Self-report
Attitudes, intentions, confidence and money-related preferences provide context. They are useful signals, but they can be affected by memory, interpretation and social expectations.
How the assessment works
Evidence-informed.
Continuously calibrated.
A transparent account of what we measure, what the results mean, and where we draw the line. The assessment combines different kinds of evidence because no single question can explain an investor.
Three sources of signal
Not a personality test.
Not observed trading behaviour.
We combine what people believe about themselves, what they know, and how they respond to the same hypothetical decisions. Each contributes something different.
Knowledge checks
Scored questions show what a respondent demonstrates on the concepts tested. They do not establish complete investment competence. Comparing related confidence and knowledge responses can identify a question worth investigating.
Decision scenarios
Standardised hypothetical situations ask respondents to make choices under uncertainty. They add behavioural context, but they are not presented as observed transactions.
Scoring and context
Not every answer
does the same job.
The assessment deliberately separates inputs that influence classification from inputs that help explain a person’s situation. This keeps a richer report from becoming an opaque score.
Seventeen scoring inputs
A defined subset of answers contributes to competence and ambiguity-tolerance scores. It combines scored knowledge answers, self-report items and four standardised hypothetical scenarios. The additional scenario weight is a design choice in the current full-assessment configuration, not a demonstrated superiority over self-report. The shorter event version uses a subset of questions.
The rest describes the journey
Background, progress, barriers, decision style and money-relationship answers enrich the individual and cohort view. They do not quietly move someone into a different profile unless they are explicitly part of the scoring set.
The distinction is operational, not cosmetic.It lets a project team trace a profile back to its scoring inputs while still using the broader survey to understand where people are, what may be getting in the way and which questions deserve investigation.
The profiling model
Two dimensions.
Nine useful starting points.
The model locates a response pattern across investment competence and tolerance of ambiguity. The profile is a lens for explanation and engagement, not a permanent label.
Lower ambiguity tolerance
Medium ambiguity tolerance
Higher ambiguity tolerance
Higher competence
Commitment-PhobeKnows the territory; hesitates under uncertainty.
Healthy RelationshipCombines capability with considered action.
One That Got AwayCapable and comfortable, but may need a clearer next step.
Medium competence
OverthinkerAnalysis can become delay.
SituationshipEngaged, but not yet committed.
Adrenaline JunkieComfort can run ahead of understanding.
Lower competence
GhosterLow confidence and uncertainty can prompt withdrawal.
Slow BurnInterest grows when the path feels manageable.
Love BomberEnthusiasm may arrive before foundations.
Names make the profiles memorable; the underlying dimensions do the analytical work.The categories support communication and intervention design. They should not be used to stereotype a person or infer characteristics the assessment did not measure.
Boundaries are treated as boundaries.When a score is close to an adjacent band, the method retains that proximity as a cusp signal. This avoids presenting a narrow scoring difference as a categorical truth about a person.
Calibration gaps
Sometimes the gap is the useful signal.
A single score can hide contradictions. We compare related answers to surface two possible calibration gaps.
Confidence versus demonstrated knowledge
Related responses can point to a possible mismatch. Check the concepts, wording and scales before interpreting it: subtracting arbitrary confidence and knowledge scores does not establish calibration. Ask what remains uncertain before proposing reassurance.
Stated comfort versus scenario choices
We compare general statements about comfort with the choices made in standardised situations. This is a comparison of survey signals, not proof of future action.
From response to action
A traceable chain,
not a leap to a recommendation.
Individual classification and cohort diagnosis are related, but they are not the same output. A client project follows a defined path from an agreed question to a measurable intervention.
1. Define the decision
Agree the population, the journey moment and the business question before fieldwork begins. The comparison should be chosen because it can inform an action.
2. Collect comparable signals
Participants receive the same structured assessment. Individual outputs remain private unless a different, explicit data arrangement has been agreed.
3. Produce reproducible profiles
Fixed answer mappings, weights and banding rules produce the two dimension scores, profile, calibration gaps and any nearby boundary signal.
