How to clean up a messy HubSpot CRM

Most HubSpot cleanups start in the wrong place. Someone exports the contacts, finds four thousand duplicates, and starts merging. Two weeks later the duplicates are back and the export looks exactly the same.

The duplicates aren't the problem. They're a symptom. A messy CRM is rarely a data problem. Most of the time it's a model problem, the shape of the thing underneath the data. Fix the shape and the data mostly cleans itself. Merge records without fixing the shape and you're mopping the floor with the tap still running.

Here's the order I work in, on every messy HubSpot I've inherited.

1. Map how the mess gets in

Before touching a single record, find the doors. Where does data enter HubSpot? Forms, imports, a Salesforce sync, a Zap someone built in 2023 and forgot, reps typing free text into fields that should be dropdowns. Every door is a place the mess regenerates. List them. Most teams have more than they think, and half of them nobody owns.

Then read the data twice. First inside HubSpot, where the interface politely hides the worst of it. Then export the whole thing to a sheet and read it again, because the export shows what the screen smooths over: the same field filled three different ways, four lifecycle stages that mean the same thing, a property that is mostly empty. Write down every value that makes no sense to you. Not to fix yet. To understand.

Then talk to the people who live in it. Almost every discrepancy is someone working around a model that never fit how they sell. The rep typing into a free-text field did it because the dropdown never had their answer. The mess is a map of the gap between how the CRM was designed and how the work happens, and you cannot read that map from the admin seat. You have to ask.

2. Kill the fields nobody uses

Open your properties. You'll find three hundred. You use forty. The other two hundred and sixty are half-filled, contradictory, and wrong - and every one is another box a rep can put garbage in. Archive the dead ones. A field that isn't required and isn't reported on is a liability with a label.

3. Fix the doors, then fix the data

Turn free text into dropdowns. Add validation to forms. Standardise the sync so two systems stop arguing about the same contact. Then dedupe - because now the duplicates stop coming back. Do it in that order or don't bother; deduping first is scheduling the next cleanup.

4. Decide what "good" means, once

A lifecycle stage should mean the same thing to marketing and to sales. Usually it doesn't. "MQL" means "raised a hand" to one team and "has a pulse" to the other. Write the definitions down. Put them where people work, on the record, in the view, not in a doc nobody opens.

5. Govern it, or watch it rot

Clean data has a half-life. HubSpot's own data puts B2B decay at about 22.5% a year - a quarter of your database goes wrong in twelve months on its own. Without a weekly check, new duplicates, stale records, fields drifting out of format, you'll be back here in a quarter. The cleanup is the easy part. Keeping it clean is the actual job, and it's the part everyone skips. It's also most of what the ongoing work is.

The uncomfortable truth: most "messy HubSpot" was built on purpose, one reasonable-sounding request at a time. Someone needed a field. Someone needed a workflow. None of it was wrong on the day. It's wrong now.

Merging duplicates takes a weekend. Building a CRM your team actually trusts takes reading it twice, asking the people in it what they need, designing the shape around how they work, and rolling it out in stages so nothing breaks while you go. Do that and the CRM starts making the work easier instead of heavier, which is the only version anyone keeps clean. If your team works around the CRM in spreadsheets, the model's the problem, and the GTM diagnostic will tell you how deep it goes. Start there.

Common questions