your website. What users search for on or off your website and whether or not they view pages (or take other actions on your web- site) are the areas where you infer their motivations based on their behavior.
What Can You Do with Segmentation?
There are two primary things you can do with segments that you have based on the research you used to create personas: explore analytics data to find patterns and answer specific questions in a more accurate way.
Exploration involves looking at data in various reports to see how your seg- ment compares to the general population of users or to another segment to see what patterns may emerge. This activity may drive further research or uncover data that you can incorporate into your personas to give them greater verisimilitude.
Answering questions more accurately simply means that for any kind of anal- ysis, you could use your persona-inspired segment rather than unsegmented data. You could examine common paths to and from a page or analyze key- words and search terms for your segment, or verify whether changes to your website have affected the right user in the desired manner. The important thing to remember is that any segment you create based on other kinds of user research is just going to narrow down the amount of data you look at—it can’t give you a perfect view of just your target users and no one else.
You may attempt to segment to answer a question more accurately when a trend is too slight to notice when you look at data for all of your users.
Personas 161
Otherwise, because your focus is on trends over time, it is often not necessary to go to the effort for this additional accuracy.
For example, AwesomePetToys.com not only serves pet owners with money to spend on their pets, it also sells to businesses that these pet owners use. If AwesomePetToys.com changes its main navigation to make it easier to identify the wholesale section of the website, they could use their business purchaser segment to filter out the noise of pet owners who accidentally go to the wholesale part of the website, and therefore get more accurate measurements of user behavior before and after they change the design.
Because pet owners probably wouldn’t actually purchase large quantities of awesome pet toys wholesale, segmenting data according to whether they visited the wholesale section wouldn’t really make any trends in sales more clear. On the other hand, changes to the wholesale section of the website may actually drive away pet owners who accidentally stray into that part of the website, lowering pageviews for all of those pages when you look at data for all users. In this situation, segmenting to show only business users would paint a better picture of performance, perhaps showing that business owners are more engaged with the wholesale section of the website, even while pet owners are fleeing.
Building Better Personas
You can use web analytics data to build more realistic personas by looking at how common your persona’s actions are and adding realistic data. The chal- lenge is to examine the right data—the data that describe the people your per- sona represent.
The way you build better personas from analytics data is through iteration—
start building a segment based on what you know about your users and then explore the data to look for other patterns that fit with what you know, such as common search keywords, pages users visit, or when they use the website.
Optimally, you go on to verify this information through further user research, otherwise this is, of course, informed guesswork.
Instead of building the Emily persona segment we described earlier, AwesomePetToys.com could build a segment showing data only for people who purchased pet toys in the consumer portion of their website. Using this segment, the analysts at AwesomePetToys.com could then study the key- words that users searched for that lead them to the website and find ways that their keywords differ from the rest of the website’s users. In the case of AwesomePetToys.com, they found that people who bought pet toys reached the website by searching for some variation on “AwesomePetToys” or by the brand of toy (e.g., “ToyCo laser pointer”). They could build up a collection of
keywords that Emily may search for that would capture pet owners who spoil their pets—both the ones who purchase expensive pet toys and the ones who don’t. In this way, they could back into what would hopefully be a similar segment.
As with other research activities, you interpret the data to derive meaning. Other people could look at the same data and draw different conclusions. There is no bulletproof way to approach this situation—in fact, debate is usually healthy. The best thing that you can do is keep track of where your data came from or how you got them, and the intellectual steps you took from raw data to conclusions. In many cases, being able to describe a thought process is enough to assuage the concerns of others.