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

# Forecast confidence

> How much weight a forecast period carries — six weighted dimensions, the bands, and the ladder that decides what Beacon may do on its own.

<Note>
  This page describes a designed model, not shipped behaviour. The dimensions, weights and thresholds below are fixed; the published names and door availability are being authored into the field catalogue and may change before these values are callable.
</Note>

## What it is

`fci_score` is a 0-100 measure of how much weight a forecast period carries, where higher means more. It is produced once per forecast period — not per account, not per segment — and its main job is to decide what Beacon is allowed to do without a person.

**This is the one number Beacon reasons rather than calculates.** Every other published figure runs through a fixed formula. This one is composed across domains, from metrics the calculation layer has already produced, and it is recorded with its full inputs so any score can be reconstructed. It never overwrites a forecast, never changes a figure, and cannot touch a sealed one.

## How it is calculated

Six dimensions are scored and combined at fixed weights.

| Dimension            | Weight | What it reads                                                                                        |
| -------------------- | ------ | ---------------------------------------------------------------------------------------------------- |
| Data quality         | 0.25   | System-level confidence in the connected sources, plus freshness, schema stability and anomaly count |
| Pipeline stability   | 0.25   | Velocity variance, stage distribution and probability drift                                          |
| Renewal stability    | 0.25   | Renewal-book concentration and movement, and usage decline in the top quartile                       |
| Financial integrity  | 0.10   | Recurring-revenue anomalies against run rate                                                         |
| Expansion visibility | 0.10   | How much of the expansion picture is observable                                                      |
| Intent strength      | 0.05   | Demand signal ahead of the pipeline                                                                  |

Data quality is the largest single lever: [system confidence](/data-integrity) is 0.60 of it. A connected system falling below 85 pulls the whole index down.

| Band        | Range    |
| ----------- | -------- |
| High        | 80-100   |
| Medium-high | 70-79    |
| Medium      | 60-69    |
| Medium-low  | 50-59    |
| Low         | 40-49    |
| Critical    | Below 40 |

**A dimension whose source is not connected contributes a neutral 50 rather than a zero**, and the period is marked as partial coverage. A gap is declared, never scored as bad news.

## What it governs

The score sets what Beacon's agents may do, directly and without a separate setting.

| Score        | What Beacon may do                               |
| ------------ | ------------------------------------------------ |
| 80 and above | Act within the grants it already holds           |
| 60-79        | Draft and wait for approval                      |
| 40-59        | Recommend only                                   |
| Below 40     | Observe. Nothing is proposed and nothing is sent |

The bottom step is unconditional. No instruction, no override and no stated urgency lifts it.

## Where it comes from

| Source            | What it supplies                                            |
| ----------------- | ----------------------------------------------------------- |
| Billing           | Required. Renewal book, recurring-revenue movement          |
| CRM               | Required. Pipeline stage, velocity and probability movement |
| Product usage     | Usage decline, which feeds renewal stability                |
| Finance or ERP    | The revenue anomaly check                                   |
| Marketing sources | Intent strength, the smallest dimension                     |

## How fresh it is

Recalculated each forecast cycle. A drop of five points or more inside twenty-four hours opens an anomaly review rather than passing unremarked. Every score is written to an append-only record with its inputs, the state of each source, and the version of the reasoning that produced it.

## Currency and rounding

Neither applies. The index holds no money value; it reads revenue only as an anomaly signal.

## What changes it

A connected system's confidence falling. Deal velocity moving more than a quarter against its benchmark. Probability drifting across the pipeline. More than a fifth of renewal revenue becoming unstable. A usage drop in the largest accounts. A recurring-revenue anomaly against run rate. Concentration also weights it: a heavily enterprise book raises how much pipeline stability counts, and a heavy renewal quarter raises how much renewal stability counts.

## What it is not

* **Not a forecast, and not a probability that the forecast is met.** It measures how much of the picture Beacon can see, not what the number will turn out to be.
* **Not [data integrity](/data-integrity).** That measures one record and is calculated. This measures a period and is reasoned.
* **Not per account or per segment.** A forecast period carries one score.
* **Not a control you set.** The thresholds are fixed and the bottom of the ladder is not negotiable.