4. Examine cohort patterns
Appropriately aggregated results are compared across relevant cohorts. The diagnostic looks for profile skews, confidence-versus-knowledge patterns and self-reported journey stages. A reported stage is not an observed drop-off. Reporting rules must address small groups and identifying combinations, not only the overall sample size.
5. Generate candidate explanations
Detected patterns are matched to a curated hypothesis library. A match identifies an explanation worth testing. It does not establish why the pattern occurred.
6. Test and review
The team selects an intervention, target segment, outcome measure and review point. What happens next is used to refine the hypothesis and inform later calibration.
Evidence language
Every statement should show
what kind of claim it is.
The method keeps observations, interpretations and outcomes separate. That distinction matters when behavioural evidence informs product, communication or customer decisions.
Response or cohort fact
What respondents selected, what the rules calculated, or how two defined cohorts differed.
Pattern or candidate hypothesis
A reproducibly detected pattern and one or more explanations that are plausible enough to investigate.
Intervention outcome
What changed after an intervention, reported only against the measure and review point agreed for that project.
No result is promoted up this ladder by better prose.Optional AI can explain an observed fact or approved hypothesis. It cannot turn a candidate explanation into a finding or a proposed intervention into a measured outcome.
Deterministic first
The facts do not move
when the prose changes.
The same answers, items and scoring version produce the same calculated result. Reproducibility is a software property; it does not establish measurement or predictive validity. AI-assisted explanations must remain tied to approved facts and be checked for factual accuracy.
Scores and classifications
Weights, banding and comparison rules are applied consistently and can be audited.
Candidate hypotheses
Cohort patterns are matched to a maintained library of explanations worth investigating, not declared causes.
Narrative explanation
AI may structure and explain approved facts. Fixed rules calculate the result; explanations require factual review. The assessment is not a source of investment recommendations.
Project outputs
Useful to the participant.
Actionable for the team.
The same assessment supports different levels of output without blurring individual and cohort evidence.
Private participant result
A profile, the two underlying dimensions, calibration gaps and a practical next step, expressed without treating the person as a fixed type.
Cohort diagnostic with reporting safeguards
Profile distribution, self-reported journey patterns, relevant comparisons and explicit evidence limits. Aggregation alone does not ensure anonymity: review small cells, combinations and free text before sharing.
Intervention brief
A target segment, candidate hypothesis, proposed change, success measure and review point that the client team can put into practice.
Continuous calibration
Review is part of the method,
not a one-time milestone.
The review process examines item performance, interpretation and potential changes to weights or thresholds. Each project needs an identified scoring version and an agreed review point. Changes should follow documented evidence, not silently alter earlier results.
Project review
We examine completion, item response, profile distribution, edge cases and how the outputs were understood in that project’s context.
Cross-project review
At regular intervals, pooled and appropriately governed aggregate evidence is used to check whether weights, thresholds or language need revision.
Versioned change
Material changes should be documented and tested. Continuous calibration means repeated evidence review, not an opaque system changing itself in real time.
Appropriate use
A behavioural diagnostic, not investment advice.
u impact does not replace a regulated suitability assessment. The assessment does not by itself establish a person’s financial situation, capacity for loss, complete investment objectives or the suitability of a financial instrument. It supports research, segmentation, communication and testable engagement design.
Limit collection
Use only data needed for an agreed purpose and define retention, access and reporting before a project begins.
Review fairness
Check outcomes by relevant cohorts, investigate unexpected skews and avoid turning correlations into assumptions about individuals.
Test interventions
Measure whether a proposed change helps. A plausible hypothesis is the beginning of a test, not the end of the analysis.
Evidence base
Research informs the model. It does not remove the need to validate each claim.
U Impact’s earlier theory-building work explored barriers and sustainable-investor types. The current nine-profile model is a later applied framework. The implementation repository documents its questions and scoring rules; it does not provide independent validation of the taxonomy, thresholds or intervention effects.
External research supports distinguishing knowledge, confidence and different forms of uncertainty. It does not validate U Impact’s own questions, weights or profiles. Evidence of reliability, construct validity, predictive performance and fairness must be established for the intended population and use. Read the referenced research briefs.
Put the method against a real question.
Start with a defined cohort, a decision you need to improve and a measurement plan.