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

# The eight signals

> What Forepost watches and why each one earned its place.

Most CX dashboards offer forty metrics and ask you to pick. Forepost takes the opposite stance: there are eight numbers worth watching every day. Adding more dilutes attention; cutting fewer drops the read.

This document is the rationale behind each one.

## 1 · CSAT

The single best lagging indicator of customer experience. If CSAT is moving, something *already* changed; the question is what and how recently. Forepost reads the trend (month-on-month delta) more than the absolute value, a 91% holding flat tells a different story than an 88% climbing.

## 2 · Weekly volume

The denominator of everything else. CSAT on 50 tickets is noise; CSAT on 5,000 is signal. Volume also drives the [capacity model](/help/horizon), every projection downstream of Horizon starts here.

## 3 · First response time

The earliest visible signal of strain. When a team is overloaded, response time degrades before CSAT does, because customers tolerate slow-but-good for a while before churning. Forepost watches it as a leading indicator of CSAT decay.

## 4 · AI deflection rate

The economic shape of your function. Going from 50% to 70% deflection without dropping CSAT is the difference between hiring and not. Watching the rate over time reveals whether your AI investments are working, or plateauing.

## 5 · Open tickets

The instantaneous queue depth. Useful as a "right now" signal that complements weekly volume. Sudden spikes flag incidents; persistent depth flags structural backlog.

## 6 · Agent utilisation

How close your team is to its ceiling. 60% is comfortable; 85% is the cliff. The Headcount calculator uses this directly to project capacity-breach week.

## 7 · Backlog age

Days since the oldest unresolved ticket. The metric that catches forgotten issues. Anything older than 21 days is a problem, it's been re-prioritised down the queue every day until it disappeared from view. Forepost surfaces this directly in the Daily Brief.

## 8 · One-touch resolution

The cleanest measure of agent efficiency. High one-touch means the team is closing things on first reply, fewer back-and-forth threads, lower per-ticket cost. Low one-touch means tickets are bouncing around. The brief uses this to read efficiency; the Headcount calc uses it indirectly via deflection.

## Plus four derived signals

Computed from the eight, surfaced on the [Watchlist](/help/watchlist):

* **Capacity headroom**: weeks to breach
* **Volume velocity**: % week-on-week
* **CSAT trajectory**: pp/month
* **Agent review queue**: count of flagged agents

These don't add inputs; they add interpretation. Twelve named things. Not forty.

## What we don't watch (and why)

| Metric             | Reason for omission                                                                                                                            |
| ------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------- |
| Time to resolution | Heavily skewed by long tail; mean is meaningless and median is hard to compute consistently across helpdesks. Backlog age is the better proxy. |
| Tickets per agent  | Reductive. Two agents handling identical volume can have wildly different difficulty mixes. We surface flagged agents instead.                 |
| NPS                | A different time horizon (quarterly) and a different audience (full customer base, not support population). Belongs in a different system.     |
| Tags / categories  | Helpdesk-specific. Once integrations land, we'll surface top tags inside the brief, but they aren't part of the core eight.                    |

The opinion is the product. Every addition costs attention.