Skip to content

Chapter 5 — Objects of Control

Outcomes are what you want. Objects of control are what you can actually manipulate.

Most bad systems fail because they are designed to control outcomes directly (“increase innovation,” “improve alignment,” “move faster”) without specifying what is being controlled to produce those outcomes.

This chapter makes a hard distinction:

  • Outcomes are goals and results.
  • Objects of control are the levers a system touches every time it runs.

If you don’t name the object of control, your system will default to controlling something accidental—often meetings, documents, or reporting.

The Failure This Chapter Prevents

Observable failure: systems are installed that create activity, but the underlying levers never change.

Symptoms:

  • Lots of planning, no improvement in delivery predictability
  • More dashboards, same surprises
  • New “alignment” cadences, same cross-team conflicts
  • Repeated strategy resets with no changes in investment behavior
  • Artifacts proliferate but decision-making does not improve

Root cause:

  • The system has goals, but no controllable mechanism.

The Core Idea: Systems Don’t Control Outcomes

You can’t directly control:

  • customer satisfaction
  • innovation
  • reliability
  • speed
  • “alignment”
  • quality

You can influence them by controlling proximate objects like:

  • constraints (WIP limits, SLAs, timeboxes)
  • interfaces (APIs, contracts, ownership boundaries)
  • work items (definition, size, sequence, acceptance)
  • incentives (what is rewarded, punished, tolerated)
  • information flow (what is visible, when, to whom)
  • domains (bounded contexts, responsibility maps)

A system that tries to control outcomes directly will drift into moralizing (“we need to care more”) or reporting (“let’s measure it harder”).

What “Object of Control” Means

An object of control is:

  • touched repeatedly by the system
  • changeable by the actors running the system
  • causally close enough to influence the failure
  • representable as an artifact or rule

A good object of control is something you can point to and edit.

If you can’t edit it, you can’t control it.

The Common Objects of Control

This book uses a deliberately short list to force precision. Most systems manipulate one or two of these.

Goals

What gets controlled:

  • goal definition
  • goal hierarchy
  • goal review cadence
  • goal acceptance criteria

Typical systems:

  • OKRs, KPI trees, strategy deployment

Misuse risk:

  • goals become wishes or slogans
  • goals multiply without tradeoffs

Work items

What gets controlled:

  • what counts as “work”
  • work item size and slicing rules
  • entry/exit criteria (DoR/DoD)
  • prioritization and sequencing rules

Typical systems:

  • agile variants, kanban policies, intake processes

Misuse risk:

  • backlog as junk drawer
  • prioritization without capacity reality

Interfaces

What gets controlled:

  • team boundaries and integration points
  • API contracts, SLAs, change protocols
  • cross-team dependency handling

Typical systems:

  • platform operating models, integration governance, contract testing norms

Misuse risk:

  • “coordination” replaces interface clarity
  • escalations become the interface

Domains

What gets controlled:

  • responsibility boundaries
  • language boundaries (what terms mean where)
  • ownership of concepts and data

Typical systems:

  • domain-driven design, team topologies, operating model definitions

Misuse risk:

  • domains become political territories
  • boundaries drawn without enforcement

Constraints

What gets controlled:

  • WIP limits, timeboxes, approval thresholds
  • quality gates
  • queueing policies
  • default rules when decisions are not made

Typical systems:

  • flow-based delivery, incident response, change management

Misuse risk:

  • constraints applied without authority
  • constraints become bureaucracy rather than throughput protection

Incentives

What gets controlled:

  • what is rewarded, punished, ignored
  • performance metrics and evaluation criteria
  • promotion narratives

Typical systems:

  • performance systems, goal systems, compensation-linked measurement

Misuse risk:

  • metric gaming
  • local optimization

Information flow

What gets controlled:

  • what is visible
  • when it becomes visible
  • to whom
  • in what form (artifact)

Typical systems:

  • dashboards, reporting cadences, decision logs, review rituals

Misuse risk:

  • legibility replaces truth
  • reporting replaces decisions

Selecting the Right Object of Control

Use this sequence to avoid choosing an object that is merely convenient.

Step 1: Start from the failure

Ask:

  • “What must change repeatedly for this failure to reduce?”

If the answer is “people need to communicate better,” you don’t have an object. Keep going.

Step 2: Identify the decision you need to optimize

Common mappings:

  • Priority decisions often control: work items, investment, goals
  • Ownership decisions often control: interfaces, domains
  • Sequencing decisions often control: work items, constraints
  • Diagnosis decisions often control: information flow, work item definitions
  • Repair decisions often control: constraints, interfaces, ownership boundaries

Step 3: Run the “edit test”

For each candidate object:

  • Can the team change it within their authority?
  • Is it updated often enough to matter?
  • Can it be represented as an artifact or rule?
  • Does changing it plausibly affect the failure within a reasonable time horizon?

If the answer is “not really,” it’s not your object of control.

The Most Common Trap: Controlling Meetings

When people don’t know what to control, they control the calendar.

Symptoms:

  • new cadences
  • more “syncs”
  • longer planning sessions
  • more reviews

This creates the illusion of governance but usually does not change:

  • ownership boundaries
  • sequencing rules
  • constraints
  • interface contracts

Meetings can be part of a system, but they are not an object of control unless the meeting produces enforceable artifacts and constraints.

Object of Control and Artifact Must Match

You can’t control an object you can’t represent.

Examples:

  • If you control interfaces, you need interface artifacts (contracts, SLAs, ownership maps).
  • If you control constraints, you need constraint artifacts (WIP policies, gates, default rules).
  • If you control domains, you need domain artifacts (maps, glossaries, context boundaries).

If your artifact is a slide deck but your object of control is interfaces, you will get theater.

Misuse Model: How “Objects of Control” Gets Misapplied

Misuse 1: Choosing too many objects

If you control everything, you control nothing.

Failure mode:

  • the system becomes complex
  • enforcement collapses
  • adoption becomes political

Correction:

  • choose 1–2 objects of control max for a system. Treat others as secondary effects.

Misuse 2: Choosing an object outside authority

You design a system that requires changing incentives or org boundaries when you can’t.

Correction:

  • redesign the system to use objects you can actually control, or define an adoption path that includes the authority holder.

Misuse 3: Picking an object that is too far from causality

Example: controlling dashboards to improve reliability.

Dashboards can improve awareness but rarely change the mechanisms that produce incidents (interfaces, constraints, quality gates).

Correction:

  • prefer objects closer to causal mechanisms even if they are less comfortable.

The Non-Negotiable Rule Introduced Here

You may not design or adopt a system until you can state:

  • The object of control is ______
  • The actors who can change it are ______
  • The artifact that represents it is ______
  • The constraint that enforces it is ______

If you can’t complete this, you are trying to control an outcome with vibes.

Exit Condition for This Chapter

Write:

  1. Your system’s 1–2 objects of control
  2. Who has authority to change each object
  3. The artifact that will make each object inspectable
  4. One constraint that prevents drift or avoidance

If you can do that, you are now operating like a system designer, not a framework consumer.