Product Stage — Quantitative Value Testing, Evidence Ladder (E0–E6) & HADI Cycles

Before committing engineering capacity to build a full Minimum Lovable Product (MLP), NBM4 requires teams to test whether the proposed value proposition actually changes customer behavior. Value testing bridges qualitative customer discovery (Stage 02) and production specification (nbm4:spec@4.0.0).

1. The NBM4 Evidence Ladder (E0 to E6)

Every product claim and hypothesis in NBM4 is graded on a seven-level empirical scale. Assumptions start at E0 and must climb the ladder through deliberate tests before unlocking larger resource commitments:

Level	Evidence Tier	Definition & Qualifying Proof	Typical Stage
E0	Founder / Team Belief	Untested internal assertion, intuition, or stakeholder opinion (HiPPO). Carries zero empirical weight.	01-idea
E1	Secondary / Desk Research	Analyst reports, competitor pricing pages, public filings, or market benchmarks.	01-idea
E2	Behavioral Discovery	5+ structured interviews focused strictly on past behavior and recent concrete episodes (The Mom Test).	02-customer
E3	Quantified Friction & Workaround	Documented time, budget, or error cost of the customer’s existing workaround in real workflows.	02-customer
E4	Behavioral Commitment / Value Test	Smoke test, interactive prototype, Wizard-of-Oz trial, or letter of intent where prospects invest time, data, or deposit.	03-product
E5	Pilot / First Lovable Delivery	Target ICP cohort completes the core workflow end-to-end on the MLP with measurable task completion or initial paid conversion.	03-product
E6	Repeatable Scale & Unit Economics	Multi-cohort retention, positive contribution margin, and repeatable acquisition loops verified across multiple cycles.	04-growth

2. Running HADI Learning Cycles (Hypothesis → Action → Data → Insight)

NBM4 operationalizes value testing through bounded HADI cycles:

1. Hypothesis (H): State a falsifiable claim linking a specific target segment, an intervention, an expected behavioral metric, and an explicit falsification threshold.
2. Action (A): Execute the smallest valid test (E3–E5) capable of disproving the hypothesis within a single strategy window (typically 1–2 weeks).
3. Data (D): Collect observed behavioral telemetry (do-data), sample size n, and completion or conversion rates rather than stated opinions (say-data).
4. Insight (I): Evaluate the data against the pre-committed threshold and record one of six terminal decisions:
   • confirm: Threshold met at the target evidence level; promote the hypothesis or expand scope.
   • refine: Direction supported but mechanism or segment boundary requires adjustment for the next cycle.
   • reject: Falsification condition triggered; discard the specific feature or mechanism.
   • pivot: Core customer problem is valid (E2/E3), but the solution architecture or channel must change.
   • pause: External dependency or sample constraint blocks conclusive measurement; freeze allocation.
   • kill: Problem or unit economics disproved; terminate the initiative and archive findings.

3. Governed Templates & Worked Examples

Template A: Product Hypothesis (nbm4:product-hypothesis@1.0.0)

---
artifact_uuid: 0192b8c0-4f10-7000-8000-000000000301
artifact_id: HYP-0014
type: product-hypothesis
artefact_type_id: NBM4-PRODUCT-HYPOTHESIS
schema_version: 1.0.0
version: 1
status: active
created_date: 2026-10-23 10:00 UTC+04:00
updated_date: 2026-10-23 10:00 UTC+04:00
authority_refs:
  - RES-0008
---

## Claim

For B2B SaaS Product Managers running bi-weekly refinement (ICP-0002), replacing freeform ticket descriptions with a mandatory 4-field UJS + falsification prompt will reduce refinement cycle time by at least 30% within two sprints.

## Test design

- Target evidence tier: E4 -> E5
- Cohort & sample size: n = 12 product teams across 2 consecutive sprints (24 refinement sessions)
- Measurement method: Automated session timer comparing baseline sprint refinement duration against treatment sprints

## Falsification signal

Reject this hypothesis if median refinement duration drops by less than 15% OR if more than 25% of participating PMs abandon the template after Sprint 2.

Template B: MutationEvent Record (Emitted at HADI Cycle Close)

{
  "event_type": "MutationEvent",
  "source_cycle": "CYC-0004",
  "target_artifact": "HYP-0014",
  "decision": "confirm",
  "reason": "Median refinement duration decreased by 34% across n=24 sessions with 92% template retention.",
  "evidence_level": "E5",
  "evidence_refs": ["VAL-0019", "RES-0012"],
  "confidence_change": "+0.25",
  "effective_release": "R-2026.11"
}

Template C: PositionClaim Record

position_claim_id: PCL-0003
target_icp: "ICP-0002 (Solo & seed-stage technical PMs)"
core_problem: "Unbounded AI-generated specs create unverified backlog sprawl."
promised_outcome: "Compile-time schema and test-matrix verification before sprint admission."
exclusions:
  - "Not an autonomous code-generation agent"
  - "Does not replace human customer interviews (E2)"
evidence_level: "E4"
falsification_condition: "Fails if setup overhead exceeds 20 minutes per release cycle."

4. Value Testing Checklist

• Every test specifies its target Evidence Ladder tier (E0–E6) before execution begins.
• Sample size (n), time window, and measurement instrument are declared in advance.
• Both a success threshold and an explicit falsification condition are written before collecting data.
• The cycle concludes with one of the six terminal decisions (confirm, refine, reject, pivot, pause, kill) and a recorded MutationEvent.