4. Lean Startup Principles
• Most startups fail because they waste $$ building
and marketing a product that no one wants
• Instead, formulate hypotheses about business
model, then launch early with a minimum viable
product, i.e., smallest feature set needed to test a
hypothesis
– Goal: validated learning via rigorous experiments
– Ries: “It’s not a test if you can’t fail”
• Pivot when a hypothesis is disconfirmed, i.e., revise
hypothesis then rapidly test it with another MVP
• Repeat process until hypotheses are fully validated and
you have product-market fit
• Don’t scale aggressively until you have PMF
4
5. Lean Startup Myths
• Lean = bootstrapped
• Lean = crappy product
• Lean = driven by customer feedback
• Lean = new approach
• Lean = only for web/mobile products
• Lean = only for startups
5
6. Hypothesis-Driven Entrepreneurship
Pivot
Viable
Feasible Perish
Desirable
Product-Market Fit:
Proceed with Scaling
Generate
Envision Test Hypothesis
Business Persevere with Next Test
Venture Using Minimum
Model
Concept Viable Product
Hypotheses
6
7. Business Model
An integrated array of distinctive choices
specifying a new venture’s unique customer
value proposition and how it will configure
activities—including those of its partners—to
deliver that value and earn sustainable
profits.
7
8. Business Model
Customer •Problem: unmet needs served
Value •Solution: launch MVP; product roadmap
Proposition •Target Customer Segments: early adopters and beyond
•Willingness to Pay vs. Total Cost of Ownership (“whole product”)
•Basis for sustained differentiation and/or cost advantage
•Complements needed: who will provide, under what terms?
•Penetration vs. skimming price
•Switching costs
Technology & •In-house vs. outsourced activities
Operations •Value proposition for key suppliers, complement providers
•Intellectual property protection
Go-to-Market •Mix of direct and indirect channels; margin/rights for partners
Plan •Mix of free vs. paid customer acquisition methods; cost for each
•Customer Lifetime Value vs. Customer Acquisition Cost (LTV/CAC)
•Incentives to race for scale, e.g., network effects, switching costs
Profit •Size, growth of Total Addressable Market
Formula •Variable contribution margin
•Fixed costs; breakeven sales as % of TAM
•Investment (PPE & WC)/sales ratio
•Cash flow profile: maximum need; net positive point 8
9. Falsifiable, Specific Hypotheses
• Falsifiable: can be rejected via test
(Ries: “It’s not a test if you can’t fail”)
• Specific: not “Product will spread by word of
mouth”; rather “Viral coefficient > 0.4”
• Measurable: Ideally, hypotheses require
quantitative measures for validation
9
11. Specify MVP Tests and Launch!
• Minimum Viable Product: Smallest set of
features/activities needed to test a business
model hypothesis
• “Launch early and often” i.e., ASAP, put a real
product in the hands of real customers in a
real world context
– Fast cycles/small batches reduce waste by
accelerating feedback and making it easier to
diagnose/fix problems
11
12. Launch Early?
Why? Why Not?
• With high uncertainty about • Disruption of mission-
both problem and critical
solution, must start learning activity, e.g., Dropbox
ASAP • Early adopters’ needs do
• Must observe radically new not match mainstream’s
product in use to refine it • Reputational risk, esp. with
– Corollary: less lean value with viral products and
“better/faster/cheaper” concentrated markets
products
– Early users = small % total
• Stealth rarely needed – Can use trial
– Ries: Try to get big company brand, e.g., Textbookflix
to steal idea
12
13. MVP Design
• Reduce product functionality to test hypotheses
about “need to have” features before building “nice
to have” features
– Example: foursquare alpha
• Rely on temporary/makeshift operations if doing so
doesn’t impact feedback quality
– Examples: Aardvark turk testing; RTR PDF test
• Smoke test: offer product that doesn’t yet exist, via
landing page test, video MVP, letter of intent, etc.
– Better for rejecting hypotheses than validating them
13
14. Partial vs. System Tests
• Partial test of a single “known unknown”
– Examples: supplier letter of intent; probation for new hire
– Give priority to tests that eliminate lots of risk at a low
cost, e.g., patent search
– Issue: run partial tests in series or in parallel? Parallel tests can
speed time to market, but this risks wasted effort if H1 must be
discarded due to failure of H2 test
• System test of entire model, at reduced scale
– Reveal “unknown unknowns”
– Explore interactions between variables
• RTR adoption odds increase sharply after a consumer has had 7-8
interactions with brand, including PR and word-of-mouth referrals
• Consumer adoption of RTR depends on designer adoption, and vice versa.
14
15. Test Design
• When does validation require passage of time?
– Retention/repurchase/referral rates
– Mainstream adoption requiring referrals
• Firms face Catch-22 when validation requires
scale, and scale requires validation
– Cannot test demand if value depends on
network size
– User base determines number of tests
possible, which determines feature
set, which determines user base…
15
16. Risk: False Positives and Negatives
• FP: sample only enthusiastic early
adopters
• FN: reaction to badly-built prototype
rather than the venture concept
– If RTR had started with PDF test, it might
have offered wrong dress assortment and
observed false negative
16
17. Lean Startup Data Sources
• Qualitative
– Customer Interviews
– Focus Groups
– Concierge MVP
– Usability Tests
– Customer Service Interactions
• Quantitative
– Customer Surveys
– Smoke Tests: Landing Pages, Letters of Intent
– A/B Tests
– Funnel/Cohort Analysis
– Viral Coefficient
– CAC vs. LTV
– Net Promoter Score
17
18. Pivot Lessons
• Pivots are NOT a goal!
• Pivots are costly, especially with a strong
reality distortion field
– Employees, investors, partners are sold on vision
– Entrepreneur has ego invested in vision
• It’s possible to pivot too quickly
• Post-PMF, pivots continue as market evolves
(e.g., Dropbox, Chegg?)
18
19. Lean Psychology
• Be ready for surprises, including information
not generated by tests
• Be wary of cognitive biases and remain open
to disconfirming data
– Optimism bias, planning fallacy, confirmation
bias, sunk cost fallacy
– Ego-defensive behavior can keep entrepreneur
from acting on disconfirming data
19
20. Some Limitations of Lean Logic
• Cannot “launch early and often” when:
– Mistakes must be limited (e.g., pacemaker software)
– Product development cycles are intrinsically long due to
science/engineering challenges (e.g., clean
tech, “Chunnel”)
• Less need to launch early/often with low demand risk
– Strong unmet demand (e.g., cancer cure)
– “Me, too” product (but, must still test execution capability)
– Founder with domain expertise (but, be wary of success-
induced biases)
20
21. B2B: Fewer Pivots
• Less feedback available
– Longer cycles due to multi-level process
– Fewer cycles due to costly system integration, training
– Fewer trial candidates
• Less feedback needed
– Failure risks mandate a more fully-evolved product
– B2B products more likely to serve existing market (?)
• Issue: Will more B2B products cross over from
B2C, like Dropbox (also, WiFi, iPhones)?
21
22. To Learn More…
• Business model analysis
– Osterwalder’s Business Model Generation
– My posts at http://bit.ly/oiWqPJ
• Lean startups
– Ries’s The Lean Startup; blog
– Blank’s Four Steps to the Epiphany; blog
• Startup management practices: my reading
lists at http://bit.ly/f9vSyP
22