Personas are not fictional user types. They are evidence-based demand clusters with spatial consequences. The distinction is the method: marketing-style personas ('Meet Sarah, a collaborative millennial') import assumptions and invite stereotype; demand clusters are derived from measured work patterns, activity profiles, mobility and environmental needs, and each carries quantified spatial consequences. Personas make a workforce of thousands designable — but only if they compress real variance rather than replace it with fiction.
- In Define, once the Dynamics Assessment provides cluster structure and the demand position is aligned
- As input to Taxonomy Alignment where population differentiation needs a governable structure
- For Workshop III and IV, where options must be tested against differentiated demand rather than an average user
- In change communication, where clusters make consequences explainable per population
- Strategist and analyst deriving clusters from the data
- Community group stress-testing clusters against ground truth — do these segments describe real colleagues?
- HR validating cluster structure against workforce composition
- Team leads confirming their populations' cluster assignments
- Demand-cluster structure from the Dynamics Assessment
- Activity profiles per population from Schedule and Activities
- Mobility and presence patterns from Balance
- Aggregate environmental and accessibility patterns from the Workplace Passport, where fielded
- Function profiles from Business Needs Interviews
- Derive clusters analytically from the demand data — work patterns, activity proportions, mobility, presence rhythm, environmental needs — rather than assembling them from job titles or narrative intuition.
- Test cluster validity: internal coherence (members genuinely resemble each other), external distinction (clusters genuinely differ on spatially consequential dimensions), and coverage (the structure accounts for the workforce without a large residual).
- Attach quantified spatial consequences per cluster: setting demand, sharing tolerance, presence rhythm, enclosure and environmental requirements, technology needs. A cluster without spatial consequences is decoration.
- Name clusters by demand signature ('anchored focus-intensive', 'distributed collaboration-led'), never by demographic or personality framing.
- Stress-test with the community group: real employees should recognise themselves without feeling typecast.
- Map cluster distribution across teams and sites — clusters cut across org charts, and the mapping is where the design value lives.
- Document the boundary honestly: clusters compress variance, they do not eliminate it; individual needs beyond the cluster remain individually valid.
- Clusters that mirror the org chart — usually a sign the analysis recycled structure instead of finding behaviour
- Demographic contamination: age or generation framing masquerading as demand segmentation
- A residual population too large to ignore, quietly dropped from the story
- Clusters differing on dimensions with no spatial consequence — statistically real, strategically useless
- The persona hardening into identity: people managed as their cluster rather than served by it
- Validated demand-cluster set with quantified spatial consequences
- Cluster distribution map across teams and sites
- Cluster-to-setting demand matrix for Taxonomy Alignment
- Communication-ready cluster descriptions for change work
The cluster-to-setting matrix differentiates the Taxonomy Alignment demand model. Workshops III and IV test options against clusters rather than averages. The Kit of Parts derives setting proportions partly from cluster distribution. In Deliver, clusters structure change communication; in Recalibrate, cluster-level behaviour is re-measured to detect demand drift.
Cluster derivation is statistical; cluster judgement is not. Deciding how many clusters the organisation can govern, when a statistically valid segment is strategically trivial, and where the compression of variance becomes distortion — these are analytical judgements. So is resisting the organisational appetite for memorable characters over defensible segments.
- Inventing narrative personas and retrofitting data to them
- Naming clusters by demographics or personality types
- Producing more clusters than the design process can differentiate for
- Presenting clusters without spatial consequences
- Treating cluster membership as fixed individual identity rather than current demand pattern
A government agency's clustering produced five demand segments; the largest ('rhythm-bound processing', 34%) cut across eleven departments and carried the highest enclosure requirement — invisible in any departmental analysis. The initial concept had provided enclosure by department seniority. The cluster evidence redirected enclosure to demand, and the community group's stress-test renamed one cluster whose draft label members found reductive: a small correction that preserved the credibility of the whole structure.