Preparing your data

Get your data ready before connecting: formatting rules, entity relationships, PII handling, and the readiness checklist.

Well-prepared data is the difference between a system that decides well and one that guesses. Clean identifiers raise match rates, complete fields keep good accounts in the ranking, and correct links between people and companies make every score sharper.

This page is about readiness. What tables you need is covered in Required tables; connecting a specific tool is covered in Your stack.

Formatting rules

The matching keys have to be clean, because signals only attach to clean identifiers.

Field	Good	Bad
Email	[e-mail] (lowercase, trimmed)	[e-mail]
Domain	acme.com	https://www.acme.com/
Phone	[phone] (E.164)	[phone]
Timestamp	ISO 8601 or Unix	01/15/2024 10:30 AM

Typical normalization, run in your warehouse before connecting:

UPDATE contacts SET email = LOWER(TRIM(email)) WHERE email IS NOT NULL;

UPDATE accounts SET domain = LOWER(
  REGEXP_REPLACE(REGEXP_REPLACE(domain, '^https?://(www\.)?', ''), '/.*$', '')
);

Uniqueness and duplicates

Primary keys must be unique: duplicate IDs cause scoring errors downstream.

SELECT account_id, COUNT(*) FROM accounts
GROUP BY account_id HAVING COUNT(*) > 1;
-- should return 0 rows

Dedupe examples that rely on ctid are Postgres-only. On Snowflake, BigQuery, Databricks, or Redshift, dedupe with a window function (ROW_NUMBER() OVER (PARTITION BY account_id ORDER BY updated_at DESC)) instead.

Link people to companies

Every contact should carry an account_id. Orphan contacts still score, but they cannot contribute to account-level decisions.

-- find orphans
SELECT COUNT(*) FROM contacts c
LEFT JOIN accounts a ON c.account_id = a.account_id
WHERE a.account_id IS NULL;

-- link by email domain (excluding free-mail)
UPDATE contacts c SET account_id = a.account_id
FROM accounts a
WHERE SPLIT_PART(c.email, '@', 2) = a.domain AND c.account_id IS NULL;

Handling PII

You choose how each sensitive field is handled at ingestion:

• Hash: one-way SHA-256, still usable for matching. The default for email, IP, and device IDs.

• Mask: partial visibility for operators (***-***-1234).

• Encrypt: reversible, for fields authorized users need to read back.

• Exclude: not ingested at all.

Hashing is consistent within your workspace, so identity resolution and ad-platform matching work without exposing the raw value. Full details live in Security model.

Readiness checklist

• Required fields populated (Required tables)

• Primary keys unique

• Emails lowercase and trimmed; domains cleaned

• Timestamps in a standard format

• Contacts linked to accounts

• updated_at columns present for incremental sync

• PII handling chosen per field

When the list is green, connect your sources (start with your CRM; see CRM sync), then watch Data health after you connect.

Outside the app

Readiness checks can be run outside the app via API and MCP as well.