When reports disagree, teams often ask for a new tool. What they usually need is fewer people editing the same field, a written definition of “complete,” and a weekly cleanup that actually happens. Dirty data is rarely a technology problem first — it is a governance problem that technology makes visible.
How to diagnose a data hygiene problem
Before blaming the ERP or CRM, check three things:
- Who can edit master records? If the answer is “everyone,” duplicates and conflicts are inevitable.
- What does “complete” mean? If nobody wrote it down, every person fills fields differently.
- Is cleanup scheduled or heroic? If fixes only happen before board meetings, the baseline never improves.
Pull a sample of 50 records and trace who created them, who last edited them, and which fields are blank or inconsistent. The pattern usually points to access, not software.
Least privilege, written down
Not everyone needs edit rights on master customers, chart of accounts, or price lists. View-only for most. Edit for owners. Admin for a short list. Review that list quarterly — access drifts quietly.
Document the permission model in plain language: role name, what they can view, what they can edit, what they can create, and who approves exceptions. When someone asks for broader access, the answer should reference a rule — not a favour.
SOPs for the dirty work
Duplicates, missing GST fields, freestyle city names — turn the top five recurring messes into one-page SOPs. Who fixes them, how, and by when. That’s how we drove measurable accuracy gains without a platform rewrite.
Example SOP for duplicate customers: how to identify (matching phone, PAN, or name fuzzy match), who merges (records owner only), what fields survive the merge, and how to notify the salesperson whose record was absorbed. Without this, merges create new errors.
Validate at the door
Required fields and simple checks at entry beat heroic cleanup later. If CRM allows blank industry or ERP allows free-text vendor names, your dashboards will keep lying.
Practical validations that pay off quickly:
- GST format check on vendor and customer creation
- Dropdown for city and state instead of free text
- Duplicate warning on phone or email before save
- Mandatory approval before a record moves from draft to active
Start with the fields that break reports most often. Perfection is not the goal — fewer surprises is.
Publish an exception list
A short weekly list of broken records with owners beats a huge data-quality project. Clear ten items a week and the system starts feeling trustworthy again.
Share the list in a standing ops meeting: record ID, what’s wrong, who owns the fix, due date. Celebrate closed items. When the same error type repeats, fix the validation or SOP — not just the record.
What good data hygiene looks like
Two finance reports match without reconciliation debates. New hires can find a customer record in under a minute. Dashboards refresh without someone manually fixing source data first. Auditors ask for access logs and you have them. That state is achievable without a data warehouse — it starts with permissions, SOPs, and a weekly rhythm.
Start where the pain is loudest
Do not launch a company-wide data quality initiative. Pick the one report that leadership questions every Monday and trace its source records. Fix access and validation there first. A visible win in one domain builds credibility for the next. Repeat until the weekly exception list stays short without heroics.
Takeaway: Clean data comes from controlled access, clear definitions, entry validation, and a weekly rhythm — not from a one-time cleanup project.