Fix problems at the deepest causal link you can reach — surface fixes only address symptoms

Why This Is a Rule

Causal chains are sequences: A causes B causes C causes D (the visible problem). Fixing D (the symptom) leaves A-B-C intact, and the chain fires again producing a new symptom. Fixing C breaks the last link but leaves A-B to potentially cause other problems. Fixing A prevents the entire chain from ever firing — every downstream link becomes inert.

The leverage point is the deepest link where you have both ability (the technical or cognitive capacity to change it) and authority (the organizational or personal permission to change it). The deepest possible link is often the root cause, but you may not have ability or authority to change it. The deepest actionable link is where the maximum available leverage lives.

Meadows's leverage-point framework establishes that deeper interventions are exponentially more effective: changing a system's rules (deep) is more powerful than changing its parameters (surface). Applied to causal chains, this means tracing the chain as deep as possible before intervening — not just fixing what's broken at the surface.

When This Fires

• During root cause analysis when deciding where to intervene
• After mapping a causal chain and choosing the intervention point
• When a surface fix has been applied but the problem keeps recurring
• Any problem-solving context where the symptom is downstream of multiple causes

Common Failure Mode

Fixing the visible symptom because it's the most accessible: "The deploy broke, so we'll add a pre-deploy check." But the deploy broke because the test suite doesn't cover the failing code path (deeper), which happens because the team doesn't write integration tests (deeper still), which happens because there's no testing culture or allocated time (deepest actionable). A pre-deploy check catches one class of failure; building testing culture prevents the entire category.

The Protocol

When a problem surfaces: (1) Trace the causal chain backward: what caused this? What caused that? Keep tracing until you reach the deepest identifiable link. (2) For each link, assess: do I have both ability (can I change this?) and authority (am I permitted to change this?)? (3) Intervene at the deepest link where both criteria are met. (4) If you can only reach a surface link → implement the surface fix, but document the deeper links as future intervention targets. Surface fixes buy time; deep fixes solve the problem.