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

# Concepts

> Channels, named sources, the review queue, confidence and how revenue is matched.

## Responses

A response is one answer to your question, in the respondent's own words, with who gave it (email, your user id, name) and optional context such as UTM parameters. Responses arrive from the snippet, the API, form webhooks, CSV and paste imports, or Attio.

## Channels

Every response is filed under exactly one channel. New workspaces start with these:

| Channel | Covers |
| - | - |
| AI assistant | ChatGPT, Claude, Perplexity, Gemini, Copilot recommended or mentioned you |
| Founder & exec social | Posts by your founders, executives or employees on LinkedIn, X, YouTube |
| Podcast | A podcast episode or host |
| Newsletter, blog & media | Newsletters, articles, creator and influencer content you don't run |
| Community | Slack or Discord groups, Reddit, Hacker News, Indie Hackers |
| Word of mouth | A friend, colleague, customer or investor |
| Search | Google, Bing and other search engines, without an ad |
| Review sites | G2, Capterra, Product Hunt, "alternatives to" lists |
| Paid ads | Search, social and sponsored placements |
| Event & webinar | Conferences, meetups, webinars, talks |
| Partner & integration | Partners, agencies, marketplace listings, integrations |
| Other / unclear | Too vague to place, or not an answer |

You can rename channels, add your own and archive the ones you don't need under **Dictionary**.

## Named sources

When an answer names something specific, Rumorly links it to a source in your dictionary: a person, an AI assistant, a podcast, a publication, a community, a company. AI assistants, search engines, review sites, Hacker News and Reddit are in every dictionary already; your founders, podcasts and competitors come from onboarding.

When an answer names something that isn't in your dictionary yet ("heard you on Lenny's Podcast"), Rumorly proposes it as a **new** source. It joins your dictionary once you approve it.

## Confidence and the review queue

Obvious answers ("ChatGPT", "google", "G2") are classified instantly by rules. Everything else goes to Claude Haiku. Each classification has a confidence, shown as a three-step meter.

An answer lands in **Review** when its confidence is below 0.6 or when it proposes a new source. From there you can keep the result, correct it, and approve or reject proposed sources. Approving a source links every answer that proposed it. Corrections stick: automatic classification never overwrites them unless you select **Reclassify** on that answer.

## How revenue is matched

Rumorly reads your Stripe customers and subscriptions, then matches each billing customer to one person who answered, in this order:

1. **`rumorly_id`** in the Stripe customer's metadata, equal to the person's `external_id` (your user id) or their Rumorly person id. This is the most reliable; set it if you can.
2. **Exact email** (case-insensitive).
3. **Company domain**, taken from the email. Free-mail domains such as gmail.com are never used for matching.

A billing customer matches at most one person. Each match shows how it was made, and every report shows the share of revenue that couldn't be matched.

## Monthly recurring revenue

MRR counts subscriptions that are `active` or `past_due`. Trials, canceled, incomplete and paused subscriptions don't count. Quantities and discounts are applied, yearly and multi-month plans are divided into months, and other currencies are converted into your reporting currency (set under **Settings → General**). A customer's revenue goes to the channel of their most recent classified answer.

## First-touch comparison

The snippet records UTM parameters, the referrer and the landing page from a visitor's first visit. Rumorly turns those into a first-touch channel and shows it next to what the customer said. The gap between the two is usually the point: AI assistants and founder posts tend to show up as "Direct" in click data.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.