Decision engine checklist: ten questions to ask before you buy

Why use a checklist before you buy a decision engine?

A checklist replaces impressions with answers. Demos are persuasive, and a fluent presentation can hide the one detail that matters: whether the result can be checked. Ten written questions put every vendor on the same footing and give your team a record to review later, when the purchase is discussed again.

The questions fall into five areas: reproducibility, inputs, output, protection, and price with first step. Each answer should be short, specific, and verifiable. A vendor who answers with documents, dates, and figures gives you something to compare, and a vague answer is information too, so it belongs in your notes.

A scoring sheet extends the checklist when several vendors stay on your short list. Each of the 10 questions feeds a weighted row, each score carries a dated source, and the totals show where a tool leads. The guide to decision software selection criteria gives a layout with gates, weights that sum to 100 and a worked comparison of two tools.

Which questions test whether the results are reproducible?

Question 1: does the same input always return the same score? Run this test yourself with a simple case, twice, on different days. Question 2: can a third party recompute a verdict? A reviewer outside the vendor should be able to follow each result from inputs to ranking. These two questions separate an explicit model from a black box. Keep the test case small enough to check by hand, so you can confirm each step.

In the EQOS Asia method, the Codex Universalis core is deterministic, and every verdict can be recomputed identically by any data scientist with access to the Codex under NDA (EQOS Asia method page, as of October 2026). Note the same two answers for each engine on your short list, and write the access conditions beside them.

Which questions check the input data a decision engine relies on?

Question 3: which inputs does the engine need, and from which sources? Expect a precise list covering measured facts, expert ratings, and context such as regulation and currency. Question 4: how is each input dated and sourced? Dated inputs let you refresh the score when one figure changes, and they keep the audit trail short.

The EQOS Asia method collects inputs through structured interviews of 2 hours at the Standard level, with documents collected and vectorized alongside them (EQOS Asia method page, as of October 2026). Any format is accepted, from PDF and video to audio and raw data. Compare that scope with what each vendor on your list can ingest.

Which questions show what the engine returns?

Question 5: what does the output look like? A useful engine returns a score, a threshold, and a recommended action, so the reader can see where a case stands and what to do next. Question 6: which risks does it name? A ranked list of priority risks tells your team where to act first. Compare the answers side by side, on the same sample case.

The Audit pillar of EQOS Asia scores a situation on 9 structural dimensions, returns a φ(S) value between 0 and 1, and lists the 3 priority risks. A score below 0.618 falls in the STOP zone (EQOS Asia method page, as of October 2026). Ask every vendor for a sample output with the same level of detail.

Which questions confirm how your data is protected?

Question 7: when is the NDA signed? The answer should be before any document is shared. Question 8: where is the data hosted, and how is it encrypted? Ask for the country, the encryption in transit and at rest, and the retention periods, and keep the replies with your notes. Written answers matter here, because a spoken assurance is hard to review afterwards.

At EQOS Asia, a systematic NDA is signed before any exchange, data is hosted in Singapore under PDPA-aligned handling, and questionnaire answers are encrypted in your browser before sending (EQOS Asia pricing page, as of October 2026). The privacy policy also lists TLS 1.3 in transit and AES-256 at rest (EQOS Asia privacy page, as of October 2026).

Which question clarifies the price and the delivery time?

Question 9: what is the price at each level, and what does each level include? Ask for a grid with delivery times and for the list of deliverables. A clear grid lets you match the tier to the size of the decision and plan the budget before the first call. Ask also how payment is organised for each level.

The EQOS Asia grid in Singapore starts with a Reading at US$358 delivered in 24-48 hours and a Decision at US$910 delivered in 48-72 hours. The three pillars of audit, modelling, and prediction activate from the Study at US$1,369 (EQOS Asia pricing page, as of October 2026). The pricing page lists every tier and the payment methods.

Which checklist question settles the first step and who owns the call?

Question 10: what is the first step, and who owns the final call? The first step should be small, dated, and tied to one real decision. The executive keeps the final call, and the engine documents the basis for it, which makes the decision easier to explain to a board.

At EQOS Asia, the first step is a free 30-minute pre-diagnostic by video in Singapore time (SGT, UTC+8) that applies the equation to a real case (EQOS Asia contact page, as of October 2026). Score each answer from 0 to 2 across the ten questions, and ask for the pre-diagnostic when one engine leads.

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