> ## Documentation Index
> Fetch the complete documentation index at: https://docs.noonum.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Signals and scores

> What each holding-level signal and the strategy-level LSI mean, how to read their ranges, and how to act on them.

## Overview

Every strategy ranks its holdings using a small set of signals. Each one answers a different question about a company's relationship to the theme.

The holding-level signals you will work with:

* **Conviction score**: the default ranking metric, for overall thematic strength.
* **Linguistic beta**: how strongly the company ties to the theme.
* **Market buzz**: recent thematic media momentum around the company.

One strategy-level metric, the **Linguistic Strength Indicator (LSI)**, tells you about the quality of the strategy as a whole. **Sector concentration** shows how the strategy is distributed.

## Conviction score (0–1)

The conviction score represents the overall strength of a company's thematic connection. It is the default sort for holdings.

Conviction is independent of company size. A small company with a strong, focused thematic connection can outrank a much larger one, so you can compare companies on thematic merit.

| Range   | Label           | Meaning                                      |
| ------- | --------------- | -------------------------------------------- |
| > 0.5   | High conviction | Strong thematic fit across multiple signals  |
| 0.2–0.5 | Moderate        | Decent fit, worth investigating further      |
| \< 0.2  | Low conviction  | Weak overall signal, candidate for exclusion |

**How to use it:**

* Rank holdings: sort by conviction to see the strongest thematic names first.
* Set an inclusion bar: drop holdings below a conviction threshold to tighten a strategy.
* Compare across sizes: because conviction is size-independent, a high score on a smaller name is a genuine thematic signal.

## Linguistic beta (0–1)

Linguistic beta is a company's stable, structural exposure to a theme. It captures how persistently the company ties to the theme over time. A higher value means a stronger, more direct tie.

| Range   | Label      | Meaning                                               |
| ------- | ---------- | ----------------------------------------------------- |
| > 0.5   | Core       | Strong, direct connection to the theme                |
| 0.2–0.5 | Supporting | Moderate connection, often through a business segment |
| \< 0.2  | Tangential | Weak connection, candidate for exclusion              |

**How to use it:**

* Defend an inclusion: use a holding's linguistic beta as the evidence that it belongs in the theme.
* Judge strategy quality by reading the distribution across holdings:
  * Most holdings above 0.5 → a tightly focused strategy.
  * Spread across the whole range → a broad or dilute strategy.
  * Clustered around 0.2–0.3 → the theme may be too niche, with few strong matches.

## Market buzz

Market buzz has no fixed 0–1 scale, so there is no range table. Compare it within a strategy, not across strategies.

Market buzz reflects recent thematic news and media activity around a company. A higher value means more current coverage related to the theme.

**How to use it:**

* Spot momentum: find companies gaining or losing thematic relevance right now.
* Surface trending names worth investigating.

## Linguistic Strength Indicator (LSI)

The LSI is a strategy-level metric. It reflects how well the overall theme is represented in current market narratives.

**How to use it:**

* Judge strategy quality: a stronger LSI means the theme is well supported by current narratives across its holdings.
* Compare iterations: watch the LSI as you refine an objective to see whether your changes sharpen or dilute the theme.

## Sector concentration

Read sector concentration from the holdings data to understand how the strategy is distributed. Watch for:

* **Over-concentration (above 40%)**: the strategy may really be a sector bet.
* **Unexpected sectors**: could be genuine thematic connections or noise; check the
  inclusion reasoning.
* **Missing expected sectors**: the strategy might be too narrow, or its exclusions too
  aggressive.

## Display conventions

* **Conviction** is the default ranking metric. Use that label in table headers.
* Say **linguistic beta**, not "beta" alone, because "beta" by itself means financial beta. In
  tight headers, "Ling. Beta" or "Ling. β" is fine.
* **Market buzz** can be shortened to "buzz" or "Mkt Buzz".
* When you narrate these scores in prose, display conviction, linguistic beta, and market
  buzz as **percentages** (0.17 → 17%, 0.0001 → 0.01%). Holdings tables from the API are
  already pre-formatted as percentages, so leave them as-is.

## Next step

Once you can read these signals, see [Portfolio optimization](/base/concepts/portfolio-optimization) for how they translate into holding weights.