The Decision Stack
How Amazon, Google, Netflix, Spotify and others have turned decision-making into a competitive advantage.
The best-run companies in the world share one habit.
They don’t just make decisions. They engineer how decisions get made, who owns them, how they’re structured, when they’re logged, and how they’re learned from over time.
And each became world-class at exactly one part of that process.
Amazon writes six-page narratives before any major call. Spotify maps every product bet to a data chain. Netflix names a single “Informed Captain” accountable for every outcome. Uber runs an RFC process that turns decisions into a searchable institutional archive. Shopify classifies every choice by reversibility before deciding how much rigor to apply.
This isn’t process for process’s sake. It’s a compounding advantage. Teams that treat decisions as first-class objects, documented, tracked, reviewed, make faster, smarter calls over time. The decision log becomes a flywheel. Some teams literally measure their decision velocity by analyzing timestamps in their logs and optimize it like any other metric.
At Jupi, we’ve spent a lot of time studying how these companies built their decision infrastructure, because we’re building the tool that makes it accessible to every team, not just ones with the engineering resources of a Netflix or an Amazon.
What follows is two things:
a synthesis of what this body of literature actually reveals when read together
a curated list of the best public resources we’ve found, company playbooks, practical templates, and academic research. No filler.
What they all got right, and what they all missed
Read all of this documentation together and the same gap appears every time.
Every company optimized for their context. Amazon is a writing culture, so they formalized structuring. Netflix decentralized fast, so they formalized ownership. Uber scaled engineering, so they formalized memory. Each solution was perfect for one fragment, in one organizational moment.
But a decision doesn’t stay in one fragment. It moves. A signal gets captured, structured into options, evaluated against beliefs, advanced to a call, executed, and then, if you’re disciplined about it, remembered in a way that makes the next decision faster.
None of these companies documented the full cycle. They documented the fragment they were struggling with.
What if you were good everywhere?
That’s the hypothesis behind Jupi: decision infrastructure that covers the full cycle, Capture, Structure, Evaluate, Advance, Execute, Remember, without every team having to build it themselves, piece by piece, the way these companies did over years.
What you’re reading below is the documentation left by those who cleared the path. What we’re building is the road.
1 — Tech company playbooks
Microsoft Engineering Playbook
Microsoft requires engineering teams to track decisions as structured ADRs, stored in version control alongside the code they describe.
Decisions that live next to the work they shaped are actually consulted. That’s the key behavioral insight.
AWS ADR Process
Amazon Web Services formalizes the decision lifecycle: Proposed, Accepted, Deprecated.
The status field alone changes behavior. Teams stop relitigating closed decisions.
Google — AI-assisted decision-making
Google Workspace outlines how gen AI is being used to surface past decisions, summarize deliberations, and flag alignment gaps across distributed teams.
The early signal of what decision infrastructure looks like with AI in the loop.
Google Cloud — Architecture Decision Records
Minimal, actionable, opinionated. Google’s take on when to write an ADR and when not to.
Knowing when NOT to document is as important as the format itself.
Spotify Engineering — When to write an ADR
Spotify’s guidance: write one when a decision is hard to reverse, broadly impactful, or carries long-term consequences. The rest? Just decide.
This triage logic is what separates a useful decision log from a bureaucratic one.
Netflix — Decision-making framework
Decentralized authority, logged decisions, validated through experimentation. Netflix’s framework is what “freedom and responsibility” looks like when it’s operationalized, not just written on a slide.
The Informed Captain model is the cleanest solution to decision-by-committee we’ve seen.
Shopify — TOMASP
Trust, Ownership, Mission, Accountability, Speed, Pragmatism. A tactical model for fast, high-quality decisions.
The key idea: classify before you deliberate. Reversible decisions need momentum, not rigor.
Uber — RFC-based decision system
Uber scaled to thousands of engineers by making decisions asynchronous and searchable.
Their RFC archive is institutional memory that compounds. The org learns faster because it forgets less.
2 — Templates, tools, and practical guides
adr.github.io
The open-source canonical reference for ADR templates, CLI tooling, and community patterns.
Start here if you’re implementing from scratch. The community has done the heavy lifting.
Atlassian — Decision Journal
Atlassian’s case for writing down not just what you decided but why you expected it to work.
The post-mortem data this creates is underrated. You can’t improve what you don’t measure.
Fellow — Ultimate Guide to Collaborative Decision-Making
The most practical long-form guide we’ve found on structuring team input without losing velocity.
Covers the tension between consensus and speed better than anything else in this list.
Notion — Lightweight ADR template
A clean starting point for teams that want structure without overhead.
Good enough to start with today. You can always formalize later.
Wrike — Decision logs for project visibility
A project management lens on decision logging: accountability, stakeholder alignment, audit trail.
Useful for product and ops teams who need decision logs without an engineering-first framing.
Coda + Product School — Decision log with Slack integration
Real-time decision capture without leaving the workflow.
The Slack-native pattern is worth studying for any team that lives in async communication.
3 — Research and thought leadership
HBR — Faster, Better Decisions
A checklist framework grounded in research: timing, stakeholders, alternatives, rationale, post-mortem.
Translates cleanly into any decision log format. Still the best single-page reference.
Cloverpop — The case for decision discipline
Data showing structured decision logging leads to more inclusive outcomes, faster execution, and clearer accountability.
The ROI argument for formalizing, useful if you’re trying to convince leadership.
IEEE — Architecture as a set of decisions
The foundational paper. Software architecture isn’t diagrams, it’s the accumulation of key decisions and their rationale.
This reframe changes how you think about documentation entirely.
Keeling — ADRs in Agile Teams
How ADRs became a lightweight alternative to heavyweight documentation in fast-moving teams.
Bridges the gap between agile velocity and architectural rigor.
Zimmermann et al. — Sustainable architecture decisions
Principles for keeping decisions maintainable as systems evolve.
The insight on “decision churn” is worth the read alone. Most teams have no idea how much they’re paying for it.
The road ahead
Every company on this list built their decision infrastructure by hand, custom processes, internal wikis, homegrown tooling, engineering culture built over years.
Most teams don’t have that runway.
That’s what Jupi is for. Decision infrastructure, out of the box, structured decision objects, The Loop (Capture, Structure, Evaluate, Advance, Execute, Remember), institutional memory that compounds automatically.
If you’re thinking seriously about how your team makes decisions, I’d love to hear about it.