One of the final sessions from last week's Gilbane conference (see our coverage here) in San Francisco focused on Personas, Market Segmentation and User Experience design. Strong assertions were made (e.g., "data without segmentation is crap") and key lessons shared. Here are the highlights and take-aways.
Personas: The Basics
A quick overview of the key aspects of personas was delivered. They are:
- Named user types with personal characteristics
- Based on research
- Tied to an organizations' market segmentation strategy
You can have as many persona's as makes sense, but to start put energy into the ones that most directly map to business goals (lead generation, sales, etc.)
Robustness is Key
According to Casburn and Woods the more robust the research backing personas and market segmentation the more likely you are to reach your business goals using them. Not all data is created equal.
There are 3 sources of persona data:
- Formal Inquiry
This includes market research, focus groups, 1-on-1 interviews, surveys, and user testing. It is considered the most robust source of persona data.
- Internal Company Knowledge
Customer service interactions, sales interactions, web analytics, search logs. This data isn't as strong as formal inquiry.
- Ambient Data
This includes social media information, user generated content, etc. This is the least robust and reliable of the data sources.
Developing Useful Personas
Casburn and Woods then described some ideas and methods behind creating effective personas. It's not just enough to have personas defined. A badly thought-out persona can be a negative in reaching business goals:
- Think about what your users want from the site, and how they would go about doing it.
- Determine what the user's motivations and pain points are.
- Recommend content and site features to address those issues.
Most Organizations Want Something from the Visitor
Woods then jumped into a topic that isn't always discussed directly. It's not enough to just look at what the visitor may want, most organizations also want something from the visitor. The site may want to:
- Sell a product
- Convert a visitor into a sales lead
- Convince a visitor to make a donation
- Convert them to a newsletter subscriber
- Have them engage by rating or commenting
Organizations need to define their preferred business outcomes and determine what metrics they can collect to track them. The website should be seen as a negotiating field between the visitor and the organization. Effective personas supported by research and metrics will help guide that visitor to the desired business outcomes.
Levels of User-facing Segmentation
Woods kicked this section off with a quote from a blogger:
Data without Segmentation is Crap
Casburn proceed to lay out some ways for sites to get started with user segmentation.
The Basics: Segmenting by Core Target Audience
If you don't have a lot of research or data to analyze you can start with segmentation by picking your top target audiences and then providing paths through the site that answer their questions and lead them towards the business goals specific to them. Casburn then provided an example of a university site where the the homepage provided different tracks for visitors in the following way (a subset):
- Students - Providing information on campus life, classes and majors, sports.
- Parents - Cost details, financial assistance data
- Alumni - Meeting information for alumni groups, Expansion plans, Ways to donate
These target audiences are looking for very different things from the website, so just getting started with the basics of segmentation holds a lot of value for the organization.
Digging a Little Deeper: Segmenting by Task
If the organization has a deeper level of knowledge about their visitors they might start segmenting based on the specific tasks the visitors are trying to accomplish. An example provided was from a travel recommendation site where the primary segmentation was:
- Traveling with Pets
- Traveling with Kids
- Traveling with Friends
Deeper Still: Relationship Segmentation
In this case, the organization knows something about the visitor, perhaps via a cookie allowing the organization to connect with what the visitor has done in the past, or if they are a registered user. Here, the site would detect this relationship when the user arrives and tailor the message specifically to them.
The example used here was the Zipcar local car sharing service. If you are a first-time visitor the site promotes information on ways to sign up, cars available in the city the visitor is coming from, etc. If the user is a returning user the site provides content specific to their past usage (new cars in locations they have picked up), links to their account, links to book a reservation, etc. The information given to the user on the homepage was very different, but presented within the same structural framework.
Categorization in Real Time
Casburn and Woods went into some technical detail relating to methods of providing real-time categorization and personalization. They began with noting that not all Web CMS products easily allow for real-time personalization.
Their examples in this segment were based on their prior work with Sitecore (news, site) which provides both a Rules Engine and associated Online Marketing Suite -- which allows marketers to both track how visitors are using the site as well as build rules to allow the site to respond automatically to specific types of usage.
They were careful to note that other WCM systems -- such as CoreMedia, Day Software, Ektron, SDL Tridion, Vignette/OpenText and others -- support similar processes, and that many other Web CMS products do not have this capability, at least not without a lot of customization.
Scoring Pages Based on Business Goals
Using the Sitecore tool set Casburn and Woods showed how you define specific goals and then score specific pages within the site as to how they advance that specific goal or show the user to be interested in one thing vs. another.
Probable Conversion Scoring
For example, in a lead generation/purchasing track a visit to a product page might be scored a 5, but a visit to the pricing page for the product might be scored an 8, to represent that the pricing page view is indicative of a stronger desire to make a near term purchase.
As the visitor navigates the website their score is updated and monitored and based on the score relating to specific goals the site can then react in real time to provide more relevant information. So, a user who has a high score as a near-term purchaser could be offered a form to contact sales, or a callback from the sales team.
Other Forms of Scoring
Scoring can also be used to remove visitors from contention or detect action of competitors. Scoring might be set to zero for known search bots or for competitor visits so that the site's analytic information doesn't get muddied with "false positives".
In a more malicious example, in one site, the system was trained to detect visits from specific competitors and when those competitors went to download white papers from the site it would allow them to download, but would give them old, outdated versions of the documents. This was meant more as a subtle tweak to competitors rather than a core business strategy, but it shows the power of these tools to react to specific business circumstances in real-time.
A key point that Casburn and Woods made is that, even though there's a lot of technology making this happen, the control of what gets tracked and how the system responds needs to be in the hands of the marketers and folks leading the business.
The site needs to react quickly based on business requirements, and this is not an IT task.
A Small Case Study
Woods and Casburn then walked through a quick case study example:
- Mike, an "urban hippie" living in Portland want to goes skiing at Mt. Hood, which is nearby
- Mike doesn't have a car, so he goes to the Mt. Hood website to look at transportation options
- He sees Zipcar listed on the Mt. Hood transportation page and thinks that's a good idea, so he clicks the link
- He arrives at the Zipcar page and looks at the cars listed but he doesn't see any information on whether those cars can hold his skis or gear, or if they have ski racks
- He leaves the Zipcar site in frustration without signing up for a membership
Zipcar, however, happens to be collecting metrics on user visits and one of the marketing staff is looking through the data to try to determine why people left the site. This person sees Mike's visit and sees that he was looking for available cars before he left, and it's also shown that Mike came from the Mt. Hood website.
This gives enough information to come up with a simple theory: visitors who are coming from skiing-related sites might be interested in knowing if the cars can handle skiing gear.
This information isn't relevant for everyone, as not all users are using Zipcar in areas where there is local access to skiing, but using real-time personalization a rule could be quickly put into place to show skiing information to users coming from skiing sites or who are in markets with local access to skiing.
Zipcar could quickly implement the new rule then check back later to see if there was a positive impact. If the theory was correct then the rule would stay and perhaps expanded on, if it didn't work out then it's quickly removed.
The Wrap -- Simple Testing is Critical
It's key to be able to effect change and quickly test the results. This allows marketers to show quick wins which might enable them to go even deeper with their market segmentation and categorization projects. More importantly, it allows them to run lots of small experiments.
All of this optimization involves a certain amount of guesswork. Only through continued testing and refinements do we make progress. Look for this when evaluating your WCM and WEM technology investments.