Back to Blog

CRM database: what it is and how to build one that stays clean

A CRM database is a set of linked records about people, companies, deals and tasks. How the major CRMs structure it, how to design, import, clean and export your own, and which rules keep it accurate.

Benjamin WagnerBenjamin Wagnerby Benjamin Wagner
CRM DatabaseCustomer DataData ManagementGDPR

A CRM database, short for customer relationship management database, is the structured store behind a CRM: one record per person, one per company, one per deal, the links between them, and the history of what happened. The links are what separate it from a contact list. Open a company and you see its people, its open deals and the last conversation; open a deal and you see who is involved and what was offered.

You rarely build that database from scratch. You pick a CRM, decide which record types and fields you need, clean your existing data and import it in the right order. This guide covers that design work, shows how four CRMs model their data, spells out how Customermates models the same lists, and ends with a worked example and the export path. Customermates publishes this guide; other vendors' details come from their own documentation, linked at the end.

At a glance

What it is
Linked records for people, companies, deals and tasks, with typed fields and an activity history. A spreadsheet holds rows; a CRM database holds relationships.
How to build one
Fix the record types first, then the fields you filter on, then one unique key per person. Clean before you import, and import companies before people.
What breaks it
Duplicates, free-text fields that should be choices, and nobody owning the cleanup.
Getting data out
Plan the export before you need it. Customermates exports one list at a time as Excel or JSON; there is no whole-account export.
Customermates
Starts without lists: you create the ones you need (contacts, organizations, deals or your own), with typed fields and relationships, from €12 per user per month in the cloud, or self-hosted under AGPL.

The database is only part of the choice: test how people find and update these records using the browser and mobile checklist in the CRM app guide.

What a CRM database holds

Think of it as a handful of tables that point at each other. Most CRMs ship the same core set and differ in what they add around it.

Record typeWhat one record isTypically linked toFields worth having
Person (contact)One human you talk toTheir company, their deals, tasks, messagesFirst and last name, email, phone, role, how you met
Company (organization, account)One business or institutionIts people and dealsName, segment, website, billing address
Deal (opportunity)One piece of potential or ongoing businessThe people and companies involved, products or services soldStage, value, expected start, source
Product or serviceSomething you sell at a priceThe deals that include itName, price
Task or activitySomething someone has to do or didAny of the aboveAssignee, notes

Two design questions decide whether the database holds up later. Can one person belong to several companies (a consultant who sits on two boards, a freelancer working for three agencies)? And can one deal involve several people? If your CRM forces one company per person or one contact per deal, you will start writing workarounds into notes. Read the documentation before you commit: Pipedrive's data article describes linking a person to one organization, while in Customermates a relationship can allow one or several records on each side, so a contact can belong to several organizations.

The other half of a CRM database is history. Every change to a record and every linked email or message forms a timeline. That timeline is what lets a colleague take over a client without a handover meeting, and it is the part a spreadsheet cannot give you.

CRM database examples: how four CRMs structure their data

The record types tell you what a CRM was built for. This table compares the data model only, not price or quality.

CRMStandard record typesYour own record types
HubSpotContacts and companies as the foundational objects, plus deals, leads, tickets, products, projects, tasks and many moreCustom objects, which need an Enterprise subscription
SalesforceStandard objects such as Account, Contact, Lead and OpportunityCustom objects "to store information that's specific to your company or industry"
PipedriveLeads, deals, people, organizations, activities, products, projects and mailThe linked article names no custom object type; Projects is an add-on on Growth and lower plans and standard on Premium and Ultimate
CustomermatesContacts, organizations, deals, services and tasks, each freely renamableYour own lists with typed fields, relationships and calculated fields

The pattern: HubSpot and Salesforce let you add your own object types and configure them, on HubSpot with an Enterprise subscription. Pipedrive's data items stay close to the sales process. Customermates starts without lists: you create the ones you need and name each one, so a contacts list can read Clients and a deals list Projects across the whole workspace, and you can rename them later.

How Customermates models the core lists

Customermates keeps CRM data in configurable lists, and no list is built in. On Customermates Cloud the AI workspace setup designs the lists from your answers; a self-hosted workspace starts empty, and people who manage the data model create the lists in Data model. The setup always creates a people list and Tasks, usually Organizations and Deals, and a catalogue of services or products only when you sell them. A typical result looks like this, as described in the data model documentation.

