Post
JA EN

Implementation Guide for ADR / Pitch / Kickoff Memos: From 'Why Write?' to Operational Design

Implementation Guide for ADR / Pitch / Kickoff Memos: From 'Why Write?' to Operational Design

Overview

ADRs (Architecture Decision Records), Pitch documents (Shape Up), and project kickoff memos are usually discussed in isolation. But viewed from the higher purpose of organizational context supply capability, they are the same purpose at different granularities.

This article extends STEP 4 of the sister piece “Implementation Guide for Organizational Context Supply Capability” into a standalone deep dive. It organizes the three document types in a Diátaxis-like layout and treats “why write,” “where to put them,” “when to retire them,” and “how to keep them searchable” as one coherent operational guide.

Roles and granularity of the three documents

DocumentGranularityPrimary purposeOrigin
ADRSingle judgment (one file per decision)Preserve the context and reasoning behind past decisionsMichael Nygard 20111
Pitch documentOutline of a proposal (pre-commit)Evidence of pre-commit deliberationBasecamp Shape Up 20192
Kickoff memoExecution agreement (at start)Clarify roles and success conditions among stakeholdersPracticed at Stripe and similar companies3

The three are also temporally complementary:

1
Pitch (pre-commit deliberation) → Kickoff memo (execution agreement) → ADR (mid-flight judgment record)

ADR implementation

1
2
3
4
5
# ADR-NNN: <decision title>
## Status: proposed / accepted / deprecated / superseded by ADR-MMM
## Context: what situation requires a decision
## Decision: what was decided
## Consequences: positive impacts / negative impacts / trade-offs / follow-ups

ADR operating principles

  • One decision per file. Don’t bundle several
  • Sequentially numbered and immutable. Don’t edit past ADRs; record supersedes in a new ADR
  • Live in the same repo as the code. Reviewed via pull request
  • Write “we won’t do this” decisions too (“we evaluated Kafka and chose not to adopt” is also valuable)
  • Repurpose-able for business decisions (BDRs: Business Decision Records)

Handling the “too many ADRs” problem

When ADR counts reach the hundreds, searchability degrades. Countermeasures:

  • Tagging (domain / status / related project)
  • Twice-yearly cleanup that explicitly deprecates stale entries
  • Importance labels (critical / important / informational)
  • Combine with architecture-document systems like the C4 model or arc42

Pitch document implementation

1
2
3
4
5
6
# Pitch: <proposal name>
## Problem: whose problem, what kind. One concrete example
## Appetite: how many weeks (fixed time) you're willing to spend
## Solution: outline of the solution (user-experience flow, not detailed implementation)
## Rabbit holes: places where digging into detail is dangerous; pre-decided stance
## No-gos: out of scope (push to next cycle)

The essence of the Pitch

The core of Basecamp / Ryan Singer’s Pitch document2 is fixed time, variable scope: “how many weeks (appetite) we’ll spend” is decided up front, and scope flexes to fit. This is the inverse of the implicit operating model at most organizations (scope fixed, time slips).

The act of writing itself is evidence of having thought it through. If you can’t write it, that’s a signal you shouldn’t start it yet.

Usable even without adopting Shape Up wholesale

You don’t need to adopt the entire Shape Up methodology (6-week cycles / cool-down / betting table) to use Pitch documents. “Write before you commit,” “explicitly state Rabbit holes and No-gos,” “declare a fixed appetite” — those are universally useful.

Kickoff memo implementation

1
2
3
4
5
6
7
## Project name / start date / projected end
## Problem statement
## Success conditions (measurable)
## Failure conditions (abandonment line)
## Stakeholders and roles (DRI / Reviewer / Informed)
## Milestones
## Open questions (unresolved points)

Why “failure conditions” matter

Writing failure conditions and an abandonment line up front is what prevents the later “we have to keep going because of sunk cost” state. This matches the logic of Defensive Pessimism4: simulate bad outcomes in advance and prepare the response.

Connection to Stripe’s writing culture

Brie Wolfson, a former Stripe employee, has documented Stripe’s writing culture3: project kickoff memos, a Google Group for retrospectives, “state” emails, shipped/unshipped catalogs. A quote from Stripe’s CTO captures the essence: “If you invest extra time to communicate ideas in clear, precise writing, you get disproportionate returns, because there are far more readers than writers.

