User Research and Persona Creation Part 2: Segmentation – Six steps to our Qualitative Personas
Posted August 30, 2010on:
In Lorraine’s last blog she described the data gathering methods used to obtain representative data from users of Edinburgh University’s Library services, the purpose of which was to identify patterns in user behaviours, expectations and motivations to form the basis of our personas. Raw data can be difficult to process and it is impossible to jump from raw notes to finished persona in one step, hence our six step guide.
There is no one right way to create personas and it depends on a lot of things, including how much effort and budget you can afford to invest. There are lots of articles on the web detailing various approaches and after much reading we decided to rely on 2 main sources of information which we felt best suited our needs.
One resource was the Fluid Project Wiki which is an open, collaborative project to improve the user experience of community source software and provides lots of useful guidance as well as sample personas. The other resource which we heavily relied on throughout the whole process was Steve Mulder’s book The User Is Always Right: A Practical Guide to Creating and Using Personas for the Web, which contains lots of great advice as well as step-by-step coverage on user segmentation.
There are three primary approaches to persona creation, based on the type of research and analysis performed:
- Qualitative personas
- Qualitative personas with quantitative validation
- Quantitative personas
There are a number of important steps to go through in order to get from raw data to personas and I will now explain the tools and methods used to generate our segments and personas for anyone who wishes to follow in our footsteps.
The first thing we did was to plan out a schedule of work which consisted of the following:
- Review and refine interview notes in the project wiki and flesh out user goals
- Write summaries for each of the participants
- Do a Two by Two comparison, to identify key similarities/differences
- Identify segments
- Write the personas
- Review personas
Step 1: Review/refine notes
We spent a day reviewing our notes in the wiki and fleshing out goals by referring to written notes taking during each interview, checking the audio recordings where necessary. We worked as a team which was beneficial as we were both present for each interview and therefore had a good grasp of all the data in front of us. Once we were happy with our set of notes, we printed out participant’s interview notes and attached each to the white board to make it easier to review all data grouped together.
Step 2: Summarise participants
- Practical and personal goals
- Information seeking behaviour
- How they relate to library services
- Skills, abilities and interests
We used different coloured post-it notes to denote each of the above categories. Once we had gone through this process for each participant, our whiteboard was transformed into a colourful mirage of notes.
We were now ready to start a two-by-two comparison of participants.
Step 3: Two-by-Two Comparisons
The next step utilised the two-by-two comparison method, a technique advocated by Jared Spool at User Interface Engineering (UIE.com). This works by reading 2 randomly chosen participant summaries and listing attributes that make the participants similar and different. We then replaced one of the summaries with another randomly chosen one and repeated the process until all summaries were read.
Below is a list of some of the distinctions identified between our participants, using this method:
- Type of library user
- Years at Edinburgh University
- Use of Edinburgh University library resources (digital and physical)
- Use of external resources
- System Preference (Classic or Aquabrowser)
- Attitude to individual systems
- Information seeking behaviour
We then created a scale for each distinction identified during the two-by-two comparison and determined end points. Doing so allowed us to place each participant on the scale and directly compare them. Most variables can be represented as ranges with two ends. It doesn’t matter whether a participant is a 7 or 7.5 on the scale; but what matters is where they appear relative to other participants. The image below provides an example of our 12 scales mapped for each of our 17 participants.
Step 4: Identify Segments
Now that we had all our participants on the scales, we then colour coded each individual to make it easier to identify groupings of participants on each of the scales. We looked for participants who were grouped closely together across multiple variables. Once we found a set of participants clustering across six or eight variables, we saw this as a major behaviour pattern which formed the basis of a persona.
After quite a bit of analysis, we identified 6 major groupings, each identifying an archetype / persona, which we gave a brief description to on paper, outlining the characteristics and identifying their unique attributes.
When carrying out this step, it is important to remember that your groups should:
- explain key differences you’ve observed among participants
- be different enough from each other
- feel like real people
- be described quickly
- cover all users
Step 5: Write the Personas
We were now ready to write up our 5 personas. For each group we added details around the behavioural traits based on the data we had gathered, describing their goals, information seeking behaviours and system usage amongst other things. We also talked about frustrations and pain points as well as listing some personal traits to make them feel more human.
We gave each persona a name and a photo which we felt best suited their narrative. We tried to add parts of participant’s personalities without going overboard as this would make the persona less credible. We kept the detail to one page and based it on a template provided by the Fluid Project wiki. It’s important to keep persona details to one page so they can be referred to quickly during any discussions. Remember that every aspect of the description must be tied back to real data, or else it’s shouldn’t be included in the persona.
Some people prefer to keep their persona details in bullet points, but we felt that a narrative would be far more powerful in conveying each of our persona’s attitudes, needs and problems. We also added a scale to each persona, detailing their behaviour and attitudes, which serves as a visual summary of the narrative and main points. It may be useful to refer to Fluid Persons Format page for example of these templates: http://wiki.fluidproject.org/display/fluid/Persona+Format
Step 6: Review the Personas
Once our personas were written, we reviewed them to ensure they had remained realistic and true to our research data. We felt that 2 personas in particular had more similar behaviours and goals than differences so we merged them into one complete persona. This left us with 4 library personas representing the students and librarians who were interviewed:
- Eve the e-book reader: “I like to find excerpts of books online which sometimes can be enough. It saves me from having to buy or borrow the book.”
- Sandra the search specialist: “In a quick-fire environment like ours we need answers quickly”
- Pete the progressive browser: “Aquabrowser and Classic, it’s like night and day”
- Baadal the search butterfly: “Classic is simple and direct but Aquabrowser’s innovative way of browsing is also good for getting inspiration.”
A full description of the personas can be found on the persona profiles page of our project wiki: https://www.wiki.ed.ac.uk/display/UX2/persona+profiles
Research has shown that a large set of personas can be problematic as the personas all tend to blur together. Ideally, you should have only the minimum number of personas required to illustrate key goals and behaviour patterns, which is what we ended up with. Finally, to ensure we had a polished product, we asked a colleague who was not involved in the persona creation, to review the personas for accuracy in spelling and grammar.
From my experience, I would say that the most difficult step of the process was getting from step 3 (Two by Two comparison) to step 4 (Identify segments). Although we had initially planned to spend 3 days creating our personas, in the end it took us 5+ days. If we were to repeat this exercise, I would allocate adequate time directly after each individual interview to write up detailed notes on the interviewee, detailing their specific goals, behaviours, attitudes and information seeking behaviour, rather than waiting until a later date to review all the notes together, as described in Step 2. In saying this, there are various different approaches which can be taken when creating personas and we would be very interested to learn what other researchers might do with the same data.
In the concluding part of this blog series, “User Research and Persona Creation Part 3: Introducing the personas”, Lorraine will discuss how we plan to keep the personas relevant and current in the future.