Description
Code interviews into behavioural traits, link participants by how many traits they share, and merge highly connected participants into labelled persona segments.
This method derives personas bottom-up from coded qualitative data rather than from researcher intuition. In the original study, survey and interview data from 32 participants were open-coded by two coders and affinity-diagrammed into 85 traits under 24 concepts. Each participant becomes a binary vector over traits; the similarity of two participants is the dot product (number of shared traits), used as an edge weight in a graph. Clusters grow agglomeratively from highly connected participants and merge until stable with no singletons. Traits shared by too many clusters are removed and the graph reclustered so that clusters differ meaningfully. Participants are then rated on two theme attributes, here knowledge and motivation, each cluster takes its modal rating, and clusters are placed and named in that two-dimensional space, giving five segments.
Attributes
- Discipline
- Qualitative Analysis & Synthesis
- Loss type
- Siloed
- Goal
- Segment users, Synthesise data
- Stage
- Discovery, Definition
- Data
- Mixed
- Effort
- Weeks
Steps
- Collect a survey and a semi-structured interview from each participant.
- Transcribe, open-code with two coders and resolve conflicts; add survey answers to the coded data.
- Affinity-diagram codes into low-level traits and higher concepts.
- Represent each participant as a binary vector over traits and build a graph with edge weight equal to the number of shared traits.
- Cluster agglomeratively from high-connectivity participants, merging until stable with no singletons; drop traits shared by too many clusters and recluster.
- Rate each participant on two theme attributes, give each cluster its modal rating, plot clusters in that space and label them descriptively.
Sources
- Privacy Personas (Dupree et al., 2016)
Also known as
- Trait-overlap graph clustering of coded interview data
- Shared-trait participant graph clustering
- Bottom-up behavioural segmentation from affinity traits
- Privacy personas method
Categories
- User Research
- Qualitative Analysis & Synthesis