How are the calculations of a decision engine verified?

What does it mean for a calculation to be verified?

A verified calculation is one that a second party has recomputed from the same inputs and found equal to the first result. The word describes a process, not a mood. It needs written inputs, a fixed procedure, and a recorded outcome of the comparison. Without those three elements, the word is only a claim, and a careful reader asks to see the record behind it.

Verification is distinct from validation. Validation asks whether the equation describes the world well, which back-tests and sensitivity checks address. Verification asks whether the software computes the equation correctly. The two together support a score, and a reader who separates them asks sharper questions of any vendor.

Validation has a checklist of its own. A reviewer defines the purpose of the engine, checks the inputs and the structure, runs back-tests on past cases, applies stress tests, and monitors the model over time. The guide to model validation for a decision engine walks through each check, so correct computation and good fit each receive a clear procedure.

Why does a deterministic core make verification possible?

A deterministic system returns the same output for the same inputs, every time. That property is the foundation of verification, because a second run can be compared line by line with the first. The method page describes the Codex Universalis as a deterministic core, with no AI, no learning, and no hidden probabilities (EQOS Asia method page, as of October 2026).

A probabilistic system returns a distribution, so two runs can differ within a stated range. That design suits some questions, and it makes exact recomputation impossible by nature. A language model adds another difficulty, since its answer to the same question can change from one run to the next. Exact recomputation therefore belongs to deterministic engines, and a buyer who needs an auditable number can choose accordingly.

Which part of a decision engine calculation is computed or AI-assisted?

EQOS Asia draws a line on its own method page. Scores, indices, and numerical verdicts, such as Λ and φ(S), come exclusively from the deterministic Python engine, never from a generative model. Research, drafting, and formatting of studies are assisted by generative AI tools, always under human review before release (EQOS Asia method page, as of October 2026).

The line matters for verification. A reader can recompute everything on the computation side of the line, and a human reviewer is accountable for everything on the drafting side. The figure below sets the two sides next to each other, with the control that applies to each.

Which checks verify a calculation step by step?

Five checks make a calculation verifiable. Freeze the inputs with their sources and dates. Run the engine and store the output. Recompute the same case independently. Compare the two results number by number. Log the engine version so the run can be repeated later. These checks come from general good practice, and each one leaves a record.

Hand-worked cases add a useful layer. A reviewer computes a simple case by hand, feeds the same case to the engine, and compares the results. A match on simple cases builds confidence before complex ones are tested, and any gap points straight to the step that needs repair. The figure below lists the sequence of the five checks.

Who can recompute an EQOS Asia decision verdict, and under what conditions?

The method page states that every verdict can be recomputed identically by any data scientist with access to the Codex, under NDA. The about page repeats the point and adds that investment committees in Singapore, Hong Kong, and Tokyo get a result they can audit (EQOS Asia method and about pages, as of October 2026).

The expanded forms and weighting coefficients of the Codex are proprietary, so access runs through a documentation request under NDA. The Codex is filed under INPI e-Soleau DSO2026020500. This arrangement protects the intellectual property and still lets a qualified reviewer repeat the computation (EQOS Asia method and about pages, as of October 2026).

How does human review fit around the calculation?

The method page lists a systematic ethical safeguard as the validation step of the Audit computation. Human review also surrounds the written material: research, drafting, and source-checking are AI-assisted with systematic human review before delivery, lighter on entry-level tiers and full on Study, Diagnostic, Mapping, and Mission tiers (EQOS Asia method page, as of October 2026).

This tiered review gives buyers a clear expectation. A Reading at US$358 receives the deterministic score with lighter review, and a Mapping at US$22,976 receives full review (EQOS Asia pricing page, as of October 2026). The same engine computes the score at every tier, and only the depth of the work changes (EQOS Asia about page, as of October 2026).

How can you ask for a verified calculation on your case?

Ask three things of any provider: which part of the output is computed, who can recompute it, and under which conditions. Then apply the same questions to EQOS Asia. A 30-minute pre-diagnostic by video, scheduled in Singapore time (SGT), is the place to ask them (EQOS Asia contact page, as of October 2026).

The documentation request sits on the method page, and the pricing page shows the tiers that carry each depth of review. Taking the questions to the call puts the verification process on the table before any purchase decision. The answers then become part of your own file on the decision (EQOS Asia method and pricing pages, as of October 2026).

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile