Product Factory Validation Process

Define an initial ideal-customer profile and a recruitment plan for 6–11 relevant participants, explicitly stating which Lane M claims that sample cannot support.

Write a non-leading interview guide that tests unprompted problem recognition, revealed cost, abandoned alternatives, and quantified value.

Convert interview notes into traceable evidence records without treating summaries, simulations, or AI-generated text as evidence.

Calculate actual validation figures from the evidence and compare them with thresholds committed before data collection.

Write a decision-trigger statement that either justifies or skips market and competitive analysis before research begins.

Design a time-boxed feasibility spike for the hardest unknown with a measurable pass/fail result.

Estimate one price, one dominant cost, one costed acquisition channel, and a plausible route to 100 target users.

Score the original and alternative concepts on common criteria, write the strongest STOP case first, and record the resulting concept decision.

Express product strategy as target user, problem, differentiated value, and one measurable proof of success.

Map each proposed MVP must-have to a specific evidence-backed hypothesis and move unmapped features to Not Now.

Write the core user journey as no more than five observable end-to-end steps.

Define acceptance criteria and analytics events that reveal whether the core hypotheses were tested.

Reduce scope until the build estimate fits the selected lane rather than extending the lane cap.

Design a real-user validation protocol measuring unassisted completion, value reached, unprompted return, and willingness to pay.

Apply the one-cycle two-strike rule to validation results and record a FIX, ITERATE, PROCEED, PIVOT-as-new-run, or STOP decision.

Specify which activities an AI agent may assist with and which evidence, scoring, challenge, and signature decisions must remain human-owned.

Model runs, phases, claims, evidence references, gates, roles, decisions, and deviations as provider-independent domain objects.

Define ports and adapters that isolate LLM providers, cloud services, evidence retrieval, identity, and persistence from process rules.

Implement in Python a numeric gate whose state transition defaults to STOP when required evidence or thresholds are missing.

Implement a provenance validator that identifies untagged claims and enforces the load-bearing-assumption ceiling.

Persist gate decisions and deviations as append-only records with actor, time, threshold, actual result, and rationale.

Build a minimal TypeScript review interface that presents sourced evidence, the STOP case before the PROCEED case, and the human decision control.

Write contract tests that demonstrate the same process behavior after replacing LLM and persistence adapters.

Prepare a launch gate with pre-launch baselines, a tested rollback, named support ownership, numeric kill criteria, and a review date.

Compare post-launch behavior with Evidence Pack claims and record both confirming and contradicting divergences.

Turn an observed signal into a cause hypothesis, a measurable experiment, and a hypothesis-linked backlog item.

Apply Phase 12 kill criteria during a monthly portfolio review to choose SCALE, OPTIMIZE, EXPAND, PIVOT, MAINTAIN, or RETIRE.

Calculate kill rate, cost per killed idea, decision lead time, survival, original-idea win rate, and lane-overrun rate from run records.

Classify a completed gate as predictive, false positive, or observable false negative, then propose a numeric threshold adjustment.

Revise and version the standard when recurring deviations show that the documented process—not the team—needs to change.

Assess a venture or client team's readiness and select a bounded pilot rather than attempting an organization-wide rollout.

Facilitate a gate-setting session that precommits numeric thresholds, decision rights, lane caps, and STOP actions.

Design governance that separates Owner, Challenger, decider, and Factory Keeper duties while handling committed ideas explicitly.

Diagnose gate gaming, fabricated provenance, performative alternatives, cap extensions, and contaminated calibration statistics in an adoption scenario.

Present an evidence-backed client recommendation to adopt, modify, or stop the process, with pilot success measures and a review cadence.