Feature
The visible or operational intervention: prompt, timer, default, offer, ranking, delay, support route or disclosure.
Shared method
Reconstruct the mechanism before applying the threshold.
The suite uses a common five-part model, then applies the specialist tests for path, variant or moment.
The five-part functional model
A feature has no fixed legal meaning in isolation. A reminder can help a consumer complete a requested task or become nagging when repeated after refusal. A recommendation can reduce search costs or become exploitative when the system adapts it around a known vulnerability. The assessment therefore reconstructs the operational chain.
Sequence
What precedes and follows it, how often it appears and whether the consumer receives a genuine opportunity to stop or reconsider.
Objective
The stated and operational optimisation target, including task completion, conversion, spend, retention, disclosure or time.
Adaptation
Whether the system changes timing, content, friction, ranking, price or support according to user, context or feedback.
Outcome
The material consumer effect, rights friction, vulnerability exploitation, harm or absence of legally significant impairment.
Core legal discipline
The five factors operate as evidential structure, not as five universal cumulative legal elements. A static and obstructive cancellation practice may fail without personal adaptation. An effects-based rule may apply where subjective intent cannot be proved. Conversely, an internal objective to increase engagement or retention does not prove unlawful manipulation by itself.
The method centres two questions:
- Material impairment: Did the design materially impair free and informed choice, obstruct a protected route or exploit a defined vulnerability event?
- Causal connection: What feature, sequence, objective or adaptation produced or materially contributed to that impairment?
Evidence hierarchy
| Evidence level | Examples | What it can establish |
|---|---|---|
| Visible artefact | Screenshot, wording, button hierarchy, timer, default | What appeared in one observed state |
| Journey record | Screen sequence, waiting, transfers, repeated prompts, completion | How the consumer moved through the route |
| Variant record | Feature flag, experiment allocation, model score, trigger rule | Why the observed experience differed |
| Objective and outcome record | Success metric, cohort effect, abandonment, spend, retention, complaints | What the architecture sought or produced |
| Governance evidence | Constraints, approvals, red-team tests, safeguards, review and rollback | Whether the organisation anticipated and controlled the risk |
Product-specific tests
Matched-pathway test
Compare entry and protective routes, classify each added burden, test justification and record the parity delta.
Protected-flow invariance
Test whether rights, redress and material information remain meaningfully available across conditions and variants.
Person-system-context test
Identify what the system knew, how it adapted, whether agency survived and how benefit and risk were allocated.
Outcome design
Each product uses five reasoned outcomes. The products do not generate one portfolio-wide compliance score. A route-level obstruction, selective disadvantage or exploitative adaptation requires a visible conclusion and cannot be averaged away.
Lifecycle and reassessment
Each conclusion attaches to a versioned service, defined journey or use case, evidence date and scope. A material change may require reassessment, including a new objective, data source, trigger rule, variant, population, market, support route, evidence gap, control failure or observed harm.
Technical implementation
- Static browser application with no necessary server-side assessment processing.
- Product questions and conditional logic held in versioned JSON schemas.
- Deterministic rules for outcome routing and visible explanations for each trigger.
- IndexedDB for optional local save and evidence attachments.
- JSON import and export, optional browser-side project encryption and print-ready PDF.
- SVG journey, decision-surface and person-system-context maps.
- No generative-AI dependency in the core conclusion.
Method status.
The framework remains a product methodology in development. It requires controlled piloting, legal review, security review, accessibility testing and usability research before commercial reliance.