Nothing in a CRM breaks when records go stale. The contacts are still there and the dashboards still count them, but a growing share of the emails are dead and a growing share of the job titles describe a role the person left a year ago.
This guide is the cleanup itself, in the order that avoids the classic mistake of a cleanup: overwriting good data with worse data. It works the same way in HubSpot, Salesforce, Pipedrive, Zoho, Airtable or a plain spreadsheet, because every step runs on an export.
Before you start: back up and pick a scope
Two things to do before you touch a single record.
- Take a full export and keep it untouched. Name it with the date, store it somewhere your import tool cannot reach. If an import goes wrong, this file is how you put things back.
- Pick the scope. Cleaning the whole database at once is how cleanups stall. Start with the segment that costs you money when it is wrong: contacts in active sequences, open opportunities, or the list behind your next campaign. A few thousand rows is a good first batch.
Also note who owns the CRM's import settings. Most of the damage in a cleanup comes from one import with the wrong "overwrite" option, so whoever runs step 8 should be the person who can undo it.
Export the fields you will actually use
Export more than the contact details. The activity fields are what tell you a record is stale; the IDs are what let you write the result back to the right record.
- Record ID (HubSpot "Record ID", Salesforce "Contact ID"). Without it, a re-import creates duplicates instead of updates.
- Identity: first name, last name, email, phone, LinkedIn profile URL.
- Company: company name, company domain or website.
- Role: job title, seniority or department if you store them.
- Activity: create date, last modified date, last activity or last contacted date, last email open or reply.
- Status: email bounce status, unsubscribed / opted out, lifecycle stage, owner.
Export as CSV with UTF-8 encoding, so accents in names survive the round trip.
Find the outdated records
Add a column called stale_reason and fill it with the first rule each row matches. In a spreadsheet each rule is one filter; in the CRM each is one saved view.
| Rule | How to spot it | What it usually means |
|---|---|---|
| Hard bounce | Bounce status is "hard" or "invalid" | The mailbox is gone, most often because the person left |
| No activity | Last activity and last modified both older than 180 days | Nobody has checked this record since it was created |
| Title never updated | Job title unchanged and record older than 12 months | A promotion or a move you have not seen |
| Missing key field | No email, or no company domain, or no title | Incomplete from day one: routing and scoring fail silently |
| Personal email on a B2B record | Domain is gmail.com, outlook.com, yahoo.com and so on | Fine for some teams, a deliverability and compliance risk for cold outreach |
| Role address | Local part is info, sales, contact, hello, admin | A shared inbox, not a person |
Leave alone any record that has a reply or a meeting in the last 90 days, whatever its age. Recent engagement beats every rule above.
Count the rows per reason before going further. That count is your before picture, and it tells you which of the next steps will carry the most weight.
Map every field before you change one
A field map is a small table that says, for each column, where it goes back in the CRM and whether new data may replace old data. Writing it takes ten minutes. Skipping it is how a cleanup wipes out the notes your reps typed by hand.
| Column in your file | CRM property (internal name) | Update rule |
|---|---|---|
| Record ID | hs_object_id / Id | Match key. Never edit. |
email | Replace only when the old one hard-bounced and the new one verifies | |
| Job title | jobtitle / Title | Overwrite: a fresh source beats an old entry |
| Company / domain | company, website | Overwrite when the person changed employer, and log the old value |
| Phone | phone, mobilephone | Fill if empty |
| LinkedIn URL | Custom property | Fill if empty |
| Lead source, owner, notes | Various | Never touch |
Use the internal names, not the labels: two properties can share the label "Phone", and the import screen matches on internal names. In HubSpot they are under Settings → Properties; in Salesforce, Object Manager → Fields & Relationships.
Add two properties if you do not have them yet: previous_company, to keep the old employer when someone moves, and last_cleaned_at, the date this record was last checked. Step 9 runs on that second one.
Normalise the formats
Duplicates and failed validations mostly come from format, not from wrong data. Fix the format first so steps 5 and 6 compare like with like.
- Emails: lowercase, trim spaces, remove trailing dots and "mailto:".
- Names: trim, fix ALL CAPS to normal case, split a full-name column into first and last.
