# Biz Validate > Runs iterative hypothesis validation loop. Triggers ONLY when: biz-think acid test fails, user explicitly asks to validate a business assumption, or user says to pivot. Do NOT trigger when: discussing business strategy casually or during product development. - Skill: `docat0209/biz-validate` (Agent Skill) - Install (CLI): `npx skillmds@latest add docat0209/biz-validate` - Raw SKILL.md: https://api.skillmd.com/api/skills/docat0209/biz-validate/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: Docat0209 (https://skillmd.com/u/docat0209) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/docat0209/biz-validate --- # biz-validate: Business Hypothesis Validation Loop A cyclical SOP based on Running Lean (Ash Maurya), Pretotyping (Alberto Savoia), and Hypothesis-Driven Entrepreneurship (HBS). This skill LOOPS — it does not run once and stop. **Framework sources:** - Running Lean 3rd Ed (Ash Maurya) — macro stages: Problem/Solution Fit → Product/Market Fit → Scale - Pretotyping (Alberto Savoia, Google) — behavioral proof before building - Reflect-Inquire-Learn (Eisenmann/Ries, HBS) — single experiment cycle --- ## Entry Conditions Run this skill when: - `biz-think` acid test fails (cannot complete the one-sentence answer) - An assumption in the business model needs validation - Pivoting after a failed experiment - User says "validate this" or "test this idea" --- ## The Loop: Reflect → Hypothesize → Experiment → Learn → Update ``` ┌─────────────────────────────────────┐ │ │ ▼ │ REFLECT │ What is the riskiest assumption? │ │ │ ▼ │ HYPOTHESIZE │ "At least X% of [target] will │ [action] if [condition]" │ │ │ ▼ │ EXPERIMENT │ Run the cheapest possible test │ (hours, not weeks) │ │ │ ▼ │ LEARN │ Did we hit the threshold? │ │ │ ├── NO → Pivot or kill ──────────────►│ │ Update canvas, re-enter │ │ │ ├── YES → Next assumption ───────────►│ │ │ └── ALL VALIDATED → EXIT ─────────────┘ ``` **Max iterations:** 5 per assumption. If an assumption cannot be validated in 5 cycles, kill the idea or pivot the fundamental approach. --- ## Step 1: REFLECT — Identify the Riskiest Assumption List all assumptions the business model depends on. Rank by: - **Impact if wrong:** would this kill the business? (High/Medium/Low) - **Uncertainty:** how confident are we? (Guess/Belief/Tested) Pick the assumption that is High-impact AND Guess/Belief. This is what you test first. Common assumption categories: - **Problem assumption:** "This problem exists and is painful enough to pay to solve" - **Solution assumption:** "Our solution actually solves this problem" - **Pricing assumption:** "They will pay $X for this" - **Channel assumption:** "We can reach them through [channel]" - **Retention assumption:** "They will keep using this after the first time" --- ## Step 2: HYPOTHESIZE — Write a Testable XYZ Hypothesis Format (Savoia's Pretotyping): > "At least **X%** of **[target audience]** will **[desired action]** if **[condition/offering]**" Rules: - X must be a specific number, not "some" or "many" - Target audience must be concrete, not "developers" but "DevOps engineers who use Terraform" - Desired action must be behavioral (click, sign up, pay), not attitudinal (say they like it) - Set the threshold BEFORE running the experiment, not after **Also set fail conditions:** what result would KILL this hypothesis? Write it down. If you can't define failure, you can't learn. --- ## Step 3: EXPERIMENT — Run the Cheapest Possible Test **Experiment types (ordered by cost, cheapest first):** | Type | What | Time | Cost | Evidence Strength | |---|---|---|---|---| | **Fake Door** | Landing page with signup/buy button, no product behind it | 2 hours | $0 | Medium (interest, not commitment) | | **Concierge** | Manually deliver the service to 3-5 people | 1-3 days | $0 | High (real usage data) | | **Wizard of Oz** | Users think it's automated, but you do it manually behind the scenes | 3-5 days | $0 | High (real behavior) | | **Pre-sell** | Take actual payments before the product exists | 1 day | $0 | Highest (money moved) | | **Smoke Test** | Ads/posts pointing to a signup page, measure conversion | 1 day | $0-50 | Medium (intent, not usage) | **Rules:** - Always prefer skin-in-the-game data (money, time invested) over opinions (surveys, "I would use this") - "Would you use this?" is worthless. "Here's the payment page, buy now" is data. - If an experiment takes more than 1 week, it's too expensive. Simplify. - Search the web for existing data before running your own experiment (someone may have already validated this) --- ## Step 4: LEARN — Evaluate Against Threshold Compare results to your XYZ hypothesis threshold: | Result | Action | |---|---| | **Exceeded threshold** | Assumption validated. Move to next riskiest assumption. | | **Close to threshold (within 20%)** | Iterate: adjust offering, pricing, or messaging. Re-run experiment. | | **Far below threshold** | Assumption invalidated. Decide: pivot or kill. | **Pivot options (Steve Blank):** - **Zoom-in pivot:** one feature of the product becomes the whole product - **Zoom-out pivot:** the whole product becomes one feature of a larger product - **Customer segment pivot:** same product, different audience - **Customer need pivot:** same audience, different problem - **Channel pivot:** same product, different distribution channel - **Revenue model pivot:** same product, different pricing/monetization **Kill criteria:** if 3+ pivots on the same core idea all fail, kill it entirely. Start fresh with a new idea through `biz-think`. --- ## Step 5: UPDATE — Revise the Lean Canvas After each cycle, update these in auto memory (`memory/validation_log.md`): - `Experiment`: hypothesis, method, result, date, learning - `AssumptionStatus`: which assumptions are now Validated/Invalidated/Untested - `LeanCanvas`: updated customer segment, problem, solution, channels, revenue based on learnings - `PivotHistory`: if pivoted, record what changed and why --- ## Exit Conditions **Exit to development (invoke `pre-code`):** - Problem assumption validated (people have this pain and will pay) - Solution assumption validated (your approach solves it) - At least 1 person has paid or committed to pay (pre-sale, LOI, deposit) - `biz-think` acid test can now be completed in one sentence **Exit by killing:** - 3+ pivots failed on the same core idea - No behavioral evidence of willingness to pay after 3 experiment cycles - Store the kill decision and learnings in auto memory (`memory/validation_log.md`) **Do NOT exit because you're tired of iterating.** Exit only when evidence supports the decision. --- ## Macro Stages (Running Lean Traction Roadmap) Track which stage you're in: | Stage | Goal | How You Know You're Done | |---|---|---| | **1. Problem/Solution Fit** | Validate the problem is real and your solution concept resonates | 5+ people describe the problem unprompted; solution demo gets "when can I use this?" | | **2. Product/Market Fit** | Validate people will pay and keep using it | Repeatable acquisition, activation rate >40%, monthly churn <5% | | **3. Scale** | Grow the validated model | Not your concern until stages 1-2 are done | **You are almost certainly at Stage 1.** Do not skip to Stage 2 thinking. ## Next Steps Report to user: "Validation [PASS/FAIL]. Assumption: [which one]. Evidence: [what we found]" Suggested next steps (user decides): - All validated → "Run brand-build, then work-breakdown" - Kill criteria met → "Start fresh with biz-think" - Need experiment → "Run copy-craft for landing page"