- Automating on top of bad data makes errors faster, not rarer.
- Start by deleting fields nobody uses. It makes every later step easier.
- Agree merge rules in writing before you touch duplicates.
- Prevent the mess from returning with required fields and checks on entry.
An automated workflow does exactly what the data tells it to do. If a contact exists three times, they receive three emails. If the lifecycle stage is wrong, the wrong sequence starts. Cleaning the CRM first is the least exciting part of an automation project and the part that decides whether the rest works.
Work through these seven checks in order. Each one makes the next easier.
1. List every field and who uses it
Export the full list of properties for contacts, companies and deals. For each one, note how many records have a value and which team relies on it. Most CRMs that have been in use for a few years contain dozens of fields created for a single campaign and never filled in again.
2. Remove or archive what nobody uses
Fields with almost no data, and no owner willing to defend them, should go. Archive instead of deleting if you are unsure. Fewer fields means shorter forms, cleaner imports and far less mapping work when you connect other systems.
3. Standardise the values that drive logic
Automation rules read fields such as country, industry, lead source and lifecycle stage. Free-text versions of these are where rules silently fail, because “UK”, “U.K.” and “United Kingdom” are three different values to a filter. Convert them to fixed drop-down options and map the old values across.
4. Agree merge rules, then de-duplicate
Before merging anything, write down the rules. Which record survives? Which email address wins when two differ? What happens to notes and deal history? With the rules agreed, merge the obvious matches in bulk and review the uncertain ones by hand. Keep a log of what was merged so that mistakes can be traced.
5. Fix ownership
Every active contact and open deal needs a current owner. Records owned by people who have left, or by nobody, are where follow-up disappears. Reassign them by territory or account, and decide what should happen automatically the next time someone leaves.
6. Check the pipeline matches reality
Walk through your deal stages with the people who use them. If reps routinely skip a stage, or park deals in one for months, the stage definitions are wrong. Rewrite each stage as a fact that can be verified, such as “proposal sent”, and close out deals that have been dead for a long time.
7. Stop the mess coming back
A clean-up without prevention lasts about a quarter. Put the guard rails in at the point of entry:
- Required fields on forms and at deal creation
- Duplicate checks when a record is created or imported
- A named owner for the data model, who approves new fields
- A short monthly report of records that are missing key values
How long should this take?
It depends on the number of records and how many systems feed the CRM, so a field audit is the honest first step. It tells you the size of the job before you commit to it. Once the data is trustworthy, automation on top of it becomes simple, including the lead routing described in our article on speed to lead.