What is the difference between a market signal and market noise?

Both are observable, but a signal changes an expectation about the future and noise does not. Whether an observation is one or the other depends on the question being asked of it.

Does a market signal prove that something happened?

A signal is the observable trace of an event, not the event itself. It establishes that something visible changed, while the underlying cause remains a hypothesis that requires separate confirmation.

How many observations does a signal need before it should inform a decision?

At least two independent observations separated in time. A single observation can be a coverage artifact, a reverted test, or a rediscovery of something long-standing.

What is a market signal — and which ones deserve tracking?

Glossary · Continuous Intelligence · 4 min read · last verified 2026-07-21

Reviewed before publication Editorial board Independent commercial review

In shortA market signal is an observable change in a market's visible state that another person could independently check and that carries information about what is likely to happen next.

A market signal is an observable change in a market's visible state that another person could independently verify and that carries information about what is likely to happen next. All three conditions are load-bearing. A change nobody can check is a rumor, and a change that predicts nothing is trivia, however interesting it is to discuss.

The three conditions

Observable. The change exists in something that can be pointed at — published text, pricing, job listings, documentation, partner announcements, the language used in a public forum. Inference about a competitor's internal state is not observation. It may be correct, and it is not a signal.

Checkable. Someone else, following the same procedure, would find the same thing. This is what separates a signal from an impression. Checkability requires that the observation be recorded with enough specificity to be re-performed: what was looked at, when, and what was found.

Informative. The change shifts an expectation about the future. If the same forecast is made with or without the observation, it carries no information regardless of how visible it is.

Signal against noise

Noise is variation the market produces without any underlying change in state. Rewording a page, rotating a headline, a routine documentation update, a departure that reflects a personal circumstance rather than a strategy — all observable, all checkable, none of it informative.

The boundary is not a property of the observation itself. It depends on the question being asked. A pricing page edit is noise for a question about category direction and a strong signal for a question about competitive packaging. This is why a stable benchmark question set does more to separate signal from noise than any filtering rule: with the question fixed, relevance stops being a judgment call made fresh each time.

Signal, event, and interpretation

Three things get conflated, and separating them prevents most bad analysis.

The event — what happened in the world; usually not directly visible
The signal — the observable trace the event left behind
The interpretation — the claim about what the event was and what follows from it

A competitor posting three roles in a new geography is a signal. The event might be an expansion, a hedge, a single executive's initiative, or a job description template being reused. The interpretation is a hypothesis about which, and it should be recorded as one. Signals are found; interpretations are argued.

Qualifying a candidate

Run a candidate observation through these in order, and stop at the first failure:

Can it be pointed at? If the answer is a feeling about momentum, it is not a signal
Is it recorded so it can be rechecked? Source, date, and captured content, or it is not checkable in three months
Is it a change? A competitor that has always described itself a certain way is a fact, not a signal; only the difference from a prior state qualifies
Against what prior state? Change is defined relative to a recorded baseline, which is what an intelligence baseline exists to supply
What expectation does it move? Name the specific forecast that changes, or classify it as noise
What would explain it innocently? List the boring explanations first; most candidates die here, and that is the step working correctly

Strength

Signals are not binary. Three properties determine how much weight one carries.

Persistence. A change still present at the next observation is substantially stronger than one that reverted. Reversions are informative too — they suggest a test rather than a commitment.

Corroboration. Independent observations pointing the same way multiply in strength. Careful: two sources that both derive from the same press release are one observation, not two, and the distinction is easy to lose.

Specificity. A signal consistent with only a few explanations is worth more than one consistent with many. "Hiring in a new region" narrows less than "hiring a regionally licensed compliance role."

Common false signals

Rediscovery — something long-standing, newly noticed; the change is in the observer, not the market
Coverage artifacts — a source added to the monitoring set makes its content appear to be new
Aggregator echo — one primary event reported across many outlets, read as many events
Seasonal regularity — a pattern that recurs annually and gets interpreted fresh each year
Confirmation-shaped — an observation noticed only because it supports an existing belief, where the contrary observation was equally available

Recording one so it stays useful

A signal recorded as a sentence in a meeting note is lost within a quarter. What survives is a record containing the observation, the source, the observation date, the prior state it differs from, the interpretations considered, and what would confirm or eliminate each. That structure is what allows the same signal, reappearing later, to be recognized rather than rediscovered — and it is the evidence trail that makes the difference.

What to watch

Watch for signals that get promoted to facts without a second observation. The characteristic sequence is an observation made once, repeated in a summary, cited in a strategy document, and finally treated as established — with no additional evidence at any step. Requiring that a signal be confirmed by an independent later observation before it can inform a commitment costs one measurement interval and removes most of this failure.