How to set up the Attio data model without making a mess

People records hold person-level data: name, email, phone, role, seniority. One human, one Person record.

Companies records hold organization-level data: domain, industry, headcount, ARR. One organization, one Company record — even if 12 of your People work there.

Deals records hold opportunity-level data: value, stage, owner, expected close date. Multiple Deals can exist per Company (renewal + expansion).

Workspaces / Users are the standard objects for product-led B2B SaaS — tracking the product accounts and end users tied to a Company.

Rule: never put person-level data on a Company record, and never put company-level data on a Person. 'Industry' lives on Company. 'Role' lives on Person. Mixing them up makes segmentation impossible later.

Workspace settings → Objects → pick object → Attributes. Read through every default. Filter by group.

Common attributes owners re-create unnecessarily: Job title (on People), Industry (on Companies), Estimated ARR, Headcount, Lifecycle stage, Lead source.

When you must create custom: click "Add attribute" → choose attribute type carefully. Attribute type is hard to change later without losing data (Text → Select is fine; Number → Select is not).

Group custom attributes into clear sections — Sales, Customer Success, Billing, Marketing Attribution. Never let custom attributes pile up in "Other."

Write a description on every custom attribute. Future-you in 18 months will not remember why `lead_score_v3` exists or what `account_health_signal` actually measures.

Text vs Rich text: Text for short single-line values (job title, lead source). Rich text for long-form notes (call summary, account brief). Do not use Text for notes — you will hit length and formatting limits.

Select (single) vs Multi-select: Select for mutually exclusive values (Lifecycle stage). Multi-select for tags (Use cases, Product modules).

Status: a special Select with workflow semantics — use for Deal stage, Lead status, Ticket status. Status powers Kanban views automatically.

Number vs Currency: Currency carries unit metadata (USD, EUR) and respects workspace primary currency. Number is just a number. Use Currency for any money field, always.

Date vs Date + time: Date for things like Close date. Date + time for Created at, Last contacted at, Demo scheduled at. Mixing the two breaks workflow timing.

Reference (Relationship): the killer feature — link records across objects. People reference Companies, Deals reference Companies + People, custom objects reference whatever you want. Set the cardinality (one-to-one vs one-to-many) consciously.

Default relationships: Person ↔ Company (auto-linked by email domain match), Deal ↔ Company, Deal ↔ Person (associated contacts).

For richer relationships, use labeled reference attributes. Add a "Champion" reference attribute on Deal pointing to People. Add a "Renewal owner" reference attribute on Company pointing to Users.

Avoid putting relationship data in free-text attributes ("Sarah is the champion" written in a notes field) — Attio cannot filter, report, or automate on text. Always use a typed reference attribute for relationships you will query.

Document the relationship map. A simple diagram (People — Companies — Deals — Workspaces, with labeled relationships) saves dozens of hours of confusion later.

Workspace settings → Objects → "Create object" (Plus / Pro tier).

Good custom object candidates: Subscriptions, Contracts, Invoices, Properties (real estate), Vehicles (automotive), Loans (fintech), Events (community), Jobs (recruiting).

Bad custom object candidates: "Sales notes" (use Rich text on Deal), "Activity logs" (use Activities), "Touchpoints" (use Interactions, built in), "Lead score components" (use Number attributes on Person).

For every "I need a custom object" idea, ask: is this a noun that has its own attributes AND relationships AND lifecycle? If only one of those, it is probably an attribute or a list, not an object.

Once created, custom objects participate in the data model exactly like standard objects — they have attributes, relationships, views, lists, workflows, and API endpoints.

Workspace settings → Objects → People → Matching. Confirm primary email is the dedup key.

For People: Attio auto-merges on email match during import. If a Person has no email, it imports as new every time — this is the #1 source of People duplicates.

For Companies: Attio auto-merges on root domain match. Add common alias domains (acme.com, acme.io) to the Company record manually to chain them together.

Before importing a CSV, deduplicate within the CSV first using Excel / Google Sheets. Attio will not catch within-file duplicates.

After import, use the Duplicates view (Companies → Duplicates tab) to find and merge any straggler records. Run this monthly for the first quarter, then quarterly.

Workspace settings → Imports → Start import. Pick the object you want to test (start with Companies).

Upload a 20-row sample CSV. Map every column to an attribute. Pay special attention to: Date formats (Attio expects ISO 8601 or M/D/YYYY), Currency fields (must specify the currency), Reference fields (must match an existing record by email or domain).

Run the test import. Open 5 random records. Are attributes populated correctly? Are relationships auto-linked? Did dedup work as expected?

If anything is wrong, fix the CSV or attribute, then re-test on the same 20 rows. Repeat until clean.

Then run the full import. Always import Companies first, then People (with Company domain column), then Deals (with Company domain + Person email columns). Otherwise relationships break.