From personas to archetypes: a living, shared understanding of the people we build for (Part 1 of 2)

Key points

Canonical's design team has published the first of two posts on "archetypes," a shared, machine-readable set of user descriptions meant to replace one-off personas across its product portfolio. This part covers what the system is for and how it is structured. A follow-up will cover how it is maintained and updated.

Why Canonical changed approach

• Canonical builds more than 60 technical products for users ranging from hobbyists running home servers to CTOs managing large fleets. Teams knew their own users but had no clear way to share that knowledge.
• Traditional personas tended to be made once, left un-updated and become inaccurate within about a year, after which other teams stopped trusting them.
• Personas generated by a large language model can sound plausible but lack evidence and lean on stereotypes, reflecting a generic average rather than Canonical's actual operators.
• The team's conclusion is that user descriptions should be maintained like a product, backed by current, sourced evidence, rather than produced as finished documents.

What an archetype is

• An archetype is a user type defined by behaviour and interests across the whole portfolio, covering both product design and brand messaging. It is distinct from marketing "brand archetypes."
• Archetypes leave out demographic detail. They describe how a user evaluates a new tool, what breaks their trust and what a bad day looks like.
• Each claim links to a source, such as an interview, a support thread or a pattern seen across many conversations, so it can be checked.
• Archetypes are organised by customer-journey stage: day 0 evaluation, day 1 deployment and day 2 operations. Several teams serving the same kind of user can then work from one file.

Two tiers

• The broad, portfolio-wide archetypes are maintained centrally by Canonical's User Science team.
• Individual teams build "contextual personas" for a specific product, feature or launch. These are named, detailed and scoped to one use case, and the team using them owns them.
• Every contextual persona starts from an archetype and inherits its evidence. New findings from team work are fed back into the archetype.
• Contextual personas are kept in a single repository, so teams can draw on each other's work.

Effects and status

• Because verified archetypes are structured files, teams can point their AI agents at them instead of prompting an untuned model.
• Agent-driven experiments are also used to test archetypes against reality, for example by using one to target a marketing campaign and measuring how it performs.
• The work is in progress. Inside Canonical, the archetypes and the process for making them are open to the whole organisation, following a changelog-style approach, so anyone can trace, dispute or extend a claim.
• The post closes with a recruiting note for the Canonical design team, pointing to canonical.com/careers.

Summary by DistroNews, based on the original post.