Appendix: capability and readiness dictionary

The canonical input is a complete declared slate, not an import of an enterprise inventory. It contains an investment envelope, an aggregate effort envelope and capability records. Each record supplies identity, name, cost, annual benefit, confidence, readiness, effort and prerequisites. The author should retain the input exactly as executed so a reader can distinguish the original proposition from a later revision. The schema and engine validate structure and declared graph conditions; they do not establish the operating truth of the slate.

A capability identifier is a machine-readable reference used by dependencies and deterministic tie ordering. Its display name is a human description. Both should be stable enough to interpret comparisons. Renaming an identifier can affect tie ordering even when the other numerical fields remain unchanged. A reader should therefore treat identifiers as part of the frozen input rather than harmless presentation detail. The study preserves them in the input, binding ledger and output.

Cost is a declared investment amount in USD thousands. Annual benefit is a declared amount in USD thousands per year. Effort is a declared aggregate amount in person-months. The author should state the cost boundary and benefit mechanism for every record. The app does not infer a currency conversion, recurring-cost schedule or actual staffing plan. Mixing scopes across capabilities can produce an exact but poorly interpretable ranking. A coherent input requires reasoning before calculation.

Confidence and readiness are percent judgments whose fractions multiply the annual benefit. They are not weights in a weighted-average maturity score. Their product can reduce a declared benefit, but it cannot enforce a mandatory operating condition. The dictionary should explain the interpretation chosen for the particular slate and identify the evidence that would change it. In this study the values are authored assumptions used to expose sensitivity, not measured adoption or calibrated probabilities.

Dependencies are references to required capabilities in the same slate. A hard dependency means the prerequisite is included in a target's closure when not already selected. The author should retain the operating reason for each arrow and identify any alternative scope that could change it. The app does not infer whether a manual route, a different architecture or a narrower service can substitute. Such alternatives require another explicitly authored input and a corresponding interpretation.

The output contains the selected set, totals, remaining resources, capability readings, initial ranking and selection steps. The selected set is ordered by the declared graph, while the steps show which target bundles were chosen. The initial ranking describes candidate closures before selection. These views answer different questions. A reviewer should follow the step record to explain the sequence and the final set to identify included capabilities, rather than assume the display order is a universal investment priority.

The capability readings retain adjusted assumed benefit and a selection explanation. A capability may be selected as a prerequisite even if its own direct benefit is absent. Another may remain unselected because its bundle exceeds a remaining envelope or because the heuristic chooses another bundle. These readings are conditional on the declared input and method. They should not be interpreted as judgements that an unselected capability is strategically unnecessary or that a selected one is operationally accepted.

A step records its target, marginal bundle, cost, effort, adjusted assumed benefit, ranking ratio and remaining envelopes. The marginal bundle excludes prerequisites already selected. This is why it should not be compared casually with the initial full closure as if the records contradicted each other. The difference reflects the model's state after prior selections. The full frozen output permits the reader to trace that state rather than rely only on a summary total.

A null ratio has a specific meaning when cost is zero. It should not be formatted as zero, an infinite number or an empirical reading. The app explicitly prioritises positive-benefit zero-cost bundles subject to effort feasibility. A zero-benefit prerequisite can still enter through a positive target's closure. An empty selected set can be valid under an unavailable envelope. These boundary meanings prevent a presentation layer from turning legitimate model states into misleading conclusions.

Warnings and limitations supplied by the canonical app belong with the study. The manuscript can explain them through cases, but should not remove them to make a recommendation appear stronger. A frozen result that omits its method identity or limits is less useful to a reader even when its totals are correct. The reproduction bundle preserves the complete output so its meaning can be assessed independently.

Resource admission uses the canonical decimal spelling of each validated input number. The engine sums prerequisite costs and effort exactly and compares those sums with the remaining envelopes; it does not add an absolute allowance to a zero or positive budget. A decimal equality can therefore be admitted without generating a negative remainder, while a small positive commitment remains a commitment beside a much larger envelope. The JSON resource account retains exact decimal capacity, used and remaining strings for both constraints. Displayed numeric totals, adjusted assumed benefits and density rankings still use Number approximations. JSON parsing may already have lost digits before the resource account begins, so the record does not preserve arbitrary precision from every submitted textual numeral. Exact admission strengthens the feasibility account within this demonstrator. It does not validate the resource estimates, combine calendar resources or prove that the greedy selection is globally best.

The app rejects unsupported fields rather than silently treating them as modelled dimensions. An author cannot add an optimisation objective, calendar, benefit synergy or legal condition and assume the engine uses it. The valid schema makes the boundary visible. Where those dimensions matter, they need a suitable separate method or an explicitly changed canonical model under its own review. The study does not alter the app to make its cases appear more complete.

Identifiers and dependencies must form a valid acyclic graph. Unknown prerequisites, self-dependencies and duplicate identities do not produce a meaningful selection. Numeric inputs must be finite and within their schema-defined ranges and precision boundary. Negative cost or effort is not a convenient way to represent a grant, credit or released resource. Such meanings require an appropriate declared method rather than a value that violates the input contract.

Structural validity is necessary but weaker than operating validity. A graph can be acyclic while encoding an unnecessary architectural preference as a hard prerequisite. A readiness judgment can be in range while unsupported by evidence. A benefit can be finite while duplicated across capabilities. The author should therefore maintain a reasoning record alongside validation. The engine checks what it is designed to check; the sponsor and reviewers examine whether the declared input represents an intelligible proposition.

The author-side invalid-fixture receipt executes deliberately malformed cases through the actual engine and records the rejection messages. It is separate from the valid study outputs. An invalid input should never be represented as a successful zero-result programme. The receipt confirms the tested rejection behaviour within its scope. It does not prove that every possible malformed input has been explored or that an accepted input is suitable for a real investment.

The distinction should persist through import and reopening. A scenario saved in the platform contains the input and supported model version, then is recomputed on reopen. It does not treat a cached user-provided output as authoritative. The manuscript's frozen bundles similarly retain exact input/output files and reciprocal scalar bindings. Reproduction is an explicit operation, not a label attached to an unverified spreadsheet or summary.

The dictionary should make it possible to ask a precise question about a record. Why does the operations capability require data and ownership? What cost scope is included? Which operating mechanism produces its declared annual benefit? What evidence would change readiness? Does the effort envelope include the necessary specialised work? These questions should have a stated answer or an explicit unresolved status. A complete numerical row without that reasoning can conceal an incomplete proposition.

The reference dictionary names data as shared operating information and ownership as the service decision route. These are illustrative capability scopes, not proof that a real enterprise can purchase authority as a software component. The declared costs include authored implementation work for comparison. Continuing operation and actual delegation remain outside the demonstrated numeric calculation. The chapter nineteen branch uses this distinction to hold production while allowing bounded discovery.

A revised input should preserve a change record. If an author changes a dependency, cost scope or judgment, the comparison should state what changed and what remained fixed. The frozen cases in the next chapter provide that record. A reader can then interpret a changed result as conditional sensitivity rather than mistakenly infer an observed change in the enterprise.

The analytical contract is deliberately narrow. It supplies exact decimal resource admission and sums, transparent assumptions and a deterministic candidate. Adjusted benefit and density retain finite numerical precision; they are outside the exact resource arithmetic. The operating dictionary supplies the meanings needed to challenge those assumptions. The commitment record supplies the authority and evidence required to act. Together they form the proposed synthesis, while remaining distinct enough that a gap in one cannot be hidden by completeness in another.