Skip to content

Feedback, Roadmap, and Customer Commitments

Draft 0.1 — 2 August 2026
Feedback is an input to product judgment, not an automatic vote. A roadmap is a planning artifact, not a binding delivery promise unless a separate authorized agreement says otherwise.

ANUKA products should make improvement participation real without turning every comment into a ticket, every vote into a mandate, or every roadmap card into a legal promise.

The protocol separates:

  • signal;
  • evidence;
  • request;
  • opportunity;
  • experiment;
  • roadmap intent;
  • funded commitment;
  • contractual commitment;
  • delivered outcome.

Products may collect feedback through:

  • in-product forms;
  • browser extension;
  • public Backstage;
  • support conversations;
  • interviews;
  • surveys;
  • issue trackers;
  • community forums;
  • sales and customer-success records;
  • contributor proposals;
  • agent-assisted analysis;
  • operational telemetry.

Every channel declares its audience, privacy level, retention, and whether submissions may become public.

A feedback record contains:

  • feedback ID;
  • product;
  • submitter or pseudonymous reference;
  • source channel;
  • audience;
  • content;
  • category;
  • affected journey;
  • severity or importance as stated by submitter;
  • evidence attachments;
  • consent for publication;
  • created date;
  • status;
  • links to duplicates, opportunities, experiments, or decisions.

The record distinguishes the submitter’s statement from product-team interpretation.

  • private to product team;
  • visible to organization members;
  • visible to selected collaborators;
  • community-visible;
  • public with attribution;
  • public without direct attribution where policy permits.

The audience is selected before submission and can be changed only through an authorized process.

Reports involving security, privacy, abuse, discrimination, harassment, payment disputes, personal data, or legal claims must route to appropriate controlled workflows.

They must not be automatically published or summarized into public AI outputs.

received → triaged → investigating → linked → decided → closed
↘ needs_information
↘ restricted_case
↘ duplicate

Closed does not necessarily mean implemented.

Triage records:

  • product area;
  • urgency;
  • potential harm;
  • affected users;
  • reproducibility;
  • available evidence;
  • security/privacy/legal routing;
  • duplicate group;
  • accountable owner;
  • next action.

Triage must not silently change the original feedback text.

Related feedback may form a Signal Group.

A Signal Group contains:

  • normalized problem statement;
  • linked feedback records;
  • source diversity;
  • affected cohorts;
  • frequency estimate;
  • severity estimate;
  • current evidence strength;
  • known bias or sampling limitations;
  • owner;
  • decision status.

Volume is informative but does not alone determine priority.

Signals should be evaluated across:

  • frequency;
  • severity;
  • strategic alignment;
  • user diversity;
  • revenue or cost impact;
  • accessibility impact;
  • security and privacy impact;
  • legal obligation;
  • evidence quality;
  • reversibility;
  • opportunity cost;
  • network reuse.

A loud customer is not automatically a representative customer.

Products should preserve the user’s requested solution while also identifying the underlying problem.

Example:

  • request: “Add CSV export.”
  • possible problem: “I need to reconcile transactions in my accounting workflow.”

The decision record may choose a different solution while acknowledging the request.

A validated or emerging user problem without a committed solution.

Research, prototype, or experiment intended to reduce uncertainty.

A planned product change with scope and owner.

Work required for security, accessibility, reliability, compatibility, or technical health.

Migration or implementation of a shared ANUKA contract.

A separately authorized promise with defined terms.

A feature-funding campaign governed by the Economy Protocol.

considering → discovery → planned → in_progress → released → measured
↘ not_planned
↘ paused
↘ superseded

Public roadmap labels must explain uncertainty.

The product confirms receipt or awareness.

No investigation or delivery promise.

The product commits to evaluating the problem.

The product currently intends to address the problem, subject to review and change.

Budget, owner, scope, and acceptance conditions are approved.

A separate authorized contract or order defines delivery obligations, remedies, and parties.

Only C3 and C4 should be described as commitments, and C4 must not be inferred from a roadmap label.

Every material roadmap decision includes:

  • decision ID;
  • problem and evidence;
  • options considered;
  • strategic context;
  • affected users;
  • accessibility, security, privacy, and operational implications;
  • estimated cost and dependencies;
  • funding status;
  • authority;
  • decision;
  • commitment level;
  • review date;
  • public explanation;
  • appeal or resubmission path where applicable.

A product may use structured methods, but no formula should replace judgment.

A prioritization profile may include:

  • user impact;
  • urgency;
  • evidence confidence;
  • reach;
  • strategic fit;
  • revenue or cost;
  • risk reduction;
  • accessibility benefit;
  • network reuse;
  • effort;
  • reversibility;
  • dependency readiness.

The exact weights and limitations should be visible to decision makers.

Votes are preference signals, not identity, payment, authority, or consent.

Controls may include:

  • one vote per eligible account;
  • organization-weighted views separated from individual views;
  • fraud and duplicate detection;
  • declared cohort filters;
  • visibility of sample size;
  • no implication that the top-voted item will ship.

