Review directions and accountability

Reviews are channels, not ceremonies

A review is a channel through which an organization lets one account of work meet another account with enough authority, evidence, and time to change what happens next. The useful question is not simply whether a review occurred. It is whether the review could alter the decision, the local plan, the resources, the authority boundary, the interface with another team, or the organization's understanding of reality.

The three directions name different accountability problems:

• Upward review carries local reality, risks, dissent, and requests toward the people who control purpose, resources, priorities, or exceptions.
• Downward review carries intent, constraints, coaching, resources, and consequences toward the people who execute or hold local knowledge.
• East/west review carries assumptions, dependencies, evidence, and commitments across peers who share a customer, system, resource, or outcome but do not report to one another.

These are directions of a review relationship, not permanent judgments about who is superior. A team can review an executive upward on strategy, receive a downward review on an agreed operating standard, and run an east/west review with a neighboring team in the same week.

What the Atlas examples make visible

The Great Leap Forward reporting system is a failure case for upward review. Production reports moved upward, but the measurement also signaled political loyalty; contradiction could therefore be punished rather than used to correct the plan.¹ The lesson is not that upward reporting is inherently good. It is that a channel cannot carry truth when the sender's safety depends on making the authority's story look true.

The Challenger launch decision shows a related failure: engineers' concern entered a formal review and still lost practical authority when the management process converted opposition into a launch recommendation.² An upward review needs a protected route for dissent, an explicit response, and a named person who must explain why a material warning did not change the decision.

The Mission Command and L. David Marquet entries show the downward problem from the other side. Clear intent can travel downward while judgment remains near the changing conditions, but only when competence, boundaries, and escalation are real rather than slogans.³⁴ A downward review should therefore inspect whether the recipient has enough purpose, authority, resources, and stopping conditions to carry the work—not merely whether the instruction was transmitted.

The Team of Teams and Japanese quality control circles entries point toward east/west review. Peer coordination is not a softer form of hierarchy; it is a way for interdependent operators to expose incompatible assumptions before the customer, system, or shared mission pays for them. The limitation is equally important: a peer forum can create alignment while still leaving a shared blind spot or excluding people who have no seat at the table.

A compact review record

Every review direction can use the same durable record. The direction changes the questions and the authority being asked to respond.

review type: upward | downward | east/west
subject and owner:
reviewer(s):
decision or work affected:
feedback window:

what is known:
what is inferred:
what is unknown or disputed:
what changed since the last review:

hard question:
material risk or dissent:
requested response:
who can change the outcome:

decision / guidance / coordination commitment:
owner:
due date:
evidence that will show whether it worked:
follow-up review:

The record should preserve the original claim, the response, and the resulting change. A summary that records only the final decision destroys the most useful part of review: what the organization knew, discounted, or failed to ask at the time.

Direction-specific templates

Upward: reality, risk, and authority

Use when the people closest to the work need a decision, resource, exception, or correction from a higher authority.

# Upward review: [subject]

## The reality we are seeing

[Observed facts, with source links and time boundary.]

## What we think it means

[Inference, confidence, and plausible alternatives.]

## What is at risk

[Customer, worker, financial, safety, mission, or coordination impact.]

## What we have already tried

[Local actions and their results.]

## What we need from authority

[Decision, resources, priority change, exception, protection, or escalation.]

## Dissent or minority account

[Who disagrees, what evidence they hold, and what remains unresolved.]

## Response required

[Accept, reject with rationale, investigate, or defer until a named date.]

Downward: intent, support, and accountability

Use when a responsible authority is setting direction or reviewing execution by people with local knowledge.

# Downward review: [subject]

## Purpose and intended outcome

[Why this matters and what success looks like.]

## Authority granted

[What the operator may decide without further approval.]

## Constraints and stop conditions

[Non-negotiables, budget, safety, policy, and escalation triggers.]

## Support supplied

[People, tools, context, time, and access.]

## Local judgment requested

[What the operator must interpret rather than merely execute.]

## Evidence of execution

[What happened, what changed, and what the operator learned.]

## Coaching, correction, or authority change

[Specific feedback and whether the operating envelope changes.]

East/west: interfaces, dependencies, and shared outcomes

Use when peers own different parts of one chain of work.

# East/west review: [shared outcome]

## Shared outcome

[Customer or system result both teams are responsible for.]

## Interface contract

[Inputs, outputs, timing, quality bar, and ownership boundary.]

## Assumptions to test

[What each team believes the other team will provide or decide.]

## Evidence and exceptions

[Signals that the handoff, dependency, or shared metric is failing.]

## Trade-offs and unresolved disagreement

[What cannot be optimized simultaneously.]

## Coordination decision

[Commitment, owner, date, and escalation path if the interface fails.]

The tactical integration for Workloop

Reviews should become typed work objects rather than an undifferentiated comment stream. At minimum, a review needs:

1. a direction and a named relationship between reviewer and reviewed work;
2. a purpose: inform, discuss, decide, accept, correct, or learn;
3. a claim or artifact under review, with evidence and uncertainty visible;
4. a response authority who can change something consequential;
5. a disposition for each material point: accepted, rejected with rationale, deferred, or escalated;
6. a follow-up condition and date;
7. an outcome record that can update doctrine, authority, a runbook, or the next execution.

This makes the distinction among review, approval, acceptance, and learning operational. A peer can review a handoff without approving a budget. An executive can review a risk without taking over execution. A beneficiary can accept or reject an output without becoming the person who designed it. A post-action review can change the next run without rewriting what was known at the time of the original decision.

The first product surface should be a small review queue attached to the work object: direction, purpose, what needs judgment, evidence, reviewer response, and next safe action. Raw discussion, provider details, and full history can remain one layer deeper. The smallest useful workflow is:

open review → collect evidence → state response authority → review →
record disposition → change work or authority → schedule follow-up → learn

The lens remains a Workloop hypothesis, not a claim that every organization should maximize all three directions. More review can slow work, chill candor, or create a false sense of control. The test is whether the review makes a material account more able to reach the authority that can respond, while preserving the agency and privacy of the people who supplied it.