# What is a CRM database? What it holds, how it’s structured, and how to keep it clean

Migaru, Nov 20, 2026 · Guides

> A CRM database holds your companies, contacts, deals and every conversation with them, linked. What goes in one, an example structure, and how to manage it.

A **CRM database** is the store of customer records behind a CRM: the companies you sell to, the people at them, the deals in progress, and the emails, meetings, calls, notes and tasks between you, each linked to the others. Because the records are linked, opening a company shows its people, its deals and its whole history in one place, which is what makes a CRM more than a list of contacts.

Every CRM is a database underneath. What makes one good is not the database software but three decisions: what kinds of records it holds, which fields each one has, and how the records get filled in and kept current.

## What a CRM database holds

- **Records,** one for each company, person or deal, grouped into kinds that CRMs call objects: accounts, contacts, deals or opportunities, and any your business needs, such as partners, projects or subscriptions.
- **Fields** on each record, also called properties or attributes: a company’s website and industry, a person’s email and job title, a deal’s stage, value and close date.
- **Relationships** between records: a contact works at an account, a deal belongs to an account and involves several contacts.
- **Activities** on a timeline: the emails, meetings, calls, notes and tasks, each filed on the records it concerns.
- **History:** who changed a field, when, and from what, so a value can be trusted or traced back.

## An example CRM database structure

For a business that sells to other businesses, a simple structure covers most of what a sales team needs:

- **Account** (the company): name, website, industry, headcount, location, owner, status such as prospect or customer, and where it came from.
- **Contact** (the person): name, email, phone, job title, department, LinkedIn, and the account they work at.
- **Deal** (the opportunity): name, account, contacts involved, pipeline and stage, value, expected close date, next step, and, once closed, why it was won or lost.
- **Activities:** emails, meetings and calls with their date, who took part and a summary, linked to the account, contacts and deal they were about.
- **Tasks:** what has to happen next, with an owner, a due date and the record it belongs to.

Start with these and add only what you will report on or act on. A field nobody fills in is worse than no field, because it makes every record look incomplete.

## CRM database vs. spreadsheet

A spreadsheet is a fine first CRM for one person. It stops working when more than one person sells, for reasons that are all about structure:

- **Rows aren’t linked.** A contact’s company is text, so renaming the company, or spelling it two ways, splits its history.
- **There is no timeline.** The emails and meetings live in inboxes and calendars, not beside the record.
- **Nothing checks for duplicates,** or stops two people overwriting each other.
- **Nothing has a history,** so nobody knows which value is current or who changed it.
- **Nothing reminds anyone** that a follow-up is due.

## CRM data management: keeping the database clean

- **Give every record a natural key.** An email address for a person and a web domain for a company. Matching on them is what stops duplicates at the door.
- **Prefer choices to free text.** A stage, an industry or a lost reason picked from a list can be counted; one typed by hand can’t.
- **Keep required fields few.** Every required field is one more reason to skip logging, or to type something meaningless to get past it.
- **Give every record an owner,** who is the one to notice when it is wrong.
- **Import with IDs.** Keep the old system’s record ID on each import, so running it again updates records rather than doubling them.
- **Archive rather than delete,** so history survives and a mistake can be undone.
- **Capture activity automatically.** The biggest source of bad CRM data is the conversation nobody logged. A CRM that reads email and calendar fills the gap without anyone typing.
- **Review duplicates and stale records on a schedule,** not when the problem becomes visible in a forecast.

[CRM hygiene](https://www.migaru.ai/blog/crm-hygiene) goes further into cleaning a database that has already drifted.

## Building a CRM database from what you already have

Most teams already hold their CRM data, scattered: in spreadsheets, an old CRM, inboxes and calendars. Bring it together in this order:

- **Companies first,** with their websites, so that people and deals have something to be linked to.
- **Then people,** with a column naming their company, so each is linked to the right account.
- **Then deals,** with their stage, value and close date, linked to their accounts.
- **Then connect email and calendar,** so the history of each relationship arrives with it rather than being typed in.

[Switching from another CRM](https://www.migaru.ai/migration) has the steps for each one.

## How Migaru’s database is put together

A Migaru workspace starts with seven kinds of record: accounts, contacts, opportunities, tasks, meetings, emails and notes. Every record has an owner, a created date, and a last and next interaction worked out from its mail and meetings. Add objects, fields and relationships of your own on any plan.

Records are built from the mail and meetings you already have: connect Gmail or Outlook with its calendar and contacts and accounts are created from the people you email and meet, people matched by email address and companies by web domain. Each thread and meeting is filed on the records it’s about. Email is kept by its subject, preview and summary, not its body.

Some fields are written for you, such as each record’s summary, and others are suggested for somebody to approve, such as a deal’s stage or an account’s industry; an admin sets which, field by field. Likely duplicates are flagged about every half hour for you to merge, and never merged by themselves. Every field shows its last 25 changes and what made each one: a person, sync, an import, the API or the agent.

And it is yours to take out: any view exports for Excel or as CSV or JSON, an owner or admin can export the whole workspace with its history, files and call transcripts, and every record can be read through the REST API or by Claude, ChatGPT and other assistants over MCP.

## Frequently asked questions

### What is a CRM database?

The store of customer records behind a CRM: companies, contacts, deals, and the emails, meetings, calls, notes and tasks between you, each linked to the others so a company’s record shows its people, deals and history together.

### Is a CRM a database?

Every CRM is built on a database, but a CRM adds what a bare database lacks: a structure designed for customers and deals, a timeline of conversations, a pipeline, reminders, reports, and permissions for the people who use it.

### What should a CRM database include?

At least accounts for companies, contacts for people, deals with their stage, value and close date, and the activities and tasks linked to each. Add other kinds of record and fields only when you will report on them or act on them.

### How do you keep a CRM database clean?

Match people by email address and companies by web domain to stop duplicates, prefer pick lists to free text, keep required fields few, give every record an owner, import with the old system’s IDs, and capture email and meetings automatically so nothing depends on someone remembering to log it.

### What is the difference between a CRM database and a marketing database?

A CRM database is organised around relationships and deals: accounts, the people at them, and every conversation in progress. A marketing database or customer data platform is organised around audiences and behaviour, such as email engagement and website visits, and is usually much larger and less personal.

---

Canonical page: https://www.migaru.ai/blog/crm-database
Last updated: 2026-11-20
