> For the complete user guide index, see [llms.txt](https://portraitresearch.com/documentation/llms.txt).

# Mosaic overview

[Canonical page](https://portraitresearch.com/documentation/mosaic/overview/)

## Applicability

- **Task:** Understand when to use Mosaic and how its research stages fit together
- **Access:** Mosaic-enabled Portrait account; Easel access; applicable credits and data entitlements
- **Audience:** Analysts with Mosaic access
- **Surfaces:** web
- **Workflow status:** current
- **Availability:** Account-gated
- **Content type:** explanation
- **Last reviewed:** 2026-09-22

Mosaic applies the same research instructions across a set of companies and keeps the resulting research so you can inspect it and build on it. Use it when you want to test a research idea across a universe, compare the findings, or collect the same information for several companies.

Mosaic is available in the web app for enabled accounts and is not available on iOS. If **Mosaic** is absent from your navigation, [contact Portrait](https://portraitresearch.com/documentation/more/contact-us/) about access. Seeing **Mosaic usage** in Settings does not mean the tool is enabled.

## What you create

You create Mosaic work by chatting with **Easel** in the Mosaic page. Describe the companies, criteria, evidence, and result you want. Mosaic saves these instructions as a **framework**. The interface lists these saved items as **Mosaic screens**; a framework is the set of instructions behind one of them.

| Kind of work | Result |
|---|---|
| **Screening** | Company research judged against your conditions, with an optional ranking against your ranking criteria |
| **Task** | Research that collects the requested information for each company, without a ranking |

For a Task, an unsuccessful result means the research task did not complete. It does not mean the company failed an investment condition.

## How the research progresses

Mosaic calls each execution of a stage a **pass**. The stages have different purposes:

1. **Quant** defines and checks the company universe using quantitative filters.
2. **Research** investigates the companies using your instructions. Start with a **trial** of up to 20 companies, inspect the findings, then expand to **full** research.
3. **Ranking**, for screening work, compares researched companies against your ranking criteria. You can rank the trial before deciding whether to expand.

The trial is real research and consumes credits. It is a small sample for checking your instructions and estimating the larger job. Review [Mosaic usage and estimates](https://portraitresearch.com/documentation/mosaic/usage/) before starting it.

## What you can inspect

Open a saved Mosaic screen to review its instructions, execution history, company outcomes, and available research files. Later research can add to the same collection of files. Selecting an older pass does not turn that collection into an isolated historical copy.

In screening work, a ranking presents relative scores and summaries. Scores are not probabilities, valuations, or expected returns. Check the underlying research and the ranking criteria before using the ordering.

## Mosaic or Screening

Use [Screening](https://portraitresearch.com/documentation/screening/introduction-to-screening/) to apply explicit quantitative and textual filters and review company-level results. Use Mosaic when you need a reusable research framework, deeper company dossiers, and successive research or ranking stages. [Compare research tools](https://portraitresearch.com/documentation/reference/compare-research-tools/) places both alongside Easel, Deep Research, and Monitors.

Continue with [Research companies with Mosaic](https://portraitresearch.com/documentation/mosaic/research-companies/) for the complete workflow, including checking results, changing instructions, and stopping work.

## Historical dates

[Mosaic backtesting](https://portraitresearch.com/documentation/mosaic/backtesting/) uses a ready completed month-end snapshot and a fixed framework date. Ask Easel for the available dates. Current company relationships and possible hindsight in authoring are explicit limits; do not interpret a historical date as a universal point-in-time guarantee.