mh-cto-sdlc-user-feedback

User Feedback Collection

Output is insight, not data. The failure mode of feedback collection is producing a dump of raw responses and calling it analysis. This skill synthesizes across four channels to produce structured insight: what users are experiencing, what they want, and what the gap is between the product's intent and their reality.

Four channels, four signal types:

• Surveys — breadth. What fraction of users feel X? Are sentiments correlated with segments?
• Interviews — depth. Why do users feel X? What is the underlying job-to-be-done they are struggling with?
• Analytics — behavior. What are users actually doing, regardless of what they say they do?
• Support tickets — pain. Where is the gap between expectation and reality causing friction?

Behavioral data (analytics) often contradicts attitudinal data (surveys, interviews). Name the contradiction — it is the most informative signal.

Step 1 — Load Context

Read Phase 1 user-research.md — the original user problems and mental models are the baseline. This feedback session measures whether the product solved what Phase 1 found.

Read outcome-report.md — know which outcomes were achieved, partial, or unmet. Feedback collection should investigate the "why" behind those assessments.

Step 2 — Survey Design and Collection

Design a short survey (5-7 questions maximum) targeting the primary user segments from Phase 1.

Survey structure:

1. One Net Promoter Score question (0-10, "how likely to recommend")
2. One overall satisfaction question
3. Two to three questions targeting specific outcome criteria that were PARTIAL or NOT ACHIEVED
4. One open-ended: "What is the one thing you wish this product did differently?"

If surveys have already been collected, skip to synthesis. If not, produce the survey instrument for the user to deploy.

NPS interpretation:

• Promoters (9-10): What is working?
• Passives (7-8): What is missing or friction-causing?
• Detractors (0-6): What failed to meet expectations? Why?

Step 3 — Interview Synthesis

If user interview notes are available, synthesize by job-to-be-done:

For each recurring theme across 3+ interviews:

• Theme: {observed pattern}
• JTBD: {underlying job the user is trying to accomplish}
• Gap: {what the product does vs. what the user needs}
• Frequency: {N of N interviews}

If no interviews are available: flag it. "No qualitative depth available. NPS detractor themes cannot be diagnosed without interviews. Recommend 5 interviews targeting the top 2 NPS detractor reasons before the next cycle."

Step 4 — Analytics Signal

For each key user flow defined in Phase 3 user flows:

Flow	Drop-off Rate	Completion Rate	Avg Time	Trend	Signal
{flow name}	{%}	{%}	{N sec}	↑/↓/→	{interpretation}

Flag any flow with drop-off > 40% — that is a product failure, not a user failure.

Identify: features shipped that show near-zero usage. A feature with <5% adoption after 30 days is a candidate for sunset (flagged for mh-cto-sdlc-roadmap-refresh).

Step 5 — Support Ticket Analysis

Categorize support tickets by theme:

Theme	Volume	% of Total	Severity	Root Cause
{theme}	{N}	{%}	High/Med/Low	{product/UX/documentation/expectation}

Root cause classification:

• Product defect: The product is doing the wrong thing
• UX gap: The product works but users cannot find or use it
• Documentation gap: Users are doing it wrong because they don't know how
• Expectation mismatch: The product does what it does, but users expected something different

Step 6 — Cross-Channel Synthesis

Produce the insight synthesis. This is the output that matters:

What users say vs. what users do: If surveys/interviews claim X but analytics show Y, this is a behavioral contradiction. Name it explicitly.

Top 3 insights: For each:

• Insight: {the finding}
• Evidence: {which channels support this}
• Implication: {what this means for the next cycle}
• Confidence: High / Medium / Low (based on data quality and consistency across channels)

Step 7 — Write Output

Save to _mh-output/cto/sdlc/user-feedback/{YYYY-MM-DD}_{r###}_user-feedback-synthesis.md:

# User Feedback Synthesis

**Product:** {name}
**Synthesis date:** {date}
**Period covered:** {launch date} to {date}
**Data sources:** {which channels had data}

---

## NPS Summary

**Score:** {N} | Promoters: {N%} | Passives: {N%} | Detractors: {N%}

## Top Insights

### Insight 1 — {title}

{body}

### Insight 2 — {title}

{body}

### Insight 3 — {title}

{body}

## Behavioral Contradictions

{attitudinal vs. behavioral gaps, or "None identified."}

## Zero-Usage Features

{List features with <5% adoption — sunset candidates}

## Support Ticket Themes

{Table from Step 5}

## Recommended Actions for Next Cycle

{Top 2-3 actions ranked by impact}

Step 8 — Present Summary

User Feedback Synthesis complete.

Channels synthesized: {N channels with data}

Top insight: {single most important finding}

Zero-usage features flagged: {N} — will feed sunset decisions in mh-cto-sdlc-roadmap-refresh.

Next step: Run /mh-cto-sdlc-defect-retro to trace escaped defects, or /mh-cto-sdlc-roadmap-refresh once all Phase 8 synthesis is complete.