product-taste-intuition

Original source / Raw SKILL.md

---
name: "product-taste-intuition"
description: "Build stronger product taste + intuition as a PM by running a Taste Calibration Sprint (benchmark set, product critique notes, intuition→hypothesis log, validation plan, practice loop). Use for “product taste”, “product sense”, “intuition”, “calibrate taste”."
---

# Product Taste & Intuition

## Scope

**Covers**
- Developing **product taste** (what “good” looks like) through deliberate exposure, observation, and critique
- Using intuition as a **hypothesis generator** (turning “gut feel” into testable hypotheses)
- Building a repeatable **practice loop** (exposure hours → analysis → validation → updated taste rules)

**When to use**
- “Help me improve my product taste / product sense.”
- “Calibrate what ‘good onboarding’ looks like for our product category.”
- “Turn my intuition about this flow into testable hypotheses.”
- “Create a structured way to study great products and extract patterns.”

**When NOT to use**
- You need to decide *what to build* (use `problem-definition`, `prioritizing-roadmap`, or `defining-product-vision`).
- You need user evidence first (use `conducting-user-interviews` or `usability-testing`).
- You want aesthetic critique only (this is product experience: value, UX, clarity, trust, speed—not just visuals).
- You can’t name any target user, use case, or the “taste domain” you want to improve (we’ll narrow first).

## Inputs

**Minimum required**
- Taste domain to improve (pick 1): onboarding, activation, navigation/IA, editor/workflow, pricing/packaging UX, notifications, retention loops, trust/safety, performance/latency feel, copy/voice
- Target user + top job-to-be-done for that domain
- 3–10 benchmark products/experiences to study (or “unknown—please propose”)
- Time box (e.g., 60–120 min sprint; or a 2–4 week practice plan)
- Constraints (platform, geography, accessibility, compliance, brand voice, etc.)

**Missing-info strategy**
- Ask up to 5 questions from [references/INTAKE.md](references/INTAKE.md).
- If inputs remain missing, proceed with explicit assumptions and provide 2 scope options (narrow vs broad).

## Outputs (deliverables)

Produce a **Taste Calibration Pack** (in-chat Markdown; or as files if requested):

1) **Taste Calibration Brief** (domain, target user/job, what “good” means, constraints)
2) **Benchmark Set** (5–10 products) + “why these” + what to study
3) **Product Study Notes** (1 page per benchmark) using a consistent critique template
4) **Taste Rules + Anti-Patterns** (do/don’t rules derived from evidence)
5) **Intuition → Hypothesis Log** (testable hypotheses + predicted signals)
6) **Validation Plan** (qual + quant checks; smallest viable tests)
7) **Practice Plan** (2–4 weeks: exposure hours + weekly synthesis cadence)
8) **Risks / Open questions / Next steps** (always included)

Templates: [references/TEMPLATES.md](references/TEMPLATES.md)

## Workflow (8 steps)

### 1) Intake + pick the taste domain (narrow the problem)
- **Inputs:** User context; [references/INTAKE.md](references/INTAKE.md).
- **Actions:** Choose 1 taste domain and 1 “moment” (e.g., first-run onboarding). Define target user + job + constraints. Set time box.
- **Outputs:** Taste Calibration Brief (draft).
- **Checks:** A stakeholder can answer: “What specific experience are we calibrating taste for?”

### 2) Define “good taste” as decision criteria (not vibes)
- **Inputs:** Domain + user/job.
- **Actions:** Draft 6–10 criteria (e.g., clarity, time-to-value, trust, agency, error recovery, perceived speed, cognitive load). Add explicit tradeoffs (what you’ll sacrifice).
- **Outputs:** Criteria list + tradeoffs section in the brief.
- **Checks:** Criteria are observable in-product (you can point to UI/behavior), not generic adjectives.