Token ownership must not determine access to basic feedback rights.

When a signal becomes a funding campaign, the record must link to:

  • campaign terms;
  • funding mode;
  • refund rules;
  • delivery scope;
  • acceptance criteria;
  • ownership and license;
  • target dates;
  • risk and dependencies;
  • current commitment level;
  • sponsor communications.

A public vote and a paid sponsorship are different acts.

Enterprise or strategic commitments require:

  • authorized parties;
  • scope;
  • dependencies;
  • timeline;
  • acceptance;
  • change control;
  • pricing;
  • support;
  • remedies;
  • data and security terms;
  • productization or exclusivity rules;
  • roadmap disclosure policy.

Sales notes or chat messages must not silently create product obligations.

Accessibility reports receive:

  • direct acknowledgement;
  • barrier classification;
  • affected assistive technology or interaction;
  • severity;
  • workaround where safe;
  • remediation owner;
  • target state;
  • testing with affected users where practical.

Accessibility requests should not compete solely on vote count.

Potential vulnerabilities route to the vulnerability disclosure process.

Public roadmap systems must not expose exploit details before remediation and coordinated disclosure.

Agents may:

  • classify themes;
  • suggest duplicate groups;
  • summarize selected feedback;
  • extract candidate problems;
  • identify missing evidence;
  • draft response options.

Agents must not:

  • publish private feedback without permission;
  • infer sensitive traits without approved purpose;
  • convert sentiment into a universal user score;
  • silently delete minority reports;
  • make final high-impact roadmap decisions;
  • invent supporting quotes or evidence.

Human owners review material summaries.

Summaries should retain links to source records and indicate:

  • number of records;
  • date range;
  • channels;
  • cohort limitations;
  • language coverage;
  • excluded sensitive cases;
  • whether the summary was AI-assisted;
  • reviewer.

Products should publish realistic expectations such as:

  • security reports: immediate controlled intake;
  • payment or access failures: support severity process;
  • accessibility barriers: acknowledgement and triage target;
  • ordinary feedback: acknowledgement without guaranteed individual response;
  • roadmap proposals: periodic review cadence.

Do not promise individual responses when the product cannot sustain them.

A good explanation states:

  • what was heard;
  • what evidence mattered;
  • what decision was made;
  • what was not decided;
  • important constraints;
  • current commitment level;
  • how the decision may be revisited.

A “not planned” decision can be respectful and useful when it is honest.

Closure reasons include:

  • delivered;
  • experiment completed;
  • solved differently;
  • duplicate;
  • unsupported use case;
  • insufficient evidence;
  • not aligned;
  • unacceptable risk;
  • legal or policy restriction;
  • product deprecated;
  • submitter withdrew;
  • superseded.

Closure does not erase the historical record.

When roadmap intent changes, affected sponsors or contracted customers receive communication according to their terms.

Public roadmap changes should include:

  • changed state;
  • reason;
  • impact;
  • replacement plan if any;
  • refund or contract process where applicable;
  • review date.

Participants should be able to export feedback they submitted, subject to third-party privacy and legal restrictions.

Products exiting ANUKA should preserve or transfer feedback according to declared policy rather than abandoning it in an inaccessible system.

Controls include:

  • spam filtering;
  • rate limiting;
  • coordinated-vote detection;
  • disclosure of paid or affiliated advocacy where relevant;
  • protection against retaliation;
  • moderation and appeal;
  • separation of security reports from public debate;
  • no purchase of hidden priority.

Commercial influence must be explicit.

Useful measures include:

  • acknowledgement time;
  • triage time;
  • unresolved high-severity signals;
  • percentage linked to decisions;
  • accessibility resolution time;
  • roadmap decision age;
  • sponsor refund rate;
  • reopened issues;
  • feedback source diversity;
  • percentage of public commitments with current status.

Avoid optimizing for total feedback volume alone.

  • calling votes democracy;
  • promising every requested feature;
  • using a roadmap as a sales contract;
  • deleting rejected feedback;
  • publishing private feedback by default;
  • prioritizing accessibility only by popularity;
  • hiding sponsored influence;
  • using AI summaries without source links;
  • changing commitment levels silently;
  • rewarding teams for closing feedback records rather than resolving problems.
  • feedback objects have audience and publication consent;
  • sensitive cases route to controlled workflows;
  • signal groups preserve source links;
  • roadmap states and commitment levels are distinct;
  • every material decision has an owner and explanation;
  • votes are labeled as signals;
  • sponsored features link to campaign terms;
  • accessibility and security feedback have dedicated routes;
  • AI summaries disclose provenance and review;
  • public commitments show current status;
  • feedback export and exit behavior are defined.
  • Protocol review date: 2 August 2026
  • Public roadmap as contract: Prohibited unless separately authorized
  • Paid influence disclosure: Required
  • Human roadmap authority: Required