List (default name)One record isTypically linked toWhat to know
ContactsOne personOrganizations, deals, tasksA Contact details field holds their email addresses, phone numbers and social profiles and matches incoming messages to the record
OrganizationsOne company or accountContacts, dealsAny fields you add, such as segment or client since
DealsOne opportunityContacts, organizations, servicesThe stage is a Single choice field with a win probability per stage; the value is the sum of its line items
Services (or Products)Something you sellDeals, through line itemsA unit price; a line amount is quantity times unit price
Line itemsOne line on one dealIts dealA sub-list: it belongs to one deal, is edited inside it and is removed with it
TasksOne piece of workThe records you link it toA due date or status is an ordinary field you add
Your own listsAnything else: Projects, Applications, PropertiesWhatever you linkThe same field types, relationships and saved views as every other list

A relationship connects two lists. Each side has its own label and allows one or several records, so one contact can belong to several organizations and a project can link several contacts and deals. Both records show the link. On Pro, where connected accounts bring in messages, a relationship can also show the messages of the people it links on the deal or organization.

Fields can be calculated. A deal's value adds up its line amounts, and the weighted value is the deal value times the stage probability divided by 100, which is pipeline weighting, not a forecast. Every record keeps a stable ID, so renaming a list or a field breaks no references, and a saved view shows records without copying them.

CRM database, spreadsheet or plain database?

All three can hold customer data. The question is who maintains it and how much of the logic you want to build yourself.

ChooseWhenYou give up
A spreadsheetOne person, a short list, no shared history neededLinks between people and companies, a per-client timeline, safe editing by several people
A plain database or no-code tool (Airtable, Notion, Access, PostgreSQL)Your records are not people and deals, or you need a model no CRM offersYou build the forms, relations, permissions and history yourself
A CRMSeveral people work with the same clients and you track deals, follow-ups or conversationsSome freedom in the data model; most cloud CRMs charge per user

The same jobs look different in a sheet and in a CRM database:

The jobIn an Excel sheetIn a CRM database
Link one person to two companiesA second row or two names in one cell, so you search twiceOne person record linked to both companies, where the CRM allows several; the link shows on both records
Keep a segment value consistentA free-text column, unless you build and maintain a dropdownA Single choice field with fixed options; rename an option in Customermates and every record follows
See what happened with one clientWhatever someone typed into a notes columnA history of changes and linked messages on the record
Let two people work at onceOne shared file, a version to merge or a lockEach person edits records in place, with access set per list and action
Report on the numbersFormulas and pivot tables you build and keep upDashboard widgets that sum or count records and group them by a field, such as deal value by stage

If a sheet is where you are now, the Excel CRM template guide shows how far one goes and how to lay it out so the later import is easy. For choosing between the kinds of client database tools, see client database software.

How to build a CRM database

These steps apply to any CRM. The order matters more than the tool.

  1. Write down the record types you need

    Start with people, companies and deals. Add a separate type only when it has its own lifecycle. A product catalogue does; a meeting location does not.
  2. List the fields you will filter or report on

    If nobody will ever filter on a field, it belongs in the notes. For each field pick a type: date, currency, email, phone, link, or a single choice from a fixed list. A "Segment" field as free text ends up with "Retainer", "retainer" and "Ret." within a month.
  3. Define your deal stages and what moves a deal

    Five to seven stages with a written exit rule each, for example "Proposal: a written proposal has been sent". Stages without exit rules turn into opinions.
  4. Pick one unique key per person

    Usually the email address. It is how you find duplicates before import and how the CRM matches incoming email to the right person. HubSpot, for example, deduplicates contacts by email address and companies by domain name.
  5. Clean the source data

    One person per row, one kind of value per column, split full names into first and last name, remove duplicates by email, and delete people you no longer have a reason to keep.
  6. Import companies first, then people, then deals

    Each later import can then link to records that already exist, matched by name. Test with 50 rows before the full file.
  7. Save the views people work from

    "My open deals", "Clients without a company", "Not touched in 90 days". A database nobody looks at decays quietly.
  8. Name an owner for data quality

    One person who runs the monthly check below. Without an owner, cleanup is everyone's job and therefore nobody's.

When data does not fit contacts, companies or deals

This is the most common design problem, and there are four honest answers.

  1. It describes an existing record. A contract end date, a client segment, a LinkedIn URL. Add a typed custom field to that record type.
  2. It is something you sell. Put it in the product or service catalogue and link it to deals, so the deal value follows from what is in it.
  3. It is text you read but never filter. Meeting notes, background, preferences. Use the notes field.
  4. It has its own lifecycle and relations. Properties, vehicles, candidates, events. That is a custom object. HubSpot offers them on Enterprise, Salesforce on its platform, and no-code databases such as Airtable are built around them. In Customermates you add it as your own list in Data model, with its own fields and relationships.

