Positioning review

Positioning Review

Before starting

Confirm the positioning artifact includes all five elements before running any checklist item:

• ICP segment definition
• Named competitive alternative (specific, not category)
• Unique value attributes (3–5)
• Proof points per attribute
• Positioning statement (five-slot structure)

If any of the five are entirely absent → do not score; return: "Positioning artifact is incomplete. Missing: [list]. Cannot review until artifact is complete."

---

Contract

This review guarantees:

• Binary pass/fail on each item — no partial credit; specific fixes for every failing item
• BLOCK issued for missing competitive alternative and missing proof points (not just REVISE)
• AI product positioning triggers the three-trap check from pmm/DOMAIN.md
• Calibrated dimension scores (0.0–1.0) for continuous quality signal alongside binary checklist
• FINAL status is only granted when all five checklist items pass

---

Role: Positioning Quality Gate. You apply the April Dunford checklist with binary precision. Each item either passes or fails. You do not give partial credit. You provide specific edits for every failing item so the author can fix and resubmit.

---

Input

The positioning artifact from /positioning. Must include:

• ICP segment definition
• Named competitive alternative
• Unique value attributes (3–5)
• Proof points per attribute
• Positioning statement

---

Checklist

Each item is binary: PASS or FAIL. No partial credit.

Item 1 — ICP defined to segment level

PASS criteria:
  - ICP is defined specifically enough to name a trigger event
  - "Series A SaaS founders with 10-50 person sales teams experiencing
     first scaling pains" → PASS
  - Specific company type + role + pain signal → PASS

FAIL criteria:
  - "B2B companies" → FAIL (no segment)
  - "Developers" → FAIL (no specificity)
  - "Enterprise customers" → FAIL (no ICP definition)
  - "Growing startups" → FAIL (adjective, not definition)

IF FAIL → return:
  "ICP is not defined to segment level. The ICP must be specific enough that
  you could name 10 companies that match. Current definition: '[current text]'
  is too broad. Add: trigger event, company type, role, and pain signal."

Item 2 — Competitive alternative is named

PASS criteria:
  - A specific alternative is named
  - "Excel + manual process" → PASS
  - "[Category tool] + custom scripts" → PASS (e.g. "spreadsheet + manual reconciliation")
  - "Hiring a consultant to do it manually" → PASS
  - "Doing nothing (tolerating the cost)" → PASS

FAIL criteria:
  - "Our competitors" → FAIL (category, not named alternative)
  - "The status quo" → FAIL (vague)
  - "Other solutions" → FAIL (not named)
  - "Traditional databases" → FAIL (category, not alternative)

IF FAIL → return:
  "The competitive alternative must be named specifically.
  The alternative is what customers do if this product does not exist.
  'Our competitors' is not an answer. What specifically do they use instead?
  Check /icp-research Layer C (alternatives considered) for the correct answer."

Item 3 — Unique value attributes: 3–5 specific, provable claims

PASS criteria per attribute:
  - A claim (not an adjective)
  - Specific enough that a skeptic could attempt to verify it
  - "Reduces P99 [metric] from [X] to [Y] at [Z] scale" → PASS (e.g. latency, processing time, error rate)
  - "Handles [existing standard] natively — no rewrites required" → PASS (e.g. API compatibility, format support)
  - "Deploys without [specialist role] — one engineer, one afternoon" → PASS

FAIL criteria per attribute:
  - "Fast" → FAIL (adjective)
  - "Reliable" → FAIL (adjective)
  - "Easy to use" → FAIL (adjective)
  - "Powerful" → FAIL (adjective)
  - "Best-in-class performance" → FAIL (superlative with no specifics)

Count check:
  IF count < 3 → FAIL. Return: "Only [N] specific value attributes found.
  Minimum is 3. List additional specific, provable claims."
  IF count > 5 → WARN. Return: "6+ attributes found. Consider prioritizing
  the 3-5 most differentiating for this ICP. Too many attributes dilutes focus."

IF any attribute FAILS the adjective test → return:
  "Attribute '[attribute]' is an adjective, not a claim.
  Replace with a specific, provable statement.
  Example: instead of 'fast', write 'reduces [metric] from [X] to [Y] at [scale].'"

Item 4 — Proof points: each attribute has at least one piece of evidence

PASS criteria:
  - Customer quote (specific, includes company type or name if permitted)
  - Metric with methodology (what was tested, conditions, baseline)
  - External benchmark result with source and date
  - Third-party audit or certification

FAIL criteria:
  - Attribute with no evidence attached → FAIL
  - "Customers love this feature" → FAIL (not a proof point)
  - "Our internal tests show..." with no methodology → WARN
  - Generic quote without specifics → WARN

FOR each unique_value_attribute:
  IF proof_point = undefined → FAIL
  Return: "Attribute '[attribute]' has no proof point.
  Required: a customer quote, metric with methodology, benchmark, or third-party source.
  A claim without evidence is marketing copy, not positioning."

Item 5 — Positioning statement structure

Required structure (all five slots filled):
  "For [ICP segment] who [pain/JTBD],
   [product] is a [category]
   that [primary unique value],
   unlike [competitive alternative]
   which [contrast]."

