How it works | Signal Bench

Start here

A useful reading begins with three questions:

What is the direction? Improving, deteriorating, or neutral describes the latest movement.
What drove it? Source contributions show which public indicators supported or opposed the reading.
How much context is available? History, source dates, and evidence help you judge how much weight to place on it.

Start with the public signal catalog or browse the research themes.

Reading a signal

A signal combines related economic readings into a score from −1 to +1. Values near zero are closer to neutral. Values farther from zero indicate a broader or stronger reading within that signal's own recent history.

Read the score with its direction, date, history, and source contributions. A single score is a summary of available evidence, not a complete economic forecast.

How signals are built

Signal Bench begins with official public economic series. Each source reading is placed in recent historical context and aligned so its contribution has a consistent interpretation. Related contributions can then be combined into public signals for growth, inflation, labor, rates, liquidity, credit, housing, and markets.

Calculations use observations available by each reading date. A monthly or quarterly signal may remain unchanged between source releases even when the website is refreshed more frequently.

Technical definitions and formulas

Normalized component score

Recent percentage movement is aligned to the source direction and bounded from −1 to +1.

directional score = clamp(percentage change × polarity / 10, −1, +1)stable-preferred score = −min(abs(percentage change) / 10, 1)

Combined signal score

Available component scores receive equal weight.

combined score = mean(available component scores)

Definitions

Direction describes whether the current score is improving, deteriorating, or neutral.
Strength groups the distance from neutral as weak, moderate, or strong.
Confidence reflects available history and source coverage.

Release surprise

Release surprise compares the latest observation with that series' own recent pattern. A positive or negative reading describes the direction of that deviation. It does not describe the overall strength of the economy.

Release surprise formula

z = (latest value − trailing mean) / trailing standard deviation

Monthly, weekly, and daily series use trailing windows suited to their release cadence. Each value is bounded from −3 to +3, aligned to its interpretation, averaged with available inputs, and divided by 3 for a score from −1 to +1.

Use the release surprise API

Evidence and API access

Evidence shows how public signals aligned with later target movement across stored historical samples. Use the dedicated Evidence page for results and sample context.

Explore Evidence Read the API docs

Public research does not require an account. Create an API key when you need API or MCP access.

Sources and limits

Source names, observation dates, descriptions, and attribution appear with the research so readers can trace the public data behind a reading.

Not trading advice. Signals provide contextual research and are not investment recommendations.
Not real-time. Readings follow source publication schedules and do not reflect intraday changes.
Not causal. Relationships and historical evidence do not establish cause and effect.

This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis.