Chapter 6 — Seeing the System You Are In

Why optimizing parts reliably harms wholes.

Most organizational failure does not begin with bad intent.

It begins with narrow success.

A team improves its metric. A function hits its target. A workflow becomes faster, cheaper, cleaner, more predictable. From inside the local boundary, progress is real and often hard-won.

Then, somewhere else, friction increases.

Handovers degrade. Risk shifts downstream. Decision latency moves to another unit. Customer burden rises while internal throughput improves. No single actor sees the whole transfer in real time, so each local improvement remains defensible.

This is how systems become more efficient and less adaptive at the same time.¹

Learning-oriented systems refuse to confuse local optimization with global learning.

Local Clarity, System Blindness

Every organization views itself through selective surfaces:

• dashboards,
• KPIs,
• reporting lines,
• process maps,
• and role-specific narratives of value.

These surfaces are necessary. Without them, coordination collapses.

But they are also partial.

They make some flows visible and hide others. They make some harms countable and others anecdotal. They elevate what can be attributed to a team and obscure what emerges between teams.

As a result, systems can become highly legible locally while becoming opaque systemically.

Actors then optimize what they can see and externalize what they cannot. 2

The Boundary Problem

System failure is often a boundary-design failure.

If boundaries are drawn too narrowly, teams are rewarded for outcomes that are only positive inside their slice. If boundaries are drawn too broadly, ownership diffuses and no one can act decisively.

Seriousness requires boundaries that are operationally useful and morally accountable.

That means asking of every performance frame:

• What consequences are included?
• What consequences are exported?
• Who absorbs exported cost?
• Who has authority to surface that cost?

Without this discipline, organizations drift into a stable pattern: internal gains financed by externalized degradation.

The system appears to improve because the bill is paid elsewhere.²

Delays, Coupling, and False Causality

Even when actors care about whole-system outcomes, complexity obscures causality.

In tightly coupled settings, small local changes propagate quickly through dependencies that were never explicitly modeled. In loosely coupled settings, effects appear so late that attribution becomes contested or politically impossible.

Both conditions distort learning:

• tight coupling creates rapid surprise;
• delayed coupling creates retrospective ambiguity.

Under either condition, teams revert to proximal signals and short horizons because those are the only signals that feel actionable.

This is understandable. It is also how system-level harm becomes normal: it arrives either too quickly to interpret or too late to own.³

Why Metrics Drift from Meaning

Metrics do not fail because they are numeric. They fail when they become proxies detached from purpose.

A target is introduced to represent a value. Over time, process pressure shifts behavior toward maximizing the target itself. The proxy hardens; the value thins.

This is not typically fraud. It is adaptation under incentive.

When careers, budgets, and legitimacy depend on measured outputs, actors move toward what is rewarded even if doing so erodes the underlying objective. The system remains accountable to the metric while becoming less accountable to reality.

They counter this by repeatedly re-linking measures to the human and operational outcomes they were meant to track.

If that relinking stops, measurement discipline can become a sophisticated form of self-deception.⁴

The Hidden Queue of Consequence

System harm often accumulates in invisible queues.

Backlogs of unresolved risk. Deferred maintenance. Unanswered incident patterns. Unintegrated learning from near misses. Emotional exhaustion that does not appear in delivery metrics until capacity suddenly drops.

These queues are easy to miss because they do not always degrade short-term outputs. They can coexist with excellent quarter-over-quarter performance.

By the time queue pressure surfaces, correction is expensive.

What looked like resilience was often debt.

Serious systems learn to treat queue growth as a first-order signal, not a side note. If unresolved consequence is compounding, performance may be improving in appearance while adaptability is deteriorating in fact.⁵

Cross-Scale Learning vs. Local Heroics

Many organizations survive fragility through local heroics.

Individuals and teams bridge boundary gaps manually. They translate across silos, absorb extra load, and compensate for design defects through discretionary effort. This keeps service running and prevents visible failure.

It also masks structural problems.

When heroics are normalized, leadership can misread extraordinary labor as evidence of robust design. Learning then remains local and embodied, not institutional and transferable.

Serious systems convert heroics into architecture.

They ask: What recurring workaround is signaling a design deficit? What cross-boundary friction is being privately absorbed? What “high performance” depends on unsustainable discretionary effort?

If those questions are not institutionalized, organizations will keep rewarding the very adaptations that prevent system-level correction.⁶

Seeing Through a Multi-Outcome Lens

To see the system you are in, single-outcome evaluation is insufficient.

At minimum, decisions should be read across four simultaneous outcomes:

• local performance (did the team hit its goal?),
• system interaction quality (did dependencies strengthen or weaken?),
• downstream cost transfer (where did burden move?),
• reversibility (did this increase or decrease future correction capacity?).

Most governance processes privilege only the first.

Serious systems institutionalize all four.

This does not eliminate tradeoffs. It makes them visible enough to be owned rather than displaced.

Seeing the whole does not mean controlling the whole. It means refusing to call an action “successful” when its gains depend on unacknowledged degradation elsewhere.⁷

The Discipline of System Sight

System sight is not intuition. It is practice.

It requires:

• review cadences that include cross-boundary effects,
• decision logs that track assumptions and exported costs,
• escalation channels for dependency stress signals,
• and incentives that reward net system improvement rather than isolated target maximization.

Most importantly, it requires leaders to tolerate ambiguity long enough to avoid premature narrative closure.

Premature closure is how local stories become system myths.

Serious systems keep interpretation open until downstream reality has had time to answer.

Footnotes

1. Meadows, Donella H. Thinking in Systems: A Primer. White River Junction, VT: Chelsea Green Publishing, 2008. ↩

2. Weick, Karl E., and Kathleen M. Sutcliffe. Managing the Unexpected: Sustained Performance in a Complex World. 3rd ed. Hoboken, NJ: Jossey-Bass, 2015. ↩

3. Perrow, Charles. Normal Accidents: Living with High-Risk Technologies. New York: Basic Books, 1984. ↩

4. Muller, Jerry Z. The Tyranny of Metrics. Princeton, NJ: Princeton University Press, 2018. ↩

5. Dekker, Sidney. Drift into Failure: From Hunting Broken Components to Understanding Complex Systems. Farnham, UK: Ashgate, 2011. ↩

6. Senge, Peter M. The Fifth Discipline: The Art and Practice of the Learning Organization. New York: Doubleday/Currency, 1990. ↩

7. Forsgren, Nicole, Jez Humble, and Gene Kim. Accelerate: The Science of Lean Software and DevOps. Portland, OR: IT Revolution Press, 2018. ↩