SuperAPI AI Directory

lean-startup-skill

General & other

View repository

Updated 10-15 23:44

About this skill

Lean Startup · Thinking Tool

"A startup is a human institution designed to create a new product or service under conditions of extreme uncertainty." — Eric Ries

Activation Conditions and Trigger Words

  • Direct invocation: "Analyze with Lean Startup...", "Design an MVP", "Validate this direction"
  • Semantic trigger: A startup/new product idea but uncertain about the direction, wanting to validate quickly
  • Combined invocation: "First validate with Lean Startup, then amplify with Growth Hacking"

Method Framework Overview

The core of Lean Startup is to learn quickly at minimum cost under extreme uncertainty. It avoids wasting resources on the wrong direction through the Build-Measure-Learn loop, falsifiable hypotheses, MVP as a learning tool, and the mindset of pivoting rather than failing.

Core Principles (3-7, each must be accompanied by ≥2 pieces of cross-domain evidence)

Principle 1: Build-Measure-Learn Loop

One-sentence definition: Do not guess; learn from real user behavior data.

Cross-domain evidence:

  1. E-commerce: Zappos' founder went to a shoe store, took photos, and put them online. Only after someone placed an order did he go buy the shoes, validating that "people buy shoes online" (Source: Hsieh, Delivering Happiness, 2010)
  2. Social tool: Buffer used two landing pages (free/paid) to validate that "some people are willing to pay for social media scheduling" (Source: Buffer Blog, 2010)

How to apply: Design a minimal experiment for each key uncertainty, measuring user behavior rather than what they say.

Limitations: Some products (such as Tesla) require huge upfront investment and cannot be validated with a rapid MVP.

Principle 2: Hypotheses Must Be Falsifiable

One-sentence definition: If you cannot design an experiment to prove the hypothesis wrong, it is not a scientific hypothesis.

Cross-domain evidence:

  1. Philosophy: Popper proposed that the distinction between science and pseudoscience is falsifiability (Source: Popper, The Logic of Scientific Discovery, 1934)
  2. Investment: Naval Ravikant emphasized that a good startup hypothesis should be able to be overturned by data within 2 weeks (Source: Naval, The Angel Philosopher, 2015)

How to apply: Write each key hypothesis as "If X holds, we should see Y data. If we do not see it within 2 weeks, X does not hold."

Limitations: Some value hypotheses (such as "users will like this design") are difficult to falsify purely with data and require qualitative feedback as a supplement.

Principle 3: MVP Is a Learning Tool, Not a Product

One-sentence definition: The goal of an MVP is to maximize learning per unit time, not product sales.

Cross-domain evidence:

  1. Cloud storage: Dropbox used a 3-minute video demo to validate demand, and 50,000 people joined the waiting list (Source: Houston, Dropbox MVP Story, 2010)
  2. Music streaming: Spotify used a very small beta version to validate demand for unlimited streaming music (Source: Ek, Spotify Origin Story, 2010)

How to apply: When designing an MVP, ask, "If I do not build this feature, will it prevent me from learning the most critical thing?"

Limitations: In some fields, an MVP of too poor quality will permanently lose user trust (such as finance and healthcare).

Principle 4: Pivoting Is Progress, Not Failure

One-sentence definition: When the data does not support the current direction, changing direction is the core of the scientific method.

Cross-domain evidence:

  1. E-commerce: Groupon pivoted from a charity crowdfunding platform to a group-buying company (Source: Mason, How I Built This, 2014)
  2. Collaboration tools: Slack pivoted from a gaming company to an enterprise communication tool (Source: Butterfield, Slack Origin Story, 2013)

How to apply: Hold a Pivot/Persevere meeting once a month. Does the data support continuing? If not, where should we pivot?

Limitations: Frequent pivots drain team morale. Before pivoting, there must be a clear decision framework.

Principle 5: Innovation Accounting

One-sentence definition: Use actionable metrics instead of vanity metrics to measure real progress.

Cross-domain evidence:

  1. Instant messaging: IMVU found that registered user count was a vanity metric, while the growth curve of active paying users was the real progress (Source: Ries, The Lean Startup, 2011)
  2. Manufacturing: Toyota used "number of problems generated per hour" rather than total output to measure improvement (Source: Ohno, Toyota Production System, 1988)

How to apply: Define 3 core metrics (acquisition, activation, revenue), and use curve trends instead of total user count.

Limitations: Over-focusing on quantifiable metrics will ignore important but non-quantifiable dimensions (such as users' emotional connection).

Agentic Protocol

Step 1: Problem classification

Applicable signals:

  • High uncertainty in the early startup stage
  • There are multiple directions but resources are only enough to validate a few
  • Need to translate product ideas into verifiable hypotheses

Non-applicable signals:

  • Mature continuous operations optimization (applies Kaizen)
  • Direction is determined and needs scaling (applies growth hacking)
  • Projects requiring huge upfront capital expenditure

Step 2: Lean Startup-style analysis

Research dimensions(each dimension must trace back to a core principle):

  1. Core hypothesis identification: Value hypothesis (do users really need it?) and growth hypothesis (will users grow naturally?)

