Skip to content

Substitution Matrix

Decision artifact — instantiates Inventory-Bounded Resource Recomposition

A table that scores candidate stand-ins against the attributes a role requires and records which swaps are acceptable under which conditions — and which resources must never be substituted at all.

When the ideal part is missing, the question is not "what else is lying around?" but "what may safely stand in for it, and under what conditions?" A Substitution Matrix is the artifact that answers this: candidate resources down one axis, the attributes the role actually demands across the other, each cell scored for fit, and a verdict column recording whether the swap is acceptable, acceptable-with-conditions, or forbidden. Its defining move is making substitution a judged, recorded rule rather than an on-the-spot guess — and, crucially, it carries the negative space too: the cells marked non-substitutable, where a resource's role is so safety- or function-critical that no stand-in is permitted regardless of convenience. It says what may be swapped for what; it does not discover the candidates or build the bridge when none fits cleanly.

Example

A craft brewery loses its supplier mid-batch and the specified bittering hop, a gasket, and a pressure-relief valve are all unavailable. The brewer builds a substitution matrix. For the hop, candidates from the store are scored on alpha-acid content, aroma profile, and quantity on hand: a substitute at a corrected weight is marked acceptable — adjust addition by ratio. For the gasket, a food-grade silicone ring of the right diameter is acceptable under condition: not on the hot-side line. The relief valve gets a hard verdict: non-substitutable — safety-rated component, no swap; halt and source.

The matrix turns three panic decisions into three recorded rules. Two swaps proceed with stated conditions written down; the third is explicitly forbidden, so no one "improves" the situation by fitting a valve that isn't pressure-rated. Weeks later, when a different brewer hits the same shortage, the matrix already holds the answer.

How it works

What makes the matrix this mechanism, rather than a shopping list, is that it evaluates fit against requirement and preserves the verdict:

  • Attributes are the requirement, not the label. Columns are the properties the role genuinely needs — rating, tolerance, purity, throughput — so a candidate is judged on what the job demands, not on sharing a name with the original.
  • Conditional verdicts, not yes/no. Most acceptable swaps come with a stated condition or adjustment ("re-torque to X", "de-rate by 20%"), and the matrix records the condition with the verdict so the swap is reproducible.
  • The forbidden column is explicit. Roles where substitution is barred — safety-critical, certified, or invariant-bearing — are marked non-substitutable up front, so scarcity pressure can't silently erode them.
  • It is a durable record. Each judged swap becomes reusable precedent the next time the same gap appears.

Tuning parameters

  • Attribute set — how many requirement dimensions each candidate is scored on. More attributes catch subtle mismatches but slow every judgement and can drown the decision in detail.
  • Acceptance threshold — how close a stand-in must score to count as acceptable. A strict bar avoids risky swaps but rejects workable ones and can stall a build; a loose bar keeps things moving but admits marginal fits.
  • Conditionality depth — whether verdicts are plain yes/no or carry adjustments and caveats. Richer conditions make swaps safe and repeatable but demand more expertise to author and to follow.
  • Non-substitutable strictness — how readily a role is placed in the forbidden column. Broad protection is safe but rigid; narrow protection is flexible but risks eroding an invariant under pressure.

When it helps, and when it misleads

Its strength is that it makes substitution disciplined and portable: a swap is judged once against real requirements, its conditions are captured, and the answer is reusable — while the non-substitutable column holds a hard line exactly where make-do pressure is most likely to cut a dangerous corner. The engineering standard it encodes is form, fit, and function: a valid substitute must match the original's physical form, its interface fit, and the function it performs, on all the dimensions that matter.[1]

Its failure mode is a matrix scored on the wrong attributes — capturing the easy-to-measure properties while missing the one that actually governs the role, so a swap reads "acceptable" and fails in service. It also ages badly: conditions drift, and a verdict trusted long after its assumptions changed becomes a rubber stamp. The classic misuse is running it backwards — starting from the substitute you want to use and picking attributes that make it pass. The discipline is to fix the required attributes from the role before scoring candidates, to treat the non-substitutable column as owned by safety rather than by convenience, and to send any load-bearing swap onward to a real test rather than trusting the score.

How it implements the components

The matrix realises the substitution-policy slice of the archetype — the rules governing what may replace what — and neither discovers candidates nor bridges gaps:

  • substitution_rule — each judged, conditioned verdict is a substitution rule: this may stand in for that, under these adjustments, on these attributes.
  • non_substitutable_constraint — the forbidden column records the roles where no stand-in is admissible, holding the line on safety- and function-critical resources.

It rules swaps in or out but does not discover the affordances it compares — that is the Affordance Inventory Walkthrough — it does not design a bridge when no clean substitute exists, which is Technical Bypass or Adapter Design, and it does not prove a chosen swap holds under load, which the Integration Test Plan does.

References

[1] Form, fit, and function is the standard interchangeability test in engineering and logistics: a substitute part is admissible only if it matches the original's form (physical shape and material), fit (how it interfaces and mounts), and function (what it does) on every dimension the application requires. The matrix operationalises the test one candidate at a time. ↩