### 3) Build the benchmark set (exposure hours, curated)
- **Inputs:** Known benchmarks (or none).
- **Actions:** Select 5–10 exemplars (direct, adjacent, and at least 1 “gold standard”). For each: what you’re studying and why it’s relevant.
- **Outputs:** Benchmark Set table.
- **Checks:** Set includes at least 2 “outside the category” references to avoid local maxima.

### 4) Study like a voracious user (structured observation)
- **Inputs:** Benchmarks; critique template.
- **Actions:** Use each product as the target user. Capture micro-moments: friction, delight, confusion, trust breaks. Record “what happened” before “why it’s good/bad”.
- **Outputs:** Product Study Notes (draft).
- **Checks:** Each benchmark note includes at least 3 concrete moments with screenshots/quotes if available (or precise descriptions).

### 5) Synthesize: turn observations into taste rules + anti-patterns
- **Inputs:** Study notes across benchmarks.
- **Actions:** Cluster patterns. Convert into rules: **DO/DO NOT**, plus rationale and where it applies. Add anti-patterns that create “AI slop” (generic, incoherent, misaligned experiences).
- **Outputs:** Taste Rules + Anti-Patterns.
- **Checks:** Each rule is backed by ≥ 2 observations from different benchmarks (or explicitly marked “hypothesis”).

### 6) Intuition as hypothesis generator (make it testable)
- **Inputs:** Rules + your gut reactions.
- **Actions:** Write intuition statements (“It feels off because…”) and convert into testable hypotheses with predicted signals and counter-signals.
- **Outputs:** Intuition → Hypothesis Log.
- **Checks:** Each hypothesis has a clear falsification condition (“If X doesn’t change after Y, we were wrong.”).

### 7) Validate with smallest viable checks (qual + quant)
- **Inputs:** Hypothesis log; available data/research access.
- **Actions:** Choose the lightest validation per hypothesis: usability task, intercept prompt, session replay review, funnel slice, A/B smoke test, copy test, etc. Define success metrics and sample.
- **Outputs:** Validation Plan with owners/cadence if known.
- **Checks:** Validation steps are feasible within the stated time box and don’t require sensitive data.

### 8) Create a practice loop + quality gate + finalize
- **Inputs:** Draft pack.
- **Actions:** Build a 2–4 week practice plan (exposure hours schedule + weekly synthesis). Run [references/CHECKLISTS.md](references/CHECKLISTS.md) and score with [references/RUBRIC.md](references/RUBRIC.md). Add Risks/Open questions/Next steps.
- **Outputs:** Final Taste Calibration Pack.
- **Checks:** A reader can follow the practice plan without additional context; assumptions are explicit.

## Quality gate (required)
- Use [references/CHECKLISTS.md](references/CHECKLISTS.md) and [references/RUBRIC.md](references/RUBRIC.md).
- Always include: **Risks**, **Open questions**, **Next steps**.

## Examples

**Example 1 (Onboarding):** “Calibrate our onboarding taste vs best-in-class. Target users are first-time PMs. Time box: 90 minutes. Output a Taste Calibration Pack.”  
Expected: benchmark set, critique notes, taste rules, hypotheses, and a lightweight validation plan.

**Example 2 (B2B workflow UX):** “My gut says our ‘create project’ flow feels slow and confusing. Turn that into testable hypotheses and a validation plan.”  
Expected: intuition→hypothesis log with falsification conditions and smallest viable checks.

**Boundary example:** “Tell me what good taste is in general.”  
Response: require a specific domain + target user/job; otherwise produce a menu of domain options and propose a narrow starting point.



---

## Referenced Files

> The following files are referenced in this skill and included for context.

