CRM Hygiene

The play in seven steps

1. 01

   Step 1: Audit, read-only

   The first pass never writes. It measures how dirty the CRM really is, so every later fix has a number behind it.

   • Export a snapshot of HubSpot or Salesforce: contacts, companies, leads, owners and the fields that matter.
   • Measure fill rates on title, LinkedIn, country, email validity and segment.
   • Find which integration or import created each record. Duplicates regenerate unless the source is fixed.
   • Treat default picklist values such as pending as blank.
   • Prove nothing was written: count records last modified by your own user.
2. 02

   Step 2: Detect the dirty sets

   Messy CRMs break in the same handful of ways, and each one gets its own rule and its own fix.

   • Duplicates: two records for the same person or company.
   • Contacts at a company that already has an account, but are not linked to it.
   • Job changers: LinkedIn shows a different current employer.
   • Bounced or invalid emails.
   • Missing fields, and leads with no account.
   • Never size a duplicate problem from a tool's suggestion count. Re-derive it per matching rule.
3. 03

   Step 3: Match duplicates carefully

   Most records that look like duplicates are different people who share a name, a phone line or a company name.

   • A name match only counts when the company domain matches too.
   • Never use a shared phone or a company name as the key. Colleagues share a switchboard, and many firms share a name.
   • Blank directory and link-in-bio domains before any domain rule, or unrelated companies merge.
   • A role inbox with a person's name stamped on it is a field to fix, not a merge.
   • Pick the survivor by rule: furthest through the funnel, then has an email, then most recent.
4. 04

   Step 4: Approve and back up before anything writes

   Nothing touches the CRM without a human yes and a way back.

   • Take a backup export of every record in the change set.
   • The CRM owner approves the exact list, as a saved list view or report.
   • Check workflows, validation rules and two-way syncs to other tools that will fire on the write.
   • Estimate enrichment credits before the run.
   • Log every change with the previous value before it is written, so any field can be restored.
5. 05

   Step 5: Fill the blanks with a waterfall

   Enrichment fills what is missing and leaves alone what someone put there on purpose.

   • Clay runs the providers in order: AI Ark for title, LinkedIn and country, then Prospeo and Findymail for email.
   • Only the misses move to the next provider.
   • BounceBan verifies every email before it lands.
   • Fill blanks only. The two exceptions: bad emails get replaced, and job changers get a new contact.
6. 06

   Step 6: Catch the job changers

   A contact who left is still worth something: the history stays, and the person becomes a lead at the new company.

   • Scan current employers for the records that matter, customers first, through Clay and AI Ark.
   • Check the new company is not the same company under another domain or another account record before moving anyone.
   • Create a new contact at the new company and mark the old one as gone. Do not overwrite the old record.
7. 07

   Step 7: Keep it on

   A cleanup decays the day it ends. A pipeline that every new record runs through does not.

   • New inbound leads, records reps add by hand and job-change signals all run through the same rules.
   • A weekly sweep re-checks fill rates and bounces.
   • Post the weekly numbers and anything held for review to Slack.