The Twenty Minute VCMike Duboe: Top 5 Lessons Scaling Stitch Fix to IPO; Why CAC/LTV is a BS Metric | 20VC #978
CHAPTERS
- 0:00 – 0:27
Cold open: The experiment one-pager that prevents “revisionist history”
Mike opens with a concrete operating artifact: a lightweight experiment one-pager that forces clarity before anything ships. The emphasis is on pre-defining success and next steps so teams don’t rationalize ambiguous results after the fact.
- •Simple one-pager template for any experiment idea
- •Required fields: objective, hypothesis, design/resources, timeline
- •Pre-defined quantitative success criteria
- •Documenting “what we’ll do if it wins/loses” to avoid post-hoc spin
- 0:27 – 2:22
Mike Duboe’s path into growth: engineering → Bain → early startup operator
Mike explains how optimizing for learning shaped his career choices, from engineering to consulting to small-team startup work. Early marketplace/product experiences became his on-ramp into what later became formal “growth.”
- •Career principle: avoid routine, maximize learning curves
- •Bain trained structured thinking but lacked personal passion alignment
- •Move to Bay Area and search for food/marketplace startups
- •Kitchit as first real ‘generalist athlete’ operating role
- 2:22 – 4:40
Tilt to Stitch Fix: from early growth chaos to scaling a profitable acquisition engine
Mike recounts joining Tilt during rapid growth, then transitioning to Stitch Fix when the company needed to build an in-house acquisition muscle later in its lifecycle. He frames Stitch Fix as a counterpoint to ‘high burn, high growth’ lessons from Tilt.
- •Cold email led to joining Tilt post-YC and a16z round
- •Role at Tilt: understand why growth was happening and accelerate it
- •Tilt experience included mentorship and eventual failure lessons
- •Stitch Fix: profitable constraints, late need for acquisition capability
- 4:40 – 7:30
Scaling Stitch Fix: two foundational lessons—objective functions & incrementality
Mike’s biggest Stitch Fix takeaways: aligning teams on the correct objective function and measuring paid marketing with incrementality, not false precision. He warns that siloed KPIs can optimize locally while harming the business globally.
- •Separate acquisition/retention KPIs can create destructive incentives
- •Redefine KPIs to down-funnel metrics that reflect true growth
- •Paid measurement should focus on lift via holdout tests
- •Attribution models often create false precision and bad decisions
- 7:30 – 9:54
What a Head of Growth actually is: accelerating learning + engineering control
Mike offers a broad definition of growth leadership that goes beyond channel management. The job is to increase the organization’s pace of learning and build systems that give the company more control over its North Star metrics.
- •Core responsibility #1: accelerate the company’s learning cadence
- •Core responsibility #2: build systems that control North Star metrics
- •Growth as experimentation with a faster ‘drumbeat’ than the org
- •Applies across product, marketing, and sometimes sales contexts
- 9:54 – 13:11
Where growth belongs in the org: single-function vs cross-functional pods
The conversation moves to org design: when growth should be a marketing-like function and when it must be a cross-functional product discipline. Mike argues growth works best under product (or CEO at times) when it requires engineering leverage.
- •Growth should ideally become ‘how the company runs,’ not a silo
- •Two common structures: channel-led acquisition org vs cross-functional pod
- •Cross-functional pods reduce dependencies and increase autonomy
- •Reporting options: CEO for leverage; product for sustainable growth
- 13:11 – 13:49
When to make the first growth hire: never before PMF, start with analytics
Mike outlines timing mistakes founders make—especially hiring “growth” to find PMF. He argues the earliest durable leverage is analytics capability that can answer basic cohort and retention questions quickly and reliably.
- •Do not hire a growth leader pre-product-market fit
- •Early ‘generalist athletes’ should focus on user learning, not scaling
- •First “growth hire” is often an analytics/SQL-oriented operator
- •Avoid building top-of-funnel before you understand retained cohorts
- 13:49 – 16:37
Analytics debt: why dashboards don’t solve bad instrumentation
Mike defines analytics debt as the high cost of answering basic business-performance questions. He emphasizes instrumentation (capturing the right events) as the prerequisite—dashboards are useless if the underlying event model is wrong.
- •Analytics debt = basic questions are slow/costly to answer
- •Dashboards (GA/Looker/Mixpanel) aren’t enough: ‘garbage in, garbage out’
- •Instrumentation means logging key user events inside the product
- •Without event capture, you can’t diagnose or run effective growth work
- 16:37 – 18:22
Who does analytics best: high signal-to-noise and explicit growth models
Pressed for examples, Mike points to Faire as a company with strong analytics habits early on. The hallmark: a small set of high-signal metrics rooted in a clearly articulated growth model.
- •Facebook as a late-stage analytics archetype (but not the only answer)
- •Faire highlighted for early clarity and dashboard discipline
- •Great analytics focuses on the few metrics that truly matter
- •Growth model clarity propagates into dashboards and operating rhythms
- 18:22 – 20:57
Finding your North Star metric by writing the growth model first
Mike advises founders to start by describing how the product grows—conceptually before spreadsheets. North Stars evolve, but a clear growth model forces teams to connect user actions to compounding outcomes.
