A practical guide to CRM data cleanup

Signs your CRM needs cleaning

You probably already know. Sales reports do not match what the team says in meetings. The same client appears three times with different spellings. Deals from two years ago sit in 'proposal sent'. Half the contacts have no phone number, and nobody knows which email address is current. The owner keeps a private spreadsheet because the CRM cannot be trusted.

These problems compound. Automations built on bad data send the wrong messages to the wrong people. Dashboards built on stale deals overstate the pipeline. Before adding any AI or automation on top of a CRM, it is worth making the underlying records reliable.

Decide what good data looks like

Start with a one-page data standard. Which fields does the business actually use to make decisions? For most service firms, the list is short: company, contact name, email, phone, lead source, service type, owner, deal stage, deal value and next step date. Everything else is optional.

For each of those fields, write the allowed values. Lead source should be a dropdown with eight options, not a free-text box with 200 variations. Service type should match how you actually price and deliver work. Agree what each pipeline stage means in plain words, including what has to be true to move a deal forward.

Keep the list of required fields short. Every required field is something a busy person must fill in before they can save a record. Ask for too much and staff will invent values to get past the form, which is worse than a blank. Five to ten required fields is usually plenty for a service firm.

Audit and back up

Export everything to a spreadsheet before you touch it. Keep that export somewhere safe. Bulk merges and deletes are hard to undo, and you will want the original if something goes wrong.

Then measure the mess. Count contacts and companies, duplicates by email and by name, deals with no activity in six or twelve months, and the share of records missing each key field. These numbers are your baseline and help you prioritize. If 40% of deals have no owner, fix that before worrying about job titles.

While auditing, also look at who has access. Former employees with active CRM logins, shared logins and admins who no longer need admin rights are common findings. Fixing them is part of good data hygiene and reduces the risk of client data leaving with someone who has left the business.

• Duplicate contacts and companies
• Deals with no activity in 6+ months
• Records missing owner, lead source or stage
• Free-text fields that should be dropdowns
• Contacts who have unsubscribed or bounced

Fix in the right order

Work from the most damaging problems to the least. First, merge duplicates. HubSpot, Salesforce, Zoho CRM and Pipedrive all include duplicate management tools, and third-party tools exist for larger cleanups. Review suggested merges before accepting them in bulk; two contacts with the same name may be different people.

Second, deal with stale opportunities. Ask each owner to close as won, close as lost, or update the next step on anything older than a set age. Anything nobody claims gets closed as lost with a reason such as 'no response'. Third, standardize picklists by mapping old free-text values to the new allowed list. Fourth, fill gaps in key fields where you can, starting with active clients and open deals rather than the whole database.

Decide what to do with contacts who have bounced, unsubscribed or asked not to be contacted. They should stay marked so they are never emailed again, which means suppressing them rather than deleting them, since deleting can allow the same address to be re-imported later from an old spreadsheet. Document the rule so whoever imports data next follows it.

Using AI and automation in the cleanup

AI tools can help with the tedious parts: suggesting likely duplicates, normalizing company names, classifying free-text lead sources into your new categories or summarizing old notes into a single next step. Many CRMs now include AI features for this, and general tools like ChatGPT or Copilot can work on an exported spreadsheet.

Keep a person reviewing changes before they go back into the CRM, especially merges. Be careful about pasting client data into consumer AI tools; use business accounts with appropriate data terms, and in regulated industries check your obligations first.

Who does the work, and how to run the project

CRM cleanup fails when it becomes one person's side project. Treat it as a short, scheduled project with a clear end date, typically two to four weeks. Name an owner, usually an operations lead or office manager, who runs the exports, applies bulk fixes and tracks progress. Each salesperson, estimator or account lead is responsible for reviewing their own open deals and contacts, because only they know which are real.

Block time for that review. Asking busy people to clean their records 'when they get a chance' does not work. A 45-minute working session per person, with the owner on hand to make changes, moves much faster. Start with active clients and open opportunities, since those matter most to revenue, and leave the long tail of old contacts for automated rules.

Communicate the why. People are more willing to clean data when they see what it enables: accurate forecasts, fewer awkward duplicate emails to clients, and automations that actually help them rather than create noise. Share the before-and-after numbers when the project ends, so the team sees the result of their effort.

• Week 1: data standard agreed, export and baseline taken
• Week 2: duplicates merged and picklists standardized
• Week 3: owner review sessions on open deals and active clients
• Week 4: rules switched on, baseline re-measured, results shared

Keep it clean

Cleanup without rules lasts about three months. Make key fields required at the point of entry. Use dropdowns instead of free text. Set an automatic task or alert for deals with no activity in 30 days. Connect web forms and booking tools directly to the CRM so records are created consistently rather than typed by hand.

Give one person ownership of data quality, with a 15-minute monthly check of the same baseline numbers. InTheory typically treats CRM cleanup as a precondition for any automation project, because reliable records make everything built on top of them easier to trust.