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

# Account Health

> How Superhawk judges account health, and how to read the verdict.

Every CS account carries a health verdict: a band, a headline, a short summary, and the specific factors behind it. It is a judgement made from evidence, not a score computed from a formula.

## The bands

| Band                  | What it means                                           |
| --------------------- | ------------------------------------------------------- |
| **Green**             | The account is in good shape on the evidence available  |
| **Amber**             | Something is worth watching, but it is not yet a crisis |
| **Red**               | There is real risk here and it needs attention now      |
| **Insufficient data** | Superhawk does not have enough to make a call           |

<Warning>
  **Insufficient data is not green.** It is shown as a neutral state precisely so a thinly-covered account cannot hide among the healthy ones. If you see it, the fix is usually a missing integration or a genuinely quiet account, not a broken health read.
</Warning>

## What a verdict contains

**Headline.** The single most important truth about the account right now, in plain words.

**Summary.** A short paragraph giving the shape of the situation.

**Factors.** The specific drivers behind the band, each labelled and each carrying its evidence. Factors are split by polarity:

* **Working** -- strengths holding the account up
* **Watch** -- risks pulling it down

A factor looks like "Creation slowing -- 49% down quarter on quarter, 0 created this month". A short name, then the fact behind it.

**Confidence.** How sure Superhawk is, given how much it could see: **high**, **medium**, or **low**. A red band at low confidence deserves a different response from a red band at high confidence -- the first is a prompt to go and look, the second is a prompt to act.

## What it reads

The judgement draws on everything connected:

* **Product usage** -- volume, trend, breadth, and which features
* **Calls** -- what was said, by whom, and what tone it was said in
* **Email** -- responsiveness and who is still in the thread
* **Support tickets** -- volume, escalations, and recurring themes
* **CRM** -- contract value, renewal date, stage, and owner
* **Memory** -- what has been true about this account over time
* **Signals** -- observations already raised

Coverage drives confidence. An account with calls, tickets, and usage gets a higher-confidence read than one where only the CRM is connected.

## Reading a verdict well

**Read the factors, not just the band.** The band is a sorting aid. The factors are the thing you act on.

**Check the polarity split.** An account with four Watch factors and no Working ones is in a different position from one with three of each, even at the same band.

**Take confidence seriously.** A low-confidence read is Superhawk telling you it is working with thin evidence. Treat it as a prompt to check, not as a conclusion.

**Look at what changed, not just where it stands.** An account that moved green to amber this week is more urgent than one that has been steadily amber for a quarter.

<Note>
  Health is recomputed on a schedule as new evidence arrives. A verdict carries the time it was computed, so you can tell whether it has caught up with a call that happened this morning.
</Note>

## When a verdict looks wrong

The usual causes, in order of likelihood:

1. **A source is not connected.** Health cannot see what Superhawk cannot read. Check [Integrations](/integrations/overview).
2. **A source has stopped syncing.** Check the last sync time on each connection.
3. **The feature catalog is incomplete**, so usage is being read against the wrong vocabulary. See [Features & Modules](/customer-success/modules).
4. **The evidence genuinely says something you did not expect.** Open the factors and follow them to the panel they came from. Sometimes the read is right and the surprise is the value.

## Related

* [Churn Risk & Renewals](/customer-success/churn-risk-and-renewals)
* [Metrics](/customer-success/metrics) -- what usage is measured against
* [Signals](/signals/overview) -- discrete observations that feed the read