Capablio Adaptive Feedback Framework

Purpose

This pack describes a more robust Capablio framework for continuous, multi-rater feedback and personal development. It is designed around five requirements:

1. Question packs that can run daily, weekly, monthly, quarterly, annually, or on events.
2. Per-question re-ask cooldowns expressed as a number of cadence periods to skip.
3. PDP-driven overrides that may shorten a cooldown or force a question into the next eligible survey.
4. Constrained rotation of questions and bundles so surveys stay short while eligible content eventually reaches full coverage.
5. Tenant-specific questions, bundles, policies, schedules, and governance.

Thomas 360 is treated as an importable external baseline, not as the governing data model. The framework preserves Thomas-compatible concepts such as multi-rater relationships, a 1–7 scale, Not Observed, competency grouping, comments, and PDP themes, while supporting a much richer longitudinal model.

Core product decision

The stable measurement unit is not an exact sentence. It is a Question Family attached to a stable Construct. A family can have variants for different cadences, observation windows, roles, respondent relationships, languages, and scales.

Example:

• Construct: feedback loop closure
• Weekly direct-report variant: “This week, my manager explained what changed after receiving feedback.”
• Quarterly peer variant: “During the past quarter, this leader consistently closed the loop after receiving feedback.”
• Annual self variant: “Over the past year, I regularly communicated what I changed after receiving feedback.”

This preserves conceptual continuity without forcing a daily question to use annual-review wording.

Important terminology

• Cadence: how often a survey window is created.
• Exposure cooldown: how many complete cadence periods must be skipped before the same question can be presented again.
• Data retention: how long response data is stored. This is a privacy setting and must never share the same field or implementation as exposure cooldown.
• Subject seat: a person whose development and longitudinal record are managed. Respondents and external raters should not consume paid seats.
• Anchor item: a stable item repeated on a predictable cadence for trend continuity.
• Rotating item: an eligible item selected by the coverage engine.
• PDP item: an item selected or forced because it is linked to an active development objective.

Recommended survey composition

A normal recurring survey should use a constrained mix rather than pure randomness:

• 20–30% anchors for trend continuity
• 40–60% rotating items for coverage
• 10–30% PDP or event-triggered items
• Optional open text, normally no more than one prompt

Daily surveys should usually contain one or two state questions. Weekly surveys should usually contain three to five questions. Leadership 360 questions should not be asked daily.

Pack map

How to use this pack

1. Review 00, 03, 05, 06, and 07 first; they contain the main architectural decisions.
2. Run a product workshop against 16_open_questions.md.
3. Convert the selected Markdown schema to JSON Schema or Protocol Buffers before implementation.
4. Pilot packs with real organizations before describing them as validated or benchmarkable.
5. Keep commercial entitlements separate from software release flags and tenant policy settings.

Design warning

“Eventually achieve 100% question coverage” is only possible when the available survey slots are sufficient for the eligible question and bundle inventory. The platform must calculate feasibility and show the projected completion period. It must not promise 100% coverage under an impossible quota.