> ## Documentation Index
> Fetch the complete documentation index at: https://docs.shareofmodel.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Analyses & Collects
> The central data unit of Share Of Model — how analyses are scoped, run, and become available.
An **analysis** is the central data unit of Share Of Model. Everything you read in dashboards or fetch through the API is produced by an analysis: a scoped, configured run that gathers signals from LLMs and search engines and turns them into metrics, perception clusters, and source lists.
Analyses come in two flavours, depending on the module:
* **Brand analyses** — answer "what does the model think about my brand?" and produce awareness, perception, and source signals.
* **Search projects** — answer "where does my brand or domain appear when users search?" and produce visibility, ranking, and citation signals.
The two share the same underlying lifecycle and ownership model.
## Scope
When you create an analysis, you define **what** is being studied. The exact fields differ by module, but the shape is the same:
| Field | Required | Notes |
| --------------- | -------- | ---------------------------------------------------------------------------- |
| **Brand** | Yes | The brand under analysis. |
| **Category** | Yes | The market in which the brand competes. |
| **Competitors** | Yes | Exactly 3 competitors, used as the comparative landscape. |
| **Attributes** | Yes | Exactly 6 attributes used for perception scoring. |
| **LLM models** | Yes | Models to query (OpenAI, Google AI, Meta AI and Anthropic by default). |
| **Regions** | Optional | Up to 8 regions per analysis. Country-prefixed prompts are added per region. |
| **Personas** | Optional | Up to 5 personas per analysis. Each persona narrows the prompt audience. |
| Field | Required | Notes |
| ---------------------------- | -------- | ------------------------------------------------------------------------------ |
| **Website** | Yes | The domain whose visibility is being tracked. |
| **Brand** | Yes | The brand whose mentions are tracked alongside the domain. |
| **Locale** | Yes | Country and language. |
| **Thematics & topics** | Yes | Categorisation hierarchy applied to every search query. |
| **Search queries (prompts)** | Yes | The queries to monitor. Each can carry funnel stage, persona, intent and more. |
| **Sources** | Yes | The engines to monitor — search engines, AI Overviews, generative engines. |
Scope drives credit usage. For brand analyses, credits scale as `regions × personas`; for search, each tracked query costs a known number of credits per run. See the [Credit System](/platform/getting-started/understanding-the-credit-system-for-share-of-model) for the full breakdown.
## Collects
A **collect** is the result of one execution of an analysis. Each time the platform runs an analysis, it produces a new collect that holds the metrics for that run.
* Multiple collects can be **aggregated** to read average performance across a window.
* Collects can also be **compared** to a previous set, to read evolution over time.
* A collect carries a status — only `FINISHED` collects are usable for reading metrics.
Through the [API](/api-reference/introduction), collects are how you ask for data:
* **single collect** — pass one ID in `collect_ids`.
* **aggregated metrics** — pass several IDs in `collect_ids`.
* **evolution** — also pass IDs in `prev_collect_ids` to compare against an earlier window.
## Lifecycle
An analysis goes through three states from creation to insight:
| Status | What it means |
| -------------- | ------------------------------------------------------------------------------------------------------------------------ |
| **Launching** | The analysis has just been created and is being prepared (typically the first 15–20 minutes). Data is not yet available. |
| **Collecting** | Data is being gathered. Trends may shift as more data arrives. Read early signals with care. |
| **Live** | Data has stabilised and is ready for insights. |
Wait at least 2–3 weeks of collection before drawing strategic conclusions from a brand analysis. Perception clusters and attribute scores only stabilise once enough data has been gathered.
## Archiving and reactivation
Analyses can be archived for later reference and reactivated when needed.
* Archived analyses don't consume credits.
* Reactivating an analysis re-consumes credits based on its original scope.
* The period during which an analysis was archived contains **no data** — data is not backfilled.
## Reading the data
Every analysis exposes its data both in the platform UI and via the [Core API](/api-reference/introduction). The same analysis, read through both surfaces, returns the same metrics.
## What's next
How brands and competitors are modelled.
The three pillars analyses produce.
Run your first analysis.
Run your first search project.