---
url: /handbook/discovery/company-research.md
description: >-
  Add a company by its website, research it, and keep the research current
  without losing what people confirmed.
---

# Company research

Company research builds a first picture of a company from public sources before anyone at the company is asked anything. It fills the Company Snapshot and the research files of the [Company Repo](company-repo.md). Every finding keeps its source and stays **Inferred** until a person at the company confirms it.

## Before you start

* Research needs a research source. Save an Exa key in **Settings › Research & AI keys**, or use a source that your operator set up on the server. Until one exists, Home shows **Connect research**.
* Research itself needs no AI connection. Other steps do. Approve an AI key for each company in **Settings › Connections › Approve firm key for this company**.

## Add a company

### Open the dialog

On **Home**, select **Add a company**. It opens **Companies › Add account**.

### Choose the tab

Choose **Already engaged in DutyGraph** for a company you already work with. Write one line in **Why is it active?**. Choose **From your book of business** for a prospect.

### Enter the details

Enter the company name, the website, the contact's name and title, and the contact's work email. Select **Add account**.

### Run research

Select **Research `<company>` now**. If you closed the dialog, open the account in **Companies** and select **Run company research**.

### Open the workspace

When the run ends, select **Open company workspace**. Read the Overview, then open **Repo**.

DutyGraph creates an account, a contact and an engagement for the company.

## What research gathers

One research run covers the company profile and the deeper lanes. There is no separate "deep" button.

| Area | Examples |
| --- | --- |
| Profile | Identity, size, headcount trend, locations, ownership, business model |
| Market | Customer segments, where the company sits in its value chain, market trends |
| Competitors | Candidates and confirmed competitors, enriched with size and name data |
| Hiring | Open roles and the skills they ask for |
| Technology | Tools that research sources detect |
| News | Dated events and announcements |

Research keeps only professional fields for people: name, title, department and location. It removes personal email addresses, phone numbers and street addresses. Each run stays inside the budget of the account's research authorization.

## Where the findings land

* **Overview (the Company Snapshot).** The company's front page. Each fact shows its basis, such as Client confirmed, Reported, Inferred, Conflicting or Not found. A percentage shows only next to its basis, for example "Inferred · 77%". Nothing unknown shows as 0.
* **Repo files.** Findings fill `company/profile.md`, `market/` and `systems/`. Open **Details** on a file to see each finding's sources.
* **The company shape.** From the findings, code proposes the kind of business, the departments and the main processes. The department files show as **Proposed** until you accept the structure.

If research finds nothing, the Overview says "Research ran but found no facts". DutyGraph does not invent a value.

::: warning
Research describes the company from the outside. It does not establish internal duties, owners or reporting lines. Those come from the [sponsor packet](sponsor-packet.md), the [executive kickoff](executive-kickoff.md) and [team asks](team-asks.md).
:::

## Facts a person confirmed always win

A fact that a person at the company confirmed always wins over research. The same is true when the advisor confirms or corrects a fact. New research cannot outvote it, relabel it or remove it.

When newer research disagrees, DutyGraph keeps the confirmed value. It shows the difference in **Repo › To review**: "A person confirmed X. Newer research says Y. The confirmed value stays. Keep it, or ask the person again."

The Company Research step page (**Discovery › Plan › Company Research**) lists these under **Kept because a person confirmed it**.

## Run research again

Select **Run research again** in the Repo's **⋯** menu or on the Company Research step page. It looks for new or missing facts. It keeps the Repo's folder structure and every confirmed fact.

To start the structure again, select **Rebuild the company structure** and confirm. Confirmed facts and answers stay. DutyGraph deletes nothing.

## Research refresh

Some files change over time. Each has a refresh plan.

| Plan | File | How often |
| --- | --- | --- |
| News and announcements | `market/signals/` | Weekly |
| Competitors | `market/competitors.md` | Monthly |
| Hiring | `company/profile.md` (hiring fields) | Every 2 weeks |
| Detected technologies | `systems/` | Monthly |
| Company profile | `company/profile.md` (other fields) | Every 3 months |

People, roles, duties, Task Cards and workflows never refresh from the web.

Each refreshing file has a `refresh:` block in its frontmatter. The block shows the plan, the cadence, the fields, the sources, `last_checked` and `next_check`.

### Refresh now

1. Open the file in the Repo and select **Details**. Or open **Refresh by file** on the Company Research step page.
2. Select **Refresh now**.
3. Read the result line, for example "3 new findings · 1 flag · nothing changed".

The same plan cannot refresh again within 10 minutes. A refresh never writes a value. New values arrive as findings to review.

### Keep research current

**Keep research current** refreshes each file when it is due. It is off by default.

1. Turn on **Keep research current** on the Company Research step page.
2. Enter the **Monthly cap ($)**.
3. Select **Save cap**.

Scheduled refreshes stop when the spend reaches the cap. A **Refresh now** does not count against the cap. The schedule skips sample companies and companies with no research source.

## Simulated research

A run with the simulated provider says "Simulated (not real data)". The Overview shows a banner. It is not real research.

Simulated research runs only for sample companies, or on a local development server. A hosted real company never gets simulated findings. Without a research source, its fields stay unknown.

::: info
**Status.** **Built:** adding a company by its website, one research run with profile and deeper lanes, the Company Snapshot, research files in the Repo, the proposed company shape, confirmed facts that win over research, **Run research again**, **Rebuild the company structure**, 5 refresh plans with **Refresh now**, and **Keep research current** with a monthly cap. **Simulated:** research on the sample companies, labelled "Simulated (not real data)". **Planned:** a live news or competitor feed, an editor for refresh sources and cadence, a notice when a scheduled refresh finds something, a single "add and research" button for an engagement, and **Delete company**.
:::