Relationship to the Diátaxis framework

Daniele Procida’s Diátaxis5 classifies technical documentation into four kinds:

  • Tutorial: leads the learner (learning-oriented)
  • How-to guide: steps to solve a problem (task-oriented)
  • Reference: specifications and facts (information-oriented)
  • Explanation: deepens understanding (understanding-oriented)

Where do ADR / Pitch / Kickoff memo sit?

DocumentDiátaxis category
ADRExplanation (why this decision)
PitchExplanation (proposal outline) + entry point to How-to
Kickoff memoReference (record of facts and agreement)

These do not function as onboarding tutorials for new hires (curse of knowledge6 risk). New-hire onboarding needs separate Tutorial / How-to documents. To get a system-level view of architecture, arc427 or a similar systematic template is a useful complement.

Searchability is the whole point

The greatest value of ADR / Pitch / Kickoff memo is being searchable later.

  • Centralize in one repo or wiki. It’s enough to split into folders like decisions/ pitches/ projects/
  • File names = date + slug
  • Tag (project name / domain / status)
  • Twice-yearly cleanup: explicitly mark deprecated entries

Make owner and “next review” date required on every document (auto-stale mark if not reviewed). Visualize read counts (access logs); low-access entries become deprecation candidates.

Anti-patterns

PatternWhat happensCountermeasure
ADR is too abstractRe-reading later, the context is unclearAlways include concrete names and numbers in Context
Pitch becomes long-formMotivation to write erodesStrict 1-2 page limit
“Every decision must be an ADR” mandateRitualizesUse the test “will this be useful as decision material in 3 months?”
Stored in a hard-to-search locationWritten but unreadCo-locate with code or centralize at wiki root
Number of ADRs as a KPIGoodhart runaway, hollows outLook at reference count, not write count
Kickoff memo without failure conditionsSunk cost prevents withdrawalMake failure conditions / abandonment lines mandatory
Treating Pitch as requiring full Shape Up adoptionAdoption bar set too highPitch is usable on its own

Summary

  • ADR / Pitch / Kickoff memo are temporally complementary: Pitch (pre-commit) → Kickoff (agreement) → ADR (mid-flight)
  • One document, one purpose; searchability is the whole point
  • ADR: one decision per file, sequential and immutable, co-located with code, “we won’t do this” included
  • Pitch: fixed time / variable scope; if you can’t write it, don’t start it
  • Kickoff memo: success conditions paired with failure conditions / abandonment lines
  • Sit at Diátaxis Explanation / Reference; new-hire Tutorials are a separate workstream
  • Don’t make volume a KPI (Goodhart runaway)

References

  1. Documenting Architecture Decisions — Michael Nygard, Relevance / Cognitect (2011-11-15). [Reliability: High] ↩︎

  2. Shape Up: Stop Running in Circles and Ship Work that Matters — Ryan Singer, Basecamp (2019). [Reliability: Medium-High] ↩︎ ↩︎2

  3. From kickoffs to retros and Slack channels — Stripe’s documentation best practices with Brie Wolfson — First Round Review (2023). [Reliability: High] ↩︎ ↩︎2

  4. Defensive Pessimism: Harnessing Anxiety as Motivation — Julie K. Norem, Nancy Cantor, JPSP, vol. 51, no. 6 (1986). DOI: 10.1037/0022-3514.51.6.1208. [Reliability: High] ↩︎

  5. Diátaxis: A systematic framework for technical documentation authoring — Daniele Procida (continuously updated). The four-way classification (Tutorial / How-to / Reference / Explanation). [Reliability: Medium-High] ↩︎

  6. The Curse of Knowledge in Economic Settings: An Experimental Analysis — Colin Camerer, George Loewenstein, Martin Weber, JPE, vol. 97, no. 5 (1989). DOI: 10.1086/261651. [Reliability: High] ↩︎

  7. arc42 Template — Gernot Starke (continuously updated). Architecture documentation system. [Reliability: Medium-High] ↩︎

This post is licensed under CC BY 4.0 by the author.