Information can fail through an implementation gap

The information-value calculation assumes that the institution can select and execute the action recommended in each branch. This assumption is often more demanding than the arithmetic. A favourable posterior may require a response that exceeds delegated authority, depends on another party or consumes a resource already committed elsewhere. A negative posterior may require withdrawal from an obligation that cannot be abandoned. The institution should review these conditions before purchasing a study whose value relies on them. Evidence becomes useful through a response path, not through an attractive branch label alone.

The restricted-action fixture provides a clean numerical test of this implementation gap. It removes Commit while retaining the prior, likelihoods and the remaining action payoffs. The instrument still changes beliefs and can still alter the action between Stage and Hold. However, its gross improvement falls because the high-upside response is unavailable. The same price now exceeds the decision benefit. This does not estimate the value of obtaining approval for Commit. It shows that an information commission justified with an unexecutable menu can be economically misleading.

The review should distinguish authority from capability. A committee may have the power to approve a redesign but lack the technical team to implement it. An operator may have the capability to change a process but lack permission under a service agreement. A sponsor may possess both and still depend on an external interface. These are different prerequisites with different owners and evidence. A generic statement that management will act does not establish any of them. The action register should identify the concrete transition from a reviewed signal to an operating or commercial change.

Implementation can also require another information decision. A positive technical result may justify a detailed design study before a commitment, rather than immediate construction. A negative result may justify a diagnosis of the failure mechanism rather than abandonment. The simple app cannot infer those sequential responses from an action name. If the committee's actual next choice is another commission, that action should have its own perimeter and analysis. Describing a broad future strategy as a single action is acceptable only when the supplied aggregate consequences and limitations make the compression explicit.

An institution should avoid building a decision rule around a response it cannot credibly commit to consider. If the sponsor has already decided that an adverse signal will be ignored, the purported benefit of avoiding failure exposure is absent from the actual policy. A model can still show what a rational response would do under its objective, but it should not claim that this is the organisation's implemented conditional policy. A credible charter needs a predeclared review process and an authority capable of changing the commitment when the evidence warrants it.

The evidence provider also needs a clear boundary. A research team can deliver a methodologically sound result without controlling the client's response. The study contract should specify the evidence object, method, scope and limitations, rather than promise the success of a later operating decision. The client should own the action architecture and acceptance process. This division prevents a provider from being credited with unimplemented benefits or blamed for a response it was never authorised to control. It also makes the commissioning rationale more honest about where value must be created.

The handoff should preserve the meaning of the signal. Operating staff need to know which state the posterior describes, the conditions under which the instrument was evaluated, and the actions that were reviewed. A summary stating that the test passed can conceal conditional uncertainty and unrepresented outcomes. The branch record should travel with the decision package, including the adverse response and the circumstances that require escalation. The app's export supports a reproducible numerical object; the institution must still organise the human handoff that makes it usable.

An action may remain feasible but lose its original payoff when implementation is delayed or incomplete. The fixture assumes the supplied response consequences occur under the selected action. In practice, a partial redesign, missed interface or weak operating adoption can change those consequences. The simple engine does not apply an implementation success probability or calculate adoption cost. A team can supply revised aggregate payoffs only when it explains their basis. If implementation uncertainty is material and distinct from the opportunity state, a broader state or sequential model may be needed.

The response path should contain acceptance criteria. For a full commitment, this may mean confirming a defined technical condition and remaining obligations. For staging, it may mean identifying what the initial phase authorises and what remains contingent. For holding, it may mean checking service duties and opportunity expiry. These criteria should be relevant to the action, proportionate to the consequence and owned by someone who can verify them. They are not universal checklist items; the institution should define them from its actual mandate and evidence rather than copy a hypothetical worksheet without adaptation.

Governance can alter information value without altering signal accuracy. A clearer delegated scope, a feasible fallback or an agreed response review can make an existing study more usable. This does not mean governance automatically creates monetary value. The action consequences and costs need their own support. It means that the evidence commission and implementation architecture should be designed together. Buying a more accurate signal while leaving response authority unresolved can spend resources on precision that the institution still cannot exploit.

The implementation record should also identify who may reject a response despite its numerical attractiveness. A protected interest, unresolved obligation or material safety condition can make an action inadmissible. The model's expected-value optimum is not a defence against such a boundary. The appropriate authority should review the constraint and the supporting evidence. This preserves the distinction between a financial comparison and an institutional permission. The paper's hypothetical roles illustrate that division without inventing a legal conclusion or an actual client's approval process.

After the signal arrives, the committee should compare the proposed response with the predeclared rule. A justified change can occur because new evidence, a changed opportunity or an uncovered constraint alters the decision. The record should describe that reason. An unexplained deviation can indicate that the original study value was overstated or that the action architecture was never accepted. Reviewing deviations is useful for improving future commissions, provided the institution does not infer empirical performance rates from a small selected set of cases without an appropriate universe and method.

Test the response path with a hypothetical favourable result delivered while Commit is not executable. If permission is absent, obtain the appropriate authority decision; if a compatible resource is absent, resolve capacity; if the result is outside the accepted population, return to instrument scope. These failures can look identical in a dashboard that simply reports no action taken, yet need different repairs. Stage or Hold may remain usable fallbacks under the supplied menu. Record the selected fallback and the reason the intended branch could not advance. A later favourable outcome does not retroactively make Commit available at the decision date, and a later adverse outcome does not identify which interface failed.

An implementation exercise should stop at the first unmet prerequisite. In a hypothetical paper exercise, branch recipients receive the accepted signal, identify the response, obtain the required authority and reserve its operating resources. If reservation fails, continuing to sign a nominal response instruction demonstrates paperwork rather than execution. Retain the failed interface and test a permitted fallback instead. Passing the exercise establishes the interfaces actually exercised, while field performance and payoffs remain unvalidated. This makes the response worksheet a discriminating instrument and shows why the restricted-action comparison is an implementation sensitivity, rather than a calculation of the value of persuading an authority to approve an action.