Segment Analytics Data Using Personas

Many teams create personas during the initial ideation and design phase of a project, but then fail to leverage them beyond settling design debates. One method of incorporating personas into the ongoing maintenance of a website is to create persona-inspired segments in your analytics tool. These segments can not only check whether the user described in a persona is characteristic of your website’s real visitors, but can also help you uncover patterns of use and trends in behavior that would otherwise be masked when lumping together the data for all visitors to the site.

What Is a Persona?

A persona is a fictional character representative of a unique group of users who share common goals. An organization should possess several personas to reflect the variety of visitors that its website attracts (most commonly, 3–7 personas will cover the majority of an audience without creating too specific or spurious distinctions in your user types). These user archetypes should ideally be based on qualitative, ethnographic user research in order to include accurate behaviors, environments, attitudes, and needs of real users. Personal details such as a name, photo, and specific contextual narratives should be combined with a description of gender, age, marital status, job title, device ownership, and other demographic information to create an easy-to-imagine, relatable character.

Benefits of Using Personas

Without personas, organizations risk creating a one-size-fits-all design expected to suit all users. For example, during our research for our E-Commerce report series, we identified 5 distinct types of e-commerce shoppers, all of which may visit the same website and expect different levels of detail and types of information about a product. Even a system as contained as a company intranet will have multiple types of users accessing it, all with different goals and tasks that they need to accomplish. Without identifying the various characteristics of the user groups visiting your site you cannot hope to design an experience that includes the key elements that each type of user needs. Instead, you will end up creating a website that doesn’t perform well for anyone.

A major strength of personas is that they focus design efforts around user types and their specific needs or behaviors, and facilitate discussions among team members and stakeholders. Framing the design conversation around a persona creates context and makes it easier to empathize with the users affected by the design decisions. (Another way to garner empathy: have the stakeholders observe a usability test.) Once there is a common understanding of customers across your team, these personas help establish a culture of verifying instead of assuming: rather than debating whether or not “people” need a proposed feature, ask, “How would this help Barbara?”

Persona-Inspired Segments

Once personas are created, it’s easy to let them fall into disuse when the site goes live. Don’t do that! A lot of time and research effort (hopefully!) went into forming these realistic representations of your audience, and return on investment increases if you can continue to leverage the personas to drive improvements. By creating a segment in your analytics tool inspired by the established personas, you are able to analyze how real users actually use the website or application. This information can validate any assumptions made during the persona-creation process. It also allows the refinement of the personas if new information is uncovered—personas need not be a document created once and never touched again. Additionally, revisiting personas through analytics is much more maintainable than continuously contacting customers to conduct interviews, diary studies, and other forms of intensive user research.

The key step to creating a persona-based segment in your analytics tool is to determine what characteristics of the persona to include in the segment filters. When reading through the persona details and any accompanying user stories, you must separate out the characteristics that are representative of that specific group of users from those added merely to lend verisimilitude. For example, a user story from a persona for a software–as–a–service (SaaS) may describe a return visitor, David, as follows:

“David is a return visitor. He receives our weekly email newsletter with tips on using new features on Monday morning at 10AM, immediately after his status meeting at the office. He clicks through from his Android phone. He has time to read a single blog post before his next meeting.”

When creating a segment to represent David, the exact time of day is likely too detailed to include as a filter, but there may be a difference in behavior between users who visit the site during the week and those who visit it during the weekend. Or maybe between visits happening during normal office hours and those during the evenings. The fact that he is a newsletter subscriber and already a customer of the service should also be included to distinguish this group of users from those who are still in the research phase and perhaps just learning about the software and its features. In this instance, the persona’s gender likely does not make a difference (but it would for an e-commerce clothing site, for example), so it should not be included when creating the representative segment.

To be meaningful and worth a separate analysis, the segment derived from a persona, once created, should display user behavior that is clearly different than the rest of the site visitors and should represent a sizeable chunk of your user population. For many, this reasonable chunk falls within 7–10% of overall visitors, but ultimately it is up to you to determine what makes sense for you and your organization. The amount of details from a persona used to create the corresponding segment can be adjusted to allow the segment to represent a larger or smaller group of users. Focus on the most distinguishing characteristics of that group of users first and foremost, and then add in more details if necessary to target a smaller subset of users.

