DOCS / UNDERSTAND

Understand

Business-state baselines, evidence collection, and structured market signals.

The Understand stage establishes what is happening before asking what should be done.

Business State Baseline

The baseline accepts longitudinal tabular data, proposes column roles, and prepares suitable numeric series for analysis. A reviewer confirms the mapping before semantic interpretations become eligible for downstream use.

The baseline can:

• align compatible time series;
• identify co-moving dimensions using principal component analysis;
• test the stability of those dimensions;
• compare conservative forecasting baselines with low-power linear challengers; and
• propose, subject to analyst approval, questions or evidence domains worth investigating.

How to interpret it

A principal component is a compressed pattern of co-movement. It may resemble concepts such as volume, value capture, or operational pressure, but the label is an interpretation rather than a computed fact. Forecasts are conditional continuations, not promises and not estimates of intervention effects.

Market evidence

Marres can collect or accept public market material and evaluate it against analyst-defined domains. A typical run distinguishes:

• total material collected;
• material relevant to the question;
• successfully scored material;
• domain coverage;
• sentiment distribution; and
• high- and low-scoring evidence examples.

Scores without coverage should not be used. A domain supported by a handful of ambiguous items is materially different from one supported by broad, recent, relevant evidence.

Human review points

Human confirmation is expected when:

• column meanings are ambiguous;
• domain definitions are proposed or changed;
• a competitor may not be genuinely comparable;
• automated relevance decisions appear systematically wrong; or
• external events may explain an observed pattern.

Appropriate claims

This stage supports claims about observed patterns, evidence distributions, and conditional baselines. It does not by itself establish why a pattern occurred or what an intervention will cause.

Related product page: Understand — product overview

Documentation draft · Source baseline 16 August 2026