Analytics, Scoring, and Reporting

Reporting principles

1. Show evidence before conclusions.
2. Show observation horizon and source.
3. Preserve relationship differences when anonymity allows.
4. Do not imply that one decimal place is meaningful without sufficient reliability.
5. Keep incompatible scales and question variants separate unless a bridge is approved.
6. Treat Not Observed as information about exposure, not as a low score.
7. Distinguish development feedback from formal performance evaluation.

Raw response model

Every response stores:

• tenant, survey, subject, respondent pseudonym, and relationship
• question family, variant, and frozen survey item
• answer value and source scale
• Not Observed / Not Applicable / declined reason
• optional confidence
• optional narrative evidence
• timestamps and channel
• consent and purpose context

Raw responses are immutable. Corrections create a new event and preserve the audit trail.

Aggregation levels

• question variant
• question family
• construct
• pack or framework dimension
• respondent relationship
• individual
• team
• organization and cohort
• time window

Minimum reporting data

Each score should carry:

• valid response count
• invited and completed counts
• Not Observed percentage
• response distribution or dispersion
• relationship composition
• observation horizon
• last updated date
• source: native, tenant, or imported

Trend rules

Exact trend

Allowed when the same question family, compatible variant, scale, relationship, and observation horizon are used.

Bridged trend

Allowed only when a framework approver has accepted a bridge between variants or scales. The report must identify the break.

Context-only comparison

Used for imported Thomas data, materially changed questions, or unlike evidence types. Show side by side, not as a continuous line.

Normalization

A 0–100 transform can help visualization:

normalized = (value - scale_min) / (scale_max - scale_min) * 100

This does not prove that a 4 on a 5-point frequency scale is equivalent to a 6 on a 7-point extent scale. Native aggregation should remain within compatible score families.

360 views

Individual summary

• strengths with sufficient observations
• development opportunities
• self-versus-others differences
• relationship differences
• Not Observed patterns
• recent trend and change confidence
• qualitative themes with concrete examples
• selected PDP goals

Blind-spot view

A “blind spot” should require more than a numeric self/others gap. Consider:

• gap magnitude
• sample size
• dispersion
• relationship consistency
• observation sufficiency
• role importance
• repeatability across cycles

Role-priority view

Show:

• self importance
• manager importance
• other-group importance
• expected impact in the next 6–12 months
• current evidence
• priority-development matrix

Anonymity and suppression

Default group threshold: three valid respondents. A tenant may require four or five.

When a group is below threshold:

• combine with a compatible group only under a predefined policy
• delay the report until more data exists
• suppress the group result
• never infer an individual response from changes in an aggregate

Manager and self responses may be shown separately only when participants were told this in advance.

Coverage analytics

Framework owners need:

• percentage of eligible families covered in current cycle
• coverage by construct and relationship
• projected completion period
• starving and repeatedly deferred questions
• impossible atomic bundles
• anchor versus rotation share
• coverage lost through Not Applicable or unavailable variants

Fatigue analytics

• questions presented per respondent over 7, 30, 90, and 365 days
• average completion time
• abandonment and skip rates
• repeated-construct concentration
• survey overlap incidents
• reminder volume
• decline and opt-out rate

Item-quality analytics

• response rate
• Not Observed rate
• Not Applicable rate
• variance and ceiling/floor effects
• comment yield
• relationship applicability
• semantic overlap with other items
• internal consistency where a valid multi-item scale exists
• stability and sensitivity to change
• translation differences

Items with poor quality should enter review, not be silently reworded in place.

PDP reporting

Show separate evidence lanes:

• planned actions
• completed actions
• self reflection
• observer evidence
• blockers and support
• relevant outcome signal
• time sustained

A simple status can summarize these, but the underlying evidence must remain inspectable.

Organization dashboard

Recommended sections:

• participation and survey health
• coverage and fatigue
• construct trends
• manager and leadership capability patterns
• team differences with suppression
• PDP activation, progress, and closure
• top recurring blockers
• action-plan ownership and closure
• item-quality alerts

Avoid individual leaderboards and simplistic ranking of teams unless the measurement program was explicitly designed and validated for that purpose.