- Company domains: strip
https://,www.and anything after the first slash, sohttps://www.acme.com/aboutbecomesacme.com. - Phones: convert to international E.164 format (
+14155550123), so the same number written three ways becomes one value. - LinkedIn URLs: keep
linkedin.com/in/handleonly, without tracking parameters or a trailing slash. - Countries: one convention, either full names or ISO codes, not both.
Merge duplicates
Look for duplicates in three passes, from the most certain key to the least:
- Same email. Almost always the same person.
- Same LinkedIn URL with different emails: the same person with an old and a new address.
- Same first name, last name and company domain. Check these by eye before merging: two Michael Smiths can work at the same company.
When you merge, keep the oldest record (it holds the history: deals, emails, meetings) and bring the newest values into it. Do the merge inside the CRM rather than by deleting rows in the file, so activity history follows the surviving record.
Validate emails and phone numbers
Run every email through a verifier and keep its verdict in a column, not just a pass/fail. The categories matter:
| Result | What it means | What to do |
|---|---|---|
| Valid | The mail server accepted the address | Keep |
| Invalid | Bad syntax, no mail server, or mailbox refused | Mark for refresh in step 7; stop sending now |
| Catch-all / unknown | The domain accepts every address, so nothing can be confirmed | Keep, but send at lower volume and watch bounces |
| Disposable | Temporary inbox service | Remove from outreach |
| Role | info@, sales@ and similar | Replace with a named contact if you can |
For phones, check that the number parses as a valid number for its country and tag it mobile or landline. A landline switchboard stored in a "mobile" field is the most common phone error in B2B CRMs.
Validation tells you an address works. It does not tell you the person is still there: an old colleague's mailbox can stay open for months after they leave. That is what the next step is for.
Refresh what changed
Take the rows flagged in step 2 and the invalid ones from step 6, and check where each person works today. The LinkedIn profile is the most current public source: compare its current employer with the company in your CRM.
- Same employer: update the title if it changed, then find the email again if the old one failed.
- New employer: move the old company to
previous_company, write the new company, title and work email. Then tag the record: someone who knew your product and just arrived in a new job is one of the warmest contacts you have. - Cannot confirm: leave the record unchanged and tag it unverified. Do not guess an email from a pattern.
For a handful of records, do it by hand. For hundreds, an enrichment tool does the lookup per row: LinkFinder AI, for example, takes the CSV and returns current employer, title and a verified email per row. Whatever tool you use, bring the results back as new columns and apply them with the rules from step 3.
Re-import without breaking anything
- Import 50 rows first. Open five of them in the CRM and check every mapped field landed where step 3 said it would.
- Match on Record ID, never on email: you just changed some emails, so matching on email would create new contacts.
- Leave out the columns marked never-touch. Remove them from the file rather than relying on an import option.
- Set
last_cleaned_atto today on every row you checked, including the ones that did not change. - Run the full import, then compare the counts per
stale_reasonwith your before picture.
Records that ended up unverified and inactive go to an archive list or a suppression list, depending on your CRM. Removing them from outreach protects your sending domain; deleting them is a separate decision, guided by your data retention policy.
Keep it clean
A one-off cleanup starts decaying the day it ends. Three habits stop the same project coming back in eighteen months:
- A 90-day sweep: a saved view of contacts whose
last_cleaned_atis older than 90 days, run through steps 4 to 8 each quarter. It is a small batch every time. - Check before every campaign: verify the segment you are about to email, not the whole database.
- Complete records at creation: required fields on forms and imports, so nothing enters the CRM without an email, a company domain and a title.
If you want the sweep to run on its own, see how to set up a scheduled CRM enrichment pipeline in n8n, Make or Zapier.
The checklist
- Full backup export saved and untouched
- Scope chosen: one segment, a few thousand rows
- Record ID, activity dates and bounce status exported
- Every row tagged with a stale reason, counts written down
- Field map written: overwrite, fill if empty, never touch
- Emails, domains, phones and LinkedIn URLs normalised
- Duplicates merged in the CRM, oldest record kept
- Emails verified, result kept per row; phones parsed and typed
- Job changes checked, previous company kept
- 50-row test import checked by eye, full import matched on Record ID
last_cleaned_atset, 90-day sweep scheduled