Which inputs does a decision engine use to score a situation?

What counts as an input in a decision engine?

An input is any fact, rating, or condition that the equation reads to produce a score. A revenue figure is an input, and so is an expert rating of leadership alignment, or a regulatory setting. Anything the equation reads can move the score, which is why the input list defines the scope of the verdict.

Defining inputs precisely keeps every later discussion concrete. Each input has a name, a unit, a range, a source, and a date. When two analysts open the same list, they find the same definitions, and a disagreement about a score becomes a disagreement about a specific input that can be checked. That habit also speeds up reviews, because each question points to one line of the list.

Which three groups of inputs does a decision engine use?

Measured facts come first: revenues, dates, headcount, exposure, and other figures with a unit. Expert ratings come second: structured judgments given on a scale, such as the 0 to 100 sliders of the EQOS Asia simulator. Context comes third: the rules of the game, such as regulation, currency, and jurisdiction. Each group answers a different kind of question.

The groups combine inside one model, and none of them works alone. A measured fact gives the equation a fixed anchor, a rating captures what no database records, and context decides how both are weighed. The table below pairs each group with an example and a typical source.

How much data does a good score need?

Quality matters more than volume. A short list of inputs with a clear definition, a named source, and a recent date gives the equation firmer ground than a large pile of loose data. Every input that enters the model should answer one question: which dimension of the situation does it help to rate?

Coverage still counts. If a dimension has no evidence behind it, the score for that dimension rests on a guess. A careful team notes the gap, collects one document or one interview to fill it, and records the new source. Over time, the list grows only where the evidence justifies it. A list built this way stays readable by a committee, a board, or an outside reviewer who joins the project later.

Two companion guides continue the theme. The guide to data quality for a decision engine lists the checks to run first and a one-page data sheet to keep. The guide to signal vs noise in decision data shows how EQOS Asia case studies report a signal-to-noise ratio and how to raise the share of useful facts in an input list.

Which sources feed the EQOS Asia decision engine?

The Modelling pillar ingests internal, external, and ecosystem data. Internal data covers finance, HR, operations, sales, and product, across entities and countries. External sources include LinkedIn, ICIS, SimilarWeb, and Crunchbase, among others (EQOS Asia method page, as of October 2026).

Qualitative documents join the list: presentations, emails, and meeting notes. The accepted formats are PDF, video, audio, and raw data. The method also describes an illustrative scenario, modeled and not a client result: a regional hub choice across Singapore, Johor, and Bangkok, scored on the same 9 dimensions, with figures in US$ (EQOS Asia method page, as of October 2026).

How are inputs collected in practice?

The standard workflow has four steps. An NDA and scoping step takes 1 day and includes a briefing of 60 to 90 minutes by video, scheduled in Singapore time (SGT). Structured collection takes 2 to 5 days, with internal documents, targeted interviews, and existing data. Analysis takes 3 to 10 days, and a deliverable with a debrief closes the cycle (EQOS Asia method page, as of October 2026).

Interviews last 2 hours at the Standard level and 3 days at the First level. Any format is accepted, digital or physical, in an encrypted isolated environment (EQOS Asia method page, as of October 2026). The figure below shows the sequence.

How do inputs stay current after the first score?

Dated inputs make refreshing simple: when a source changes, the team updates that line and recomputes. The EQOS Asia model is reusable monthly through Sentinel, which re-measures the 9 dimensions and sends alerts when a tipping threshold gets close (EQOS Asia method and pricing pages, as of October 2026).

A monthly rhythm also reveals drift. If one input moves for three months in a row, the team can investigate the cause before the score crosses a line. The trend then becomes a signal of its own, next to the score itself. Sentinel carries a minimum commitment of 6 months and is priced on quote (EQOS Asia pricing page, as of October 2026).

Which inputs should you gather before a first conversation?

Start with one decision, written in a single sentence. Then list the handful of facts that move the answer, the countries and currencies involved, and the person who owns the final call. The contact form of EQOS Asia asks for exactly this short description of the situation, including countries and currencies (EQOS Asia contact page, as of October 2026).

A 30-minute pre-diagnostic by video, scheduled in SGT, then applies the equation to your real case. Bring what you already have, in any format, and the first call sorts it into the three groups of this guide. The pricing page shows which tier matches the depth of inputs you can provide, from a Reading at US$358 to a Mapping that integrates 100% of your data (EQOS Asia pricing page, 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