### references/INTAKE.md

```markdown
# Intake (question bank)

Ask up to **5** questions, then proceed with explicit assumptions.

## A) Goal + scope
1) What **taste domain** are we calibrating? (pick 1) onboarding, activation, navigation/IA, editor/workflow, pricing/packaging UX, notifications, retention loops, trust/safety, performance/latency feel, copy/voice
2) Who is the **target user** and what is their top **job-to-be-done** in this domain?
3) What decision will this work inform? (e.g., redesign direction, critique standard, experiment backlog, PM skill-building)
4) What’s the **time box**? (single 60–120 min sprint vs 2–4 week practice plan)

## B) Constraints + context
5) Platform + constraints: web/mobile, accessibility level, compliance, brand voice, markets/geos, technical limitations
6) Where are you starting from? (existing flow, new concept, competitor catch-up, quality bar reset)

## C) Benchmarks + references
7) Any known benchmark products you admire (or dislike) for this domain?
8) Should we include “outside-category” references? (default: yes, 2–3)

## D) Evidence + validation access
9) What evidence is available today? (analytics, funnel, session replays, support tickets, interviews)
10) What validation methods are feasible this week? (usability calls, prototype tests, A/B, copy test, dogfooding)

## E) Output preferences
11) Should the output be **in-chat Markdown** or written as **files**? If files, what folder path?
12) Any preferences for tone/voice (e.g., direct, brand-aligned), or a rubric style you already use?


```

### references/TEMPLATES.md

```markdown
# Templates

Use these templates to produce the Taste Calibration Pack.

## 1) Taste Calibration Brief (template)

### Context
- **Taste domain:** (e.g., onboarding)
- **Target user + job:** (who / what they’re trying to do)
- **Decision this informs:** (redesign direction, quality bar, experiment backlog, etc.)
- **Time box:** (e.g., 90 minutes; or 4-week practice plan)
- **Constraints:** (platform, accessibility, compliance, brand voice, technical constraints)

### “Good” decision criteria (6–10)
- Criteria list (observable, not vibes)

### Tradeoffs / non-goals
- What you will intentionally sacrifice or deprioritize

## 2) Benchmark Set (table)

| Benchmark | Why it’s relevant | What to study | Notes / links |
|---|---|---|---|
|  |  |  |  |

## 3) Product Study Notes (1 page per benchmark)

### Benchmark
- Name + platform + user persona used

### Quick summary
- What this product does well in this domain (1–2 sentences)

### Moments (capture “what happened”)
| Moment | What I did | What happened | Emotion / friction | Why it might work (hypothesis) |
|---|---|---|---|---|
|  |  |  |  |  |

### Pattern candidates
- Potential rules (draft): DO / DO NOT

### Copy/pasteable artifacts
- Key microcopy, interaction details, defaults, empty states, error handling, etc.

## 4) Taste Rules + Anti-Patterns

### Taste rules (DO / DO NOT)
| Rule | Rationale | Evidence (benchmarks/moments) | Applies to | Exceptions |
|---|---|---|---|---|
| DO:  |  |  |  |  |
| DO NOT: |  |  |  |  |

### Anti-patterns (“slop filters”)
| Anti-pattern | How it shows up | Why it’s harmful | Replacement rule |
|---|---|---|---|
|  |  |  |  |

## 5) Intuition → Hypothesis Log

| Intuition statement (“it feels…”) | Hypothesis (testable) | Predicted signal | Counter-signal | Smallest viable test |
|---|---|---|---|---|
|  |  |  |  |  |

## 6) Validation Plan (smallest viable checks)

### Hypotheses to validate (top 3–7)
- Link to rows in the hypothesis log

### Tests
| Hypothesis | Method | Sample | Success metric | Decision rule | Owner | When |
|---|---|---:|---|---|---|---|
|  |  |  |  |  |  |  |

### Instrumentation / tracking notes
- Events, funnels, qualitative tags, logging plan

## 7) Practice Plan (2–4 weeks)

### Cadence
- **Exposure hours:** (# sessions/week × minutes/session)
- **Weekly synthesis:** (time + output)
- **Peer calibration:** (optional: 1 session/week)

### Weekly plan
| Week | Focus | Benchmarks | Output |
|---:|---|---|---|
| 1 | Baseline + criteria |  | Taste brief + benchmark set |
| 2 | Deep studies |  | Study notes + rule drafts |
| 3 | Hypotheses + tests |  | Hypothesis log + validation |
| 4 | Consolidate |  | Final rules + next steps |