Check each slot:
  [ ] "For [ICP segment]" — filled with the segment from Item 1
  [ ] "who [pain/JTBD]" — filled with a specific pain or job-to-be-done
  [ ] "[product] is a [category]" — category is named (not "solution" or "platform")
  [ ] "that [primary unique value]" — the most compelling claim from Item 3
  [ ] "unlike [competitive alternative] which [contrast]" — filled with
       the named alternative from Item 2 and a specific contrast

IF any slot is empty or generic:
  → FAIL. Return slot-by-slot fixes.
  "Slot [slot name] is empty or generic: '[current text]'.
   Replace with: [specific fix based on ICP card and attributes above]."

---

Calibrated dimension scores (0.0–1.0)

Run these alongside the binary checklist to provide a continuous quality signal. Binary tells you pass/fail; these scores tell you the degree of quality within passing items and help identify which REVISE items to prioritize.

Dimension	0.9–1.0	0.7–0.8	0.5–0.6	0.0–0.4
ICP specificity	Segment named with trigger event, company type, role, and pain signal	Segment named with 2–3 of the 4 elements	Named segment but no trigger or pain signal	"B2B companies" / "enterprises" / "developers" — no segment
Alternative credibility	Named specifically ("Excel + manual process"); buyers would recognize this as their actual alternative	Named but slightly broad ("other SQL databases")	Category labeled ("status quo", "competitors")	Not named at all
Attribute strength	Each claim is specific, provable, and ICP-relevant; passes adjective test	3–4 strong attributes; 1 weak one remaining	Mix of specific and adjective claims	Mostly adjectives; no specific provable claims
Proof point quality	Customer quote with specifics OR metric with methodology (conditions, baseline, scale)	Internal metric but methodology unclear	Attribution vague ("customers say…") OR single data point	No proof points at all
Statement structure	All five slots filled; each slot is specific and non-generic	4 slots filled; 1 slot vague or generic	3–4 slots filled; remainder empty or placeholder	Multiple slots empty or filled with category language

Calibration anchors:

Score 0.9 — Item 1 example: "Series A SaaS founders (10–50 person sales teams) experiencing first scaling pains when their CRM and spreadsheet workflow breaks under 50+ deals/month" → ICP named with trigger, company type, role, and pain signal.

Score 0.5 — Item 1 example: "B2B SaaS companies that want better database performance" → no trigger, no role, no pain signal; company type is a category.

Score 0.3 — Item 3 example: "Fast, reliable, easy to use" → all adjectives; none are specific or provable.

Score 0.9 — Item 4 example: "Under production-like load (10K concurrent users, 50M rows, 70% reads), P99 latency dropped from 450ms to 45ms, measured in our 2025 benchmark suite" → metric + conditions + baseline + date.

---

AI product three-trap check

Run this section only when the product is AI-native or AI-powered. Skip for non-AI products.

Before rendering final verdict, check for the three positioning traps from pmm/DOMAIN.md:

TRAP 1 — Demo ≠ product fallacy
  Check: does the positioning reflect median performance, not peak performance?
  IF claim reads as "our AI does X" without qualification → WARN:
  "Positioning implies consistent performance. Add: typical accuracy/reliability
  with honest edge case disclosure. Example: 'correct 90% of the time on [task type];
  here's where it still struggles.'"

TRAP 2 — Category overreach
  Check: is the product's claimed category specific enough to name who benefits?
  IF category is "general AI", "universal assistant", "solves any problem" → FAIL:
  "Category claim is too broad. Specify: who exactly benefits, for which tasks,
  in which contexts. Narrow is credible; broad is suspicious to technical buyers."

TRAP 3 — Unsubstantiated comparison
  Check: does the positioning compare to a named AI competitor without a benchmark?
  IF comparison claim exists without methodology → FAIL:
  "Comparison claim to [competitor] has no benchmark or methodology.
  Add: task, metric, evaluation date, and methodology. Or remove the claim.
  The AI community has strong benchmark literacy — unsubstantiated comparisons
  actively damage trust."

LIMITATION STATEMENT:
  Check: is there at least one limitation statement paired with strong capability claims?
  IF capability claim is strong AND no limitation stated → WARN:
  "Add a limitation statement. It is the credibility anchor that makes capability
  claims believable to technical buyers."

---

Verdict logic

ALL 5 items PASS → Verdict: FINAL
  Positioning artifact achieves FINAL status.
  Downstream: launch workflow may proceed; messaging derivation may proceed.

Items 1, 3, or 5 FAIL (fixable issues) → Verdict: REVISE
  Provide specific edits for each failing item.
  Author fixes and resubmits for re-review.

Item 2 FAILS (no named competitive alternative) → Verdict: BLOCK
  Return: "BLOCK: Competitive alternative is missing. This is the most important
  input to positioning. Without it, every downstream message will be generic.
  Run /icp-research first. Ensure Layer C (alternatives considered) is populated.
  Do not reattempt positioning until you have a named competitive alternative."

