CRM data governance: data-entry rules your team will actually follow

Every CRM cleanup ends the same way: the data looks great for a few weeks, then slowly drifts back. The cleanup fixed the records, not the habits that created them. Data governance is the set of rules that keeps new data as clean as the data you just fixed, and the best rules are the ones the CRM enforces for you.

Start with ownership

Governance fails when everyone is responsible for the data, because then nobody is.

• One person owns the data model. Usually RevOps or a CRM manager. They decide which properties exist, what each one means and who may create new ones.
• Every record has an owner. A contact, company or deal without an owner is a record nobody will update. Assign owners by rule, through routing or workflows, rather than by memory.
• Each property has a source. For each important field, decide where its value comes from: a form, the sales rep, an integration, an enrichment. Two sources writing the same field is how values flip back and forth.

Write a short data dictionary

A data dictionary doesn't need to be a hundred-page document. One table is enough: property, meaning, allowed values, who fills it, when it becomes required. Keep it next to the CRM, not in a forgotten folder, and update it whenever a property is added.

The most useful entries are the ones people argue about: what counts as a lead versus an MQL, when a deal's close date must be set, what "industry" means for a holding company.

Let HubSpot enforce the rules

Rules written in a document are suggestions. Rules set in the CRM are enforced. HubSpot offers several of them:

• Validation rules on text, number, date and phone properties: no leading or trailing spaces, minimum or maximum length, numbers only, no special characters, dates in a range. Regex rules require a Professional or Enterprise subscription. HubSpot applies them when values are edited or imported, but not in workflows, chatflows, meeting scheduling pages or forms built with the legacy editor, and it doesn't reformat existing values (HubSpot: set validation rules for properties).
• Required properties per pipeline stage. A stage can make a property required, so a record can't be created or updated into that stage until it has a value (HubSpot: set up and customize pipelines).
• Dropdowns instead of free text wherever the list of values is known: country, industry, lifecycle stage, lead status. Free text is where "France", "FR" and "Fr." start living side by side.
• Automatic company association. HubSpot can create companies and associate contacts based on the email domain, and lets you exclude up to 1,000 domains (HubSpot: automatically create and associate companies with contacts). Decide deliberately whether you want it, because it also creates one company per subdomain.

Control the doors data comes in through

Sales reps typing in a hurry are rarely the whole story. Bad data also arrives in bulk, through forms, imports and integrations.

• Forms should ask for the fields your segmentation depends on, with dropdowns where possible.
• Imports should go through one person, with a template that maps to existing properties and includes identifiers: email for contacts, domain for companies. Without them, HubSpot creates duplicates instead of updating records (HubSpot: set up your import file).
• Integrations should write to the properties they own, and only those. Document each one in the data dictionary.

Measure, every week

Governance without measurement is a matter of faith. Pick a handful of indicators and follow them:

• share of contacts and companies without an owner;
• share of companies without a domain;
• open deals without a close date, an amount or a next activity;
• number of duplicate groups.

A single score per object makes the trend obvious. Forgeo runs 40 checks after each sync and turns them into a data quality score for contacts, companies and deals: the share of records with no high or medium issue. When the score drops, you know which door let the bad data in.