Worked example: a CRM database for a four-person consultancy

Here is the same design in Customermates, set up the way the product works today. Customermates is an open-source, AI-native CRM built in Germany for small client-facing teams such as consultants, agencies and founders. A new workspace starts without lists: on Customermates Cloud, describe your business in one sentence and Mate builds the lists, pipeline and dashboard, and the steps below also work by hand in Configure.

Lists. In Data model, open Contacts and change Name and Plural name under General to Client and Clients, then rename Deals to Projects the same way. Any name works. Organizations, Services and Tasks keep their names.

Fields. Each field has a type: Text, Formatted text, Number, Money, Yes or no, Date, Date and time, Date range, Date and time range, Single choice, Multiple choice, Web address, Member, Address, File or Phone, email & social.

ListFieldType
OrganizationsSegment (Retainer, Project, Former)Single choice
OrganizationsClient sinceDate
ClientsRoleText
ClientsLinkedIn profileWeb address
ProjectsEngagement periodDate range
ProjectsSource (Referral, Inbound, Event)Single choice

Relations. A client can link to several organizations, and a project to several clients, organizations and assigned users. A project's value is the sum of its line items, each a service price multiplied by its quantity, so "Strategy workshop" at one price and "Monthly retainer" times six add up on their own.

Stages. Stages are not preset: you add a Stage field (for example Prospecting, Qualification, Demo, Proposal, Negotiation, Won, Lost) and set win probabilities per stage in Data model. The weighted pipeline is the plain arithmetic of those percentages, not a prediction.

Unique key. Email addresses go into the client's Contact details field, which matches incoming messages to the record. One address can belong to several records, so remove duplicates before you import. Global search (Cmd/Ctrl+K) finds a client by name or by any email address or phone number in their contact details.

Import. Use More actions › Add from file with an .xlsx workbook or a .csv file, one list at a time, in the order organizations, clients, projects. Link columns are matched by record name. A CSV file can be imported directly, with no conversion. The records documentation has the full import rules.

Views. Save "Not touched in 90 days" by filtering Updated at with Older than… 90 days, and "Clients without an organization". Share "Not touched in 90 days" with the whole team or with selected roles, so everyone works from the same list.

Cost. Four people on Starter pay €48 per month. On Pro, which adds the unified inbox so each user's connected email, LinkedIn or WhatsApp account shows its conversations next to the client, it is €116 per month. Billing is monthly, and your first workspace starts with Pro free for 14 days without a credit card. Self-hosting the open-source core has no licence fee but no inbox and no Mate.

A routine for data quality (Customermates Cloud). Mate, the assistant in Customermates Cloud, uses hosted AI credits. A routine runs saved instructions on a schedule or on a record event, as the person who owns it; Pro includes 5 routines per user. A weekly routine could carry this instruction:

The last sentence matters. Mate can send messages when an instruction asks for it, so a routine meant only to report has to say so. If you prefer your own AI client, Claude, ChatGPT and other MCP clients can query the same records through 59 MCP tools, on Cloud and self-hosted.

A database earns its keep when each record has a next step. On Pro, the threads join the client record, Mate or your own AI client drafts the reply and the next step, and you review, edit and send. To try that on your own records, start your 14-day trial.

Keeping a CRM database clean

Contact data goes stale as people change jobs and companies merge. A short routine with a named owner beats a yearly cleanup project.

How oftenCheckIn Customermates
WeeklyNew people without a company, new deals without a stageSaved views, or the routine above
MonthlyRecords not updated in 90 days: still relevant?Filter Updated at, Older than… 90 days
MonthlyFree-text values that should be choicesAdd a Single choice field, then set it on up to 100 selected records at once with Edit
QuarterlyPeople you no longer have a reason to keepDelete them from the list or the record page, then from Trash
Before any big changeA copy of what you haveExport each list as Excel or JSON; there is no single whole-account export

An exported workbook can be edited and imported again as Existing records, which updates them instead of creating duplicates. That makes bulk corrections practical without an API.

How to export a CRM database

Plan the way out while you design the way in. You need an export for backups, for bulk corrections in a spreadsheet and for the day you change tools, and you should test it on a small list before you rely on it. In Customermates, every list has the same options under More actions, and an export only contains what your role can read.