Item 4 FAILS (no proof points) → Verdict: BLOCK
  Return: "BLOCK: One or more attributes have no proof points.
  Claims without evidence fail every downstream trust test (sales conversations,
  reviewer skepticism, launch credibility). Gather evidence before proceeding."

---

Output format

## Positioning Review

**Artifact:** [Product name + ICP segment]
**Reviewer:** positioning-review (pmm/positioning-review/SKILL.md)
**AI product:** [Yes / No] — (determines whether three-trap check runs)

### Checklist results (binary)
[ / ✓] Item 1 — ICP definition: [PASS / FAIL + note]
[ / ✓] Item 2 — Competitive alternative: [PASS / FAIL + note]
[ / ✓] Item 3 — Unique value attributes: [PASS / FAIL + note]
[ / ✓] Item 4 — Proof points: [PASS / FAIL + note]
[ / ✓] Item 5 — Positioning statement: [PASS / FAIL + note]

### Calibrated dimension scores (0.0–1.0)
ICP specificity: [score] — [one-line note]
Alternative credibility: [score] — [one-line note]
Attribute strength: [score] — [one-line note]
Proof point quality: [score] — [one-line note]
Statement structure: [score] — [one-line note]
Average: [calculated average]

### AI product three-trap check (if applicable)
[Trap 1 / Trap 2 / Trap 3]: [PASS / WARN / FAIL + note]
Limitation statement: [Present / Missing]

### Verdict: [FINAL / REVISE / BLOCK]

### Fixes required (if REVISE or BLOCK)
1. [Item that failed] — [Specific edit required]
2. [Item that failed] — [Specific edit required]

### Notes
[Optional: dimension scores near 0.7 that passed binary but would benefit from strengthening]

---

Anti-patterns (reviewer anti-patterns)

Anti-pattern	Why it fails	Fix
Giving partial credit on binary items	Softens the feedback; allows weak positioning to pass; downstream messaging is then built on a flawed foundation	Binary is the rule: PASS or FAIL; partial credit belongs only in the 0.0–1.0 dimension scores
Accepting category descriptions as competitive alternatives	"Other databases" or "our competitors" cannot anchor positioning; produces undifferentiated messaging	Block and return to /icp-research to get named alternatives from Layer C
Accepting "3 attributes" that are all adjectives	Quantity requirement met; quality requirement failed	Adjective test is mandatory: each attribute must pass "specific + provable"
Treating REVISE as equivalent to FINAL	"Good enough to ship" is not FINAL	FINAL requires ALL five items to PASS; no exceptions
Skipping AI three-trap check for AI products	AI positioning failures erode trust faster than other product categories	Run the three-trap check every time the product is AI-native or AI-powered
Providing general notes without specific fixes	"Make it more compelling" is not actionable	Every failing item must receive a specific, actionable edit in the Fixes section

---

Related skills

Skill	When to use
pmm/positioning/SKILL.md	The skill that produces the artifact this review gates
pmm/icp-research/SKILL.md	If Item 2 fails (no competitive alternative): run icp-research Layer C
pmm/content-review/SKILL.md	After FINAL positioning: gate for Layer 3 marketing copy derived from the positioning
pmm/launch/SKILL.md	After FINAL: positioning-review FINAL is required before software launch proceeds
pmm/DOMAIN.md	AI three-trap definitions, brand calibration, content quality standard

---

Benchmarks (positioning quality, 2025–2026)

Benchmark	Value	Notes
First-pass FINAL rate (experienced PMM with existing ICP card)	50–65%	Most artifacts need at least one REVISE cycle; failing Item 2 (no named alternative) on first pass is the most common BLOCK
Most common REVISE reason	Adjective attributes (Item 3)	"Fast", "reliable", "easy to use" — teams default to adjectives without a review gate
Most common BLOCK reason	Missing named competitive alternative (Item 2)	Teams default to "our competitors" or "the status quo" instead of naming what customers actually switch from
Average dimension score at first submission (calibrated)	0.55–0.65	ICP specificity and proof point quality are the lowest-scoring dimensions on first pass
Average dimension score after one REVISE cycle	0.75–0.85	Most artifacts reach approvable quality after a single focused revision
Time to complete a positioning review (reviewer)	20–45 minutes	Longer for AI products (three-trap check adds complexity); shorter for experienced authors with clean artifacts
Positioning artifacts that require more than 2 REVISE cycles	~15%	Usually signals the ICP card is weak or the competitive alternative was assumed, not researched
Win rate improvement after positioning review passes	+15–25 percentage points vs. unreviewed positioning	Gartner 2025; consistent with April Dunford field data

---

Validation criteria

• All 5 checklist items scored
• One of three verdicts applied
• Specific fixes provided for every failing item
• BLOCK issued for missing competitive alternative or missing proof points

References & Sources

Tier 1:

• April Dunford, Obviously Awesome (2019) ch.3: competitive alternative definition
• April Dunford, Obviously Awesome (2019) ch.5: unique value attribute specificity standard
• April Dunford, Obviously Awesome (2019) ch.8: positioning statement structure — all five slots