    • Source principle: Principle 2
    • How to apply: List the 3 most critical hypotheses, each written in a falsifiable format
  2. MVP design: What is the cheapest way to validate these 3 hypotheses?

    • Source principle: Principle 3
    • How to apply: Format "experiment content → minimum cost → minimum learning → success criteria"
  3. Validation metric definition: Which metric will be used to judge whether the hypothesis holds?

    • Source principle: Principle 5
    • How to apply: Define 1 real metric and judgment threshold for each hypothesis
  4. Pivot/Persevere judgment: Based on current data, should we persevere or pivot?

    • Source principle: Principle 4
    • How to apply: Preset "if data < threshold → consider pivoting to [direction]"

Step 3: Lean Startup-style output

  • 3 core hypotheses (falsifiable format)
  • Recommended MVP design plan (including cost and time estimates)
  • Validation metrics and success/failure thresholds
  • Alternative Pivot directions

Applicable/non-applicable judgment

Problem Type Applicability Description
New product direction validation High Classic scenario
Feature prioritization decisions High Replace subjective ranking with experimental data
B2C internet startup High Natively suited to the Lean Startup methodology
Hard tech startup Medium The ideas apply, but the MVP approach is different
Incremental optimization in mature enterprises Low Kaizen is more suitable

Classic case library

Success cases

  1. Dropbox video MVP(Cloud storage/2008)

    • Problem: needed to validate "ordinary people need file synchronization" without writing a single line of backend code
    • Method application: produced a 3-minute product demo video
    • Result: beta waiting list jumped from 5,000 to 75,000
    • Source: Houston, Dropbox Startup Story (2011)
  2. Zappos shoe store MVP(E-commerce/1999)

    • Problem: validate "people will buy shoes online"
    • Method application: took photos in physical shoe stores and put them on the website, then bought the shoes only after orders came in
    • Result: validated that demand was real and obtained external financing
    • Source: Hsieh, Delivering Happiness (2010)
  3. Buffer two-page MVP(Social tool/2010)

    • Problem: validate "some people are willing to pay for social media scheduling"
    • Method application: two landing pages (free version/paid version); clicking the paid button showed "still in development"
    • Result: the paid version had a significant click-through rate, and coding began
    • Source: Buffer Blog, "How We Validated Buffer" (2010)

Failure cases

  1. Color's sky-high failure(Social/2011)
    • Problem: raised $41 million to build a photo-sharing social network
    • Misuse: large-scale development without MVP validation; after the app launched, no one understood how to use it
    • Lesson: having a lot of money does not mean validation can be skipped. Funding amount should not replace user validation
    • Source: TechCrunch, "Color's $41M Failure" (2011)

Misuse detector (≥3 misuse patterns)

Misuse signal Detection logic Warning message Suggested action
Method does not match the problem A hard tech project with huge CAPEX wants to use a rapid MVP "Hard tech projects cannot simply apply the internet MVP. But demand assumptions can be validated—do people really need it?" Adjust the approach
Skipping key steps Designed an MVP but did not define validation metrics "An MVP without validation metrics is just product development, not learning. First define what counts as passing validation." Guide definition of metrics
Ignoring the learning loop Did 3 rounds of iteration but did not record the key insights learned each time "After 3 experiments you still have not recorded what you learned; you are developing a product, not learning." Guide learning review

Honesty boundaries (≥3 specific limitations)

  1. Not applicable to large-scale CAPEX projects: Projects requiring facility construction, such as chip fabs and pharmaceutical plants, cannot be validated with a two-week MVP
  2. May overlook long-term vision: Overly focusing on rapid validation may cause the team to lose commitment to its long-term vision
  3. Sample bias risk: Early users may not represent the mainstream market, and validating with early users may mislead the direction for scaling
  4. Quality threshold issue: Industries such as finance, healthcare, and security have high MVP quality requirements; a poor-quality MVP will permanently damage brand credibility

Research sources (proportion of primary sources must be >50%)

# Source Type Primary/Secondary Weight
1 Ries, The Lean Startup (2011) Book Primary High
2 Blank, Four Steps to the Epiphany (2005) Book Primary High
3 Maurya, Running Lean (2012) Book Primary High
4 Hsieh, Delivering Happiness (2010) Book Primary Medium
5 Dropbox MVP Case Study (2011) Case Primary Medium
6 Buffer MVP Blog Post (2010) Blog Primary Medium
7 TechCrunch: Color Failure Report (2011) Report Secondary Medium

This Skill is generated by Forge Skill — Forging Thinking Tools Generated by