From data swamp to clean CRM: a 30-day reset

Before you migrate, before you automate, and definitely before you point AI at anything: clean the data. This is a 30-day plan that a small team can run alongside their day jobs. It won't make your CRM perfect. It will make it trustworthy enough to build on.

Week 1 — Measure before you touch anything

Resist the urge to start deleting. Spend the first week establishing what 'bad' actually means in your system.

• Count duplicate accounts and contacts using a defined match rule (domain plus normalized name works for most B2B).
• Measure fill rate on the ten fields your reports depend on.
• Count records with no activity in 24 months.
• Count opportunities with a close date in the past and an open stage.
• List every field on your main objects and how many records populate each one.

Publish these numbers. They are your baseline and, later, your evidence that the effort was worth it.

Week 2 — Decide the rules

Cleanup without agreed rules just moves mess around. Get the business owners in a room and write down definitions, in plain language, for each of the following:

1. 1What makes two records duplicates, and which one survives a merge.
2. 2What makes an account active, dormant, or archived.
3. 3Which fields are genuinely required, at which stage — not which fields someone would like filled in.
4. 4Who owns a record when the original owner leaves.
5. 5What gets archived rather than deleted, and where it goes.

One page, signed off. This document outlives the cleanup and becomes your governance baseline.

Week 3 — Execute in a safe order

Order matters. Do it in this sequence and you avoid re-doing work:

1. 1Take a full backup. Non-negotiable.
2. 2Standardize formats first — country names, phone formats, industry values, job titles.
3. 3Then deduplicate, since standardized data matches far more accurately.
4. 4Then reassign ownership of orphaned records.
5. 5Then archive dormant records out of the working view.
6. 6Finally, delete the fields nobody populates. Fewer fields, higher quality.

Run each step on a small batch first and have a business owner eyeball the result before you run it at volume.

Week 4 — Make it stick

A cleanup with no governance is a project you will repeat annually. Spend the final week on prevention:

• Turn on duplicate prevention at the point of entry.
• Enforce the required-field rules you agreed in week two — and only those.
• Add validation on the formats you just standardized.
• Build a one-screen data-health dashboard showing the week-one metrics, refreshed weekly.
• Name a data owner per object and put a 15-minute monthly review in the calendar.

What good looks like after 30 days

Duplicates under two percent on your core objects, fill rate above ninety percent on report-critical fields, no open opportunities with past close dates, and a dashboard that tells you within a week if things start slipping again. That's a foundation you can migrate, automate, or apply AI to with confidence.

Clean data is not a project outcome. It's a weekly fifteen-minute habit with a dashboard attached.