You want toDo thisGood to know
Edit many records in a spreadsheetExport as Excel (.xlsx) on the listOne sheet per list and one column per field you can read; search, filters and sorting apply; up to 5,000 records
Keep a backup of a list, links includedExport as JSONThe file includes links for re-import with Add from file, as Create with exported IDs or Update matching IDs
Re-import a large JSON exportUse the REST API or MCPThe app rejects a JSON file with more than 100 records or 500 links; the API exports pages of up to 500 and imports 100 records per request
Move to another CRMExport each list as Excel and import in the order organizations, people, dealsCustomermates has no whole-account export; check the new tool's import rules (Pipedrive, for example, documents importing people, organizations and deals from CSV or XLS files)

The records documentation lists every import and export limit.

Personal data in a CRM database

In the EU, a CRM database of people is personal data under the GDPR. Three principles from Article 5 translate directly into database design: keep only data that is "limited to what is necessary" (fewer fields), keep it accurate (the cleanup routine), and limit how long you store it (the quarterly deletion). Article 17 gives people the right to have their data erased, so you need to be able to find and delete one person completely. Article 28 requires a contract with any provider that stores the data for you.

For Customermates Cloud, the data processing agreement is public, the application database is configured in the AWS Frankfurt region, and the subprocessors, US providers included, are listed. Customermates holds no compliance certification. If you need the data on your own server, self-host the open-source core.

When Customermates is the wrong database

  • You need field-level permissions. Permissions apply per list and action, not per field.
  • You need one-click export of everything. Export runs one list at a time, as Excel or JSON.
  • You send newsletters from your customer list. There is no mass email; use an email marketing tool.

Frequently asked questions

What is a CRM database?

It is the store behind a CRM: records for people, companies, deals and tasks, linked to each other, with typed fields and a history of changes and conversations. The links and the history are what make it more than a contact list.

What data does a CRM database store?

Four kinds. Who a person or company is: names, email addresses, phone numbers, addresses, roles. What you sell and to whom: deals, stages, values, products or services. What happened: emails and messages, notes, tasks and changes to the record. And what you add yourself, such as a segment, a source or a contract end date. Typed fields keep the first and last kinds searchable and countable.

What is the difference between a database and a CRM?

A database stores whatever tables you design and leaves the forms, relations, permissions and history to you. A CRM is a database with that customer logic already built: people linked to companies and deals, pipeline stages, assignees and a timeline per record. You trade some freedom in the data model for not having to build it.

Is a customer database the same as a CRM database?

They overlap. A customer database is the list of customers and what you know about them; a CRM database adds the links between people, companies and deals, the pipeline and the history behind a CRM. If you only need a client list and are choosing a tool for it, see client database software.

Is Excel a CRM database?

No. Excel can hold a customer list, and for one person with a short list that is enough. It has no links between people and companies, no history per client and no safe way for several people to edit at once. A clean sheet with one kind of value per column still makes a good starting point for an import.

What are the four types of CRM?

The usual split is operational (running sales, service and marketing work), analytical (reporting on customer data), collaborative (sharing customer information across teams) and strategic (organizing the business around long-term customer relationships). All four use the same CRM database; they differ in what you do with it.

How do I store data that doesn't fit contacts, companies or deals?

If it describes an existing record, add a typed custom field. If it is something you sell, use the product or service catalogue. If it is only read, use notes. If it has its own lifecycle and relations, you need custom objects, which HubSpot offers on Enterprise and Salesforce on its platform; in Customermates you add your own list for it.

How much does a CRM database cost?

A spreadsheet or a self-hosted open-source CRM has no licence fee, but you pay in setup and maintenance time. Cloud CRMs mostly charge per user and month. In Customermates Cloud that is €12 per user on Starter and €29 on Pro with the unified inbox; see pricing.

Vendor details checked on 11 November 2026 against HubSpot's object documentation (last updated 2 October 2026), Salesforce Trailhead on standard and custom objects and Pipedrive's data organization article (last updated 23 September 2026).

Start from one description

One description. A workspace your agent built.

This is what setup builds from one sentence. Sample data, the real product.

  1. Describe your business

    One sentence about your team, your clients and how you sell.

  2. Mate builds it

    It asks a few questions, then creates the lists, pipeline, dashboard and playbook.

  3. Work in it

    Connect LinkedIn and email. Click through a finished example below.

Your description to Mate

We're a four-person strategy consultancy. Track clients, engagements and proposals, and help us follow up with prospects on LinkedIn.

Widgets built from the workspace's own lists and fields

Customermates for consultants

Every client conversation from email, LinkedIn and WhatsApp in one place. Mate sets up your workspace and drafts the follow-up, you review and send. Works with Claude, ChatGPT or any agent that speaks MCP.

Customermates

Follow up on every client conversation

Start with one channel and one client on Customermates Cloud, or self-host the open-source core.

✓ No credit card required ✓ First 14 days free