OpenLabor Jeff Bezos - Operations Flywheel
Use these frameworks when designing operations, setting up metrics, organizing teams, or improving decision-making velocity. Apply the flywheel model to map self-reinforcing business loops. Replace output metrics with input metrics. Structure teams for speed and autonomy.
Routes when user asks about: flywheel effect, business model loops, input metrics, output metrics, high-velocity decisions, disagree and commit, two-pizza teams, operational excellence, Amazon operations, how to build self-reinforcing growth, measurement systems
How to Use This Skill
When organizing workflows, designing processes, setting up metrics, or restructuring teams, apply the relevant framework below. The flywheel is the strategic design tool; input metrics are the measurement system; decision velocity and team structure are the execution layer.
Designing Flywheels
Every business should map its virtuous cycle. Reference the Amazon flywheel (Bezos's napkin sketch, 2001):
Lower Prices → More Customers → More Volume → More Sellers Attracted
→ Wider Selection → Better Customer Experience → Lower Prices (loop)
AND simultaneously:
More Volume → Lower Cost Structure → Lower Prices
Every element reinforces every other element. Push on any part and the whole thing accelerates.
Flywheel Design Process
Step 1 — Identify core customer value: What single thing does the customer value most?
Step 2 — Map what creates that value: What inputs produce that customer value? (Lower cost structure → lower prices)
Step 3 — What does customer satisfaction produce?: More customers → more volume → more leverage → lower costs? Network effects? More data?
Step 4 — What does more volume produce?: Lower unit costs? Attract suppliers/partners? Data advantage?
Step 5 — Connect the loop: Draw the full virtuous cycle. Verify each element genuinely feeds the next.
Step 6 — Identify the bottleneck: Where is the flywheel weakest? What constraint, if removed, would accelerate the whole loop?
Flywheel Quality Checklist
- Each element in the loop causes the next — test causality, not just correlation
- The loop returns to the starting point (actually circular)
- Adding more to any element accelerates the entire loop
- The flywheel creates a self-reinforcing moat over time
- Competitors cannot easily enter the loop at any point
Input Metrics vs. Output Metrics
Most companies measure outputs. Bezos insisted on measuring inputs — this is the single most important operational insight.
The Distinction
Output metrics (results): Revenue, profit, customer satisfaction, NPS, churn rate
- Problems: lagging indicators (learn about problems too late), can be gamed, don't tell you what to fix
Input metrics (drivers): First-contact resolution rate, in-stock rate, delivery speed, selection breadth
- Advantages: leading indicators, tell you exactly what to change, harder to game
"We need to be able to know whether we are delivering a great customer experience, and we need to be able to measure that directly — not through proxies." — Bezos
Input → Output Mapping Template
| Output Metric | Driving Input Metrics |
|---|---|
| Revenue | Leads → conversion rate → avg. order value |
| Customer satisfaction | First-contact resolution rate, delivery time, defect rate |
| Retention/NPS | Onboarding completion, feature adoption, support ticket rate |
| Profit margin | Unit economics, operational efficiency, supplier terms |
Amazon's Input Metric Examples
- In-stock rate (% of items available)
- Detail page view weight (% of views that convert)
- Perfect order percentage (orders with no problems)
- Contact rate (customer contacts per unit shipped — rising = problems, falling = improving)
- Glance views (how often items are seen)
Building Your Input Metric System
- List top 5 output metrics
- For each output: what is the single most important input that drives it?
- Is this input measurable weekly or daily (not monthly/quarterly)?
- Build dashboards around inputs, not outputs
- Review input trends in every operational meeting
High-Velocity Decision Making
"Speed matters in business. Many decisions and actions are reversible and don't need extensive study." — Bezos
Why Companies Slow Down
They apply the same process to all decisions. Type 1 (irreversible, high stakes) process applied to Type 2 (reversible, low stakes) decisions kills velocity.
The 70% Rule
"Most decisions should probably be made with somewhere around 70% of the information you wish you had. If you wait for 90%, in most cases, you're probably being slow."
Going from 70% to 90% certainty takes 3x longer but only improves quality by ~20%. For reversible decisions, this is a bad trade.
Apply the 70% rule to: product features, marketing decisions, operational changes, hiring for most roles, pricing tests, partnership pilots.
Do NOT apply to: major acquisitions, irreversible architecture, legal/regulatory commitments, large capital commitments.
Decision Velocity Checklist
Before any decision:
- Is this reversible? If yes → decide now with 70% information
- Who is the right decision-maker? Person closest to the problem — not necessarily most senior
- Do we have 70%? If yes → decide
- Cost of 30-day delay? If high → decide now
- Failure modes if wrong? If recoverable → decide now
Disagree and Commit
"If you have conviction on a particular direction even though there's no consensus, it's helpful to say, 'Look, I know we disagree on this but will you gamble with me on it?'" — Bezos
The Problem This Solves
Consensus cultures: slow to the pace of the most reluctant person, produce watered-down decisions, create passive resistance, mistake silence for agreement.
The Protocol
- Voice disagreement clearly: "I disagree because [specific reason]. I think we should [alternative] because [reasoning]."
- Get genuinely heard: Decision-maker must engage with the objection, not dismiss it.
- Decision-maker decides: Once the argument is heard, make the call.
- Commit fully: "I disagree, but I'm committed to making this succeed. Here's how I'll support it."
What it is NOT: doing the minimum, bringing up objections later, letting it fail to say "I told you so."
When to Use
- Technical architecture disagreements where one approach must be chosen
- Product prioritization when two valid paths exist
- Any decision where continued debate costs more than the risk of being wrong
Two-Pizza Teams
"If you need more than two pizzas to feed a team, it's too large." — Bezos
Design Rules
Team size: 5-8 people maximum.
Structure: One owner with clear accountability. All necessary skills represented. Dedicated resources — not borrowed.
Scope: Team owns a coherent domain end-to-end. Clear APIs to other teams. Success metrics the team controls.
Autonomy: Team makes its own operational decisions. Escalation for conflicts between teams, not within.
Team Size Diagnostic
| Symptom | Likely Cause | Fix |
|---|---|---|
| Meetings > 8 people regularly | Team too large | Split the team |
| Nobody knows who owns a decision | Unclear scope | Redefine charter |
| Can't ship without waiting for another team | Dependency problem | Bring skill in-house or restructure |
| Individual contribution invisible | Team too large | Split or restructure |
Operational Excellence Audit
Run quarterly to assess operational health.
Flywheel Health
- Can you draw your flywheel in 5 minutes?
- Are all loop elements measurable?
- Is the flywheel accelerating quarter-over-quarter?
- Is the bottleneck identified and being addressed?
Metrics Quality
- Do you have daily/weekly input metrics for top 5 outputs?
- Are input metrics reviewed in operational meetings?
- Is there a clear owner for each input metric?
- Has any metric been gamed recently? (metric improving while experience degrades)
Decision Velocity
- Average time from "idea raised" to "decision made" for Type 2 decisions?
- What % of decisions made at the right level (not escalated unnecessarily)?
- How many decisions delayed for consensus when they only needed alignment?
Team Structure
- Are all teams genuinely two-pizza sized?
- Does each team have a clear owner and charter?
- Can teams ship independently?
Sources
- Amazon Shareholder Letters 1997-2020
- Working Backwards — Colin Bryar & Bill Carr
- Good to Great — Jim Collins (flywheel concept origin)
- Amazon Leadership Principles (amazon.jobs)
- Bezos interview at re:Invent 2021
- The Everything Store — Brad Stone