- •North Star changes as a company matures
- •Start by writing a conceptual growth model (pre-spreadsheet)
- •Examples: content loops (Pinterest) and system mechanics
- •Harry shares 20VC’s ‘great guests’ loop as a practical illustration
- 20:57 – 28:20
Hiring for growth: defining success, calibrating, and asking the right questions
Mike lays out a practical hiring sequence: define what success means, calibrate against similar companies, and write a spec that both sells the role and filters for fit. He then shares interview questions that reveal systems thinking and learning orientation.
- •Step 1: define success outcomes (systems/KPIs) before hiring
- •Step 2: calibrate by speaking with leaders from similar growth patterns
- •Good job specs market the moment + clarify the real challenges
- •High-signal questions: ‘How does your favorite product grow?’; ‘Tell me about a failure’; ‘Show me your daily dashboard’
- 28:20 – 31:25
Growth hiring exercises: experiment roadmaps and growth models (with time limits)
The discussion turns to take-home or timed exercises that test prioritization and holistic thinking. Mike prefers tasks that reveal simplification skill, loop-based reasoning, and ability to propose a coherent first 90-day plan.
- •Exercise: prioritized experiment roadmap for a specific goal/feature
- •Avoid ‘laundry list A/B tester’ mindset—test prioritization and ROI logic
- •Exercise: write a growth model for a comparable product/company type
- •Give candidates time; watch for overly complex or purely financial models
- 31:25 – 35:26
Compensation and the Keith Rabois ‘barrels vs ammunition’ framing
Mike frames great growth hires as potential “barrels”—multipliers that increase how much the company can execute in parallel. He then offers rough Series B compensation ranges and urges founders to be generous with equity for true leverage hires.
- •Barrels vs ammunition: barrels determine parallel execution capacity
- •Great early growth leaders can become ‘first barrels’ who hire more barrels
- •Series B-ish range: ~$150k–$250k salary (sometimes higher), meaningful equity
- •Equity guidance: roughly ~0.5% to 1%+ depending on seniority and dilution
- 35:26 – 39:28
Loops vs funnels: why compounding systems beat linear optimization
Mike defines growth loops and contrasts them with funnel thinking (AARRR). He argues funnels encourage silos and decaying returns, while loops force teams to design compounding systems that integrate product, channels, and monetization.
- •Loop definition: closed system where outputs feed back as inputs
- •Funnels (AARRR) are linear and often create organizational silos
- •Loops usually require product/engineering leverage, not just marketing
- •Operationalizing loops = embed distribution into product strategy
- 39:28 – 45:16
Building an experimentation operating system: intake, triage, reviews, and knowledge sharing
Mike describes the mechanics of running growth as an experimentation machine while keeping the org aligned. The key is lowering friction for ideas, using structured templates and prioritization, and broadcasting learnings so other teams compound them.
- •Principles: best ideas can come from anywhere; failure = not learning
- •Mechanics: experiment one-pagers + centralized prioritized spreadsheet
- •Triage framework: ICE (Impact, Confidence, Effort)
- •Cadence: weekly experiment reviews; share major learnings at all-hands
- 45:16 – 55:29
Paid marketing done right: accelerant not crutch, diversify at scale, measure incrementality
Mike shares paid marketing principles from Stitch Fix: avoid early addiction, expect performance degradation with scale, and lean on incrementality testing over attribution stories. He gives a detailed TV holdout-test example that reveals channel interactions and true CPA.
- •Paid as a PMF health check: turning it off shouldn’t kill the business
- •Paid can distract from harder cross-functional product growth initiatives
- •At scale, diversify channels to reduce concentration risk (e.g., Facebook shocks)
- •Incrementality via holdouts beats last-click or black-box attribution
- •Stitch Fix example: TV local holdout test to measure halo and true incremental CPA
- 55:29 – 58:19
Why CAC/LTV is ‘BS’ (or at least flawed): lifetime assumptions, lack of granularity, and better alternatives
Mike critiques CAC/LTV as overly seductive and often misused to justify bigger budgets. He proposes payback period (using incremental paid CAC) as a more practical, granular decision metric, and echoes Bill Gurley’s warning that LTV is a tool—not a strategy.
- •LTV ‘lifetime’ is unknowable early and invites bad extrapolation
- •Blended CAC/LTV hides major variance by channel/keyword/audience
- •Prefer payback period and compute using incremental paid CAC
- •Granularity matters: different intent segments justify different CAC thresholds
- •Bill Gurley: ‘Dangerous Seduction of the LTV Formula’—don’t confuse tool with strategy
- 58:19 – 1:01:55
Quick-fire: mistakes hiring growth, Tilt lessons, and standout growth strategies (Faire referrals)
In the closing rapid round, Mike revisits how growth can mask problems, the top hiring mistakes, and a key Tilt lesson on over-incentivizing behavior. He ends by praising Faire’s B2B referral mechanics as an unusually powerful growth engine.
- •Changed belief: ‘growth solves everything’—it can also mask issues
- •Hiring mistake: weak referencing + mis-scoping growth vs marketing/product
- •Tilt lesson: over-incentivizing usage creates dependency on promos
- •Wish for growth: fewer micro-optimizations, more holistic systems and big bets
- •Most impressed by: Faire’s cross-side B2B referral engine