Coming back to our SaaS example, in order to create a segment based on David, we would need to know what other details from the persona description may be meaningful. Is the fact that he uses his mobile phone to access the site relevant for the group he stands for? How about the fact that he uses an Android phone as opposed to an iOS or Windows phone? To know for sure why a particular detail was included, we might need to revisit the original user research conducted when the persona was developed. Additionally, we’d have to determine whether that behavior actually distinguishes the group of interest from another: do a lot of David-like users access the site on their phone and behave differently than other mobile users?  Or perhaps all types of users sometimes access the site on smartphones and do a small number of similar things? If the latter, it would be more useful to analyze mobile usage using that group’s overall segment rather than creating a separate mobile segment for each.

Examples of ways to segment users that may be derived from a persona include:

• Demographics: Age range and/or gender
• Geographic locations: specific countries, regions, urban vs. suburban, and so on
• Device and/or browser
• New vs. returning visitor, logged-in user vs. not logged in or doesn’t have an account
• Source: arriving from an email, search engine, a specific social network, or set of referring sites
• Whether the visitor has searched for branded vs. unbranded keywords or terms
• Whether the visitor has reached a specific set of pages: for example, visited a product detail page, visited the customer service section, or landed in the content section aimed at trade professionals or wholesale buyers

This list is by no means exhaustive, and the set of characteristics included in a segment will vary depending on the site, its audience, as well as what is technically feasible to define in the chosen analytics tool.

Use Segments to Avoid Drowning in Data

Interpreting analytics data to answer specific questions becomes much easier when you use segments to narrow down the volume of data available and hone in on the relevant statistics. Once a persona-inspired segment has been created, most reports within your analytics tool of choice can be filtered to show only the data pertaining to that group. Trends in behavior and site usage for individual user types will emerge more clearly than they could when viewing data for all site traffic.

For example, when looking at a metric such as bounce rate, it is easy to get distracted by the overall rate of all visitors who bounce after reaching a certain page or group of pages. However, this number by itself isn’t actionable information, since several different types of users land on the page from different sources, with different expectations for the page content. Each bounce rate must be analyzed separately in order to uncover anything meaningful. Let’s consider two groups of users that are represented by two personas: David, the loyal visitor and newsletter subscriber, and Mary, a marketing manager at an organization with no in-depth technical knowledge. Are David-like users (repeat, frequent visitors coming from the email newsletter) bouncing from blog posts about tips for using existing features? That’s okay, we expect loyal content consumers to bounce because they have already previously consumed the majority of content on the website­. How about users in Mary’s segment, who arrive on the page from relevant queries on search engines?  If they are also immediately bouncing, maybe there is a mismatch between what users expect from the search-engine results page (SERP) and what content is actually delivered on the page, which is something that should be investigated further. On the other hand, if Mary-type users are in fact not bouncing, nothing may need to be done to the page or to the links leading to it. Such insights can only come from segmenting the data to unmask the helpful information.

Simplified example of analyzing bounce rate of a page: An overall bounce rate of 65% masks the difference between the rates for loyal visitors and visitors searching specifically for information on the page topic. Only when you segment the data can you see such distinct differences and better understand how the page performs given the specific goals of each audience.

Not only do segments enable us to analyze specific metrics in detail, but they can also uncover behavioral patterns and answer questions such as: “Do new visitors landing on article pages from Google also visit any other types of pages?” and “Do newsletter subscribers and nonsubscribers behave differently? Specifically, are subscribers more likely to download whitepapers, contact us for consulting services, or upgrade their membership?” A significant difference in the conversion rate for a segment not only indicates that the segment is likely a valid representation of a distinct user type, but also reveals what group of users you should continue to pursue in your content strategy and aim to grow. For example, if you found that newsletter subscribers upgrade their membership level to access more features than the nonsubscribers, it would be worthwhile to research ways of increasing subscription rates.

While quantitative data from analytics tools can never tell you why users performed in a certain way—only what they did—on your website, uncovering these behavioral patterns can focus and inform user-research activities such as qualitative usability testing. Armed with this triangulated data, you will be well on your way to optimizing your site and its content to better satisfy the needs of your real users.

Learn more about how to extract meaningful UX insights from analytics data in our full-day training course on Analytics and User Experience.

Persona photo credit: "Paul reading HTML5 For Web Designers" by Jeremy Keith, used under CC BY / Modified from original