Pat’s Point of View

I’ve recently been helping a client with developing personas for their website audience. During the process I have created a couple of ideas that have helped.

First, an idea to help illustrate the concept of taking explicit user research and shaping this into discrete personas. Most people I have mentored find that the most difficult step is going from research into producing the draft personas. My idea is to use Venn diagrams to show how you take the information you find out about each individual you research (through interviews, focus groups, contextual enquiry, customer data etc) and come up with an archetypal description of them all. Remember, the key is for the resulting persona to have attributes that are a common subset of the attributes possessed by all the users it represents, whilst still being accurate for each of them. So in effect you need to ‘knock the corners’ off the data and look at the overlap, to make something that fits all users in the group.

The Venn diagram shows multiple overlapping circles, each of which represents the information we know about an individual. The intersection of the circles represents the commonality between them. A large proportional area of overlap (top of diagram on right) would indicate a good match or ‘good fit’ and thus tells you that these users would likely be well represented by the same persona. Conversely, a small area of overlap (bottom diagram on right) would suggest the users aren’t very similar and thus probably wouldn’t be well served by the one same persona.

Now, this is simply a method of illustrating the concept behind the persona creation process. I can say with confidence is that the people with whom I have used this method of communication (and they have had varying degrees of expertise in user-centred design) have found it very useful in terms of explaining what we were trying to do. What if there was a way of plotting user attributes collected through research, such that you could actually use Venn diagrams to group users together and work out your personas?

Well I have an idea for this too. A popular method of illustrating comparitive data in an easily digestible way is to plot it on a simple two dimensional graph. Product marketplace maps are a good example, as are Jessica Hagy’s examples on Indexed. And Steve Mulder talks about doing this in his persona book The User is Always Right (chapter 6) for pretty much the same purpose. The trouble is that your users will almost always have many more dimensions that you want to consider, than just the two that could be plotted on these type of graphs. Of course, you can extend to a third dimension but it gets difficult to produce and can be difficult to visualise and comprehend.

Exploring this further, you could move from Venn diagrams to a radar graph which allows you to plot more than just a X and Y axis. Plotting a score for each dimension for each user would produce a pattern that represents that user (diagrams on left). Overlaying several such graphs would allow you to look for overlaps, as with the Venn diagrams. The visual patterns help highlight where a persona could be found. But I suspect this might become impractical to produce such graphs with more than 4 or 5 dimensions (labeled A-E here).

So, my idea is to use a graph based on the slider controls of a stereo equaliser (’EQ’) such that any user’s dimensions, or at least the ones we think are important, can be plotted in one simple chart as shown by the vertical lines in the diagram on the right. Again we have two extremes for each dimension and a number of intervening points (having an infinite scale would make things too difficult). As with other graphs, you assess where the user would ’sit’ on each dimension, effectively giving them a score out of five, for example. Dimension A might represent the user’s level of computer literacy, ranging from low to high. A ’super user’ would be plotted nearer the top of the dimension, whereas a ’soccer mom’ might be plotted nearer the bottom. Doing this for each dimension (A-H) allows you to illustrate that user on the EQ map or chart.

So it follows that you would create an EQ map for each user who participated in the research. It shouldn’t be that big a leap to see how you could combine several such maps in one diagram (as shown in the paragraph above on the right) which would allow you to recognise patterns in the maps. Similarities in the lines between the dots would suggest a persona could make sense for the users involved. A line of best fit could also be produced which would then form an EQ map for that persona.

So, various squiggly lines can help you find possible personas in the data you have collected through research. I will post further findings as I continue to use and refine this graphing technique. Please do give me feedback, especially if you use an EQ map.

NB: As I alluded to in my last post, during Todd Warfel’s tutorial on data-driven persona development, I realised these ideas weren’t really that new. Todd demonstrated a similar approach using a box plot, which is also used by various other people for the same purpose. Whilst I did some up with the above completely independently—and I think the EQ map has some advantages—I guess it goes to show a thorough literature search is a good idea.

I guess it also goes to show that if you have a good idea, blog it as soon as possible. And it’s nice to think that smart people agree with me :)

See what people are blogging about: personas, qualitative segmentation, user research, data analysis

Popularity: 54% [?]