```

### references/CHECKLISTS.md

```markdown
# Checklists

Use these before finalizing the Taste Calibration Pack.

## 1) Pack completeness (must-have)
- [ ] A single, explicit **taste domain** is chosen (not “product quality” in general).
- [ ] Target **user + job** are stated.
- [ ] “Good” criteria are **observable** and include **tradeoffs / non-goals**.
- [ ] Benchmark set includes **5–10** items and at least **2 outside-category** references (or justification for not doing so).
- [ ] Study notes include **concrete moments** (what happened) before interpretation.
- [ ] Taste rules are written as **DO / DO NOT** and reference supporting evidence.
- [ ] Hypotheses are **testable** with predicted signals + counter-signals.
- [ ] Validation plan uses the **smallest viable tests** that fit the time box.
- [ ] A 2–4 week **practice cadence** is specified (even if the immediate output is a 90-minute sprint).
- [ ] Includes **Risks / Open questions / Next steps**.

## 2) “Taste rule” quality
- [ ] Rule is specific enough to apply to a UI decision this week.
- [ ] Rule describes *when* it applies and includes exceptions.
- [ ] Rule is not merely “make it simpler / better / clearer” without a mechanism.
- [ ] Rule has ≥ 2 evidence points or is explicitly marked “hypothesis”.

## 3) Hypothesis quality
- [ ] Hypothesis is falsifiable (clear fail condition).
- [ ] The predicted signal is measurable (qual tags or quant metrics).
- [ ] The smallest viable test does not require sensitive or unavailable data.

## 4) Validation plan realism
- [ ] Tests fit the time box and staffing realities.
- [ ] There is a decision rule (what you’ll do if it passes/fails).
- [ ] You’re not over-indexing on one method (mix qual + quant where feasible).


```

### references/RUBRIC.md

```markdown
# Rubric (1–5)

Score the Taste Calibration Pack. A “ship-ready” pack is typically **≥ 24/35** with no category below 3.

## 1) Domain focus (1–5)
- **1:** Vague (“improve product taste”) with no specific domain or moment.
- **3:** Domain is named; scope is mostly clear.
- **5:** Domain + moment + target user/job are crisp and clearly bounded.

## 2) Criteria + tradeoffs (1–5)
- **1:** Generic adjectives; no tradeoffs.
- **3:** Criteria are mostly observable; tradeoffs exist but are thin.
- **5:** Criteria are concrete, prioritized, and include explicit tradeoffs/non-goals.

## 3) Benchmark set quality (1–5)
- **1:** Random list; no rationale; no outside-category references.
- **3:** Reasonable set with some rationale.
- **5:** Curated set with clear “what to study” per benchmark and diversity across categories.

## 4) Observation depth (1–5)
- **1:** Opinions without moments.
- **3:** Some concrete moments captured; inconsistent structure.
- **5:** Consistent, moment-based notes across benchmarks with precise details and pattern candidates.

## 5) Rules + anti-patterns (1–5)
- **1:** Rules are vague or ungrounded.
- **3:** Rules exist but lack applicability guidance or evidence.
- **5:** DO/DO NOT rules are evidence-backed, scoped, and immediately reusable in product work.

## 6) Hypotheses + validation (1–5)
- **1:** Intuition remains subjective; no falsification.
- **3:** Some hypotheses and tests; decision rules missing.
- **5:** Hypotheses are testable with predicted/counter-signals and realistic smallest-viable validation steps.

## 7) Practice loop (1–5)
- **1:** One-off critique with no habit plan.
- **3:** A cadence is described but not operational.
- **5:** A 2–4 week plan with exposure hours + weekly synthesis + optional peer calibration is ready to execute.


```