How to Audit CRM Data Quality

Most data quality reviews stall at the first meeting, because they open with a debate about what quality means. The six measures below avoid that entirely: each requires no interpretation, no agreed target, and no judgement — only arithmetic.

Scoped correctly, the whole audit is a day's work and produces a baseline you can repeat.

Scope it before you start

The reason these audits fail is scope. Auditing every field in a mature CRM instance is weeks of work that produces a spreadsheet nobody acts on. Audit the ten to fifteen fields your executive reporting actually depends on — trace back from the reports leadership looks at and list the fields underneath them. That list is almost always shorter than people expect, and it is the entire surface that matters.

The six measures

Method: the one-day audit

1. List the reports leadership actually uses. Not every report — the ones that inform decisions.

2. Trace each to its underlying fields. Deduplicate the list. This is your scope, and it should be roughly ten to fifteen fields.

3. Run completeness per field, not in aggregate. An overall figure conceals which specific field is breaking routing or scoring.

4. Run staleness on open opportunities. Proportion with a close date in the past is the single most diagnostic measure available and needs no interpretation at all.

5. Run consistency on controlled fields. Count the distinct values present. Free-text creep in a picklist shows up here immediately.

6. Identify conflicts by asking, for each field, which system writes it. Any field where the answer is more than one system is a finding regardless of current disagreement rate.

7. Record every figure with the date and the method. This is the baseline, and without it no later work can be shown to have improved anything.

Reading the results

Each measure implies a different remediation, which is why six numbers beat one.

• High incompleteness on fields automation depends on explains routing and scoring failures without further investigation. Fix at the point of entry rather than by backfilling.

• High conflict indicates undeclared field ownership. Declare a writer per field; this is usually the highest-leverage finding.

• High staleness inflates pipeline coverage and makes every conversion metric downstream unreliable. Treat conversion figures as provisional until it is corrected.

• Low consistency in a controlled field means the control is not enforced in the system, only stated in a document.

• High duplicate rate matters relative to the questions you ask. A hundred duplicates in a thousand records is a broken database; the same hundred in a hundred thousand is noise.

Where this audit does not help

It measures the state of the data, not whether the data model is right. A CRM where every field is complete and consistent can still be modelling the business incorrectly — the wrong objects, the wrong relationships, stages that do not describe the process. That is a design review rather than a quality audit, and the two are frequently conflated.

It also says nothing about whether the fields should exist. A field that is 100% complete and read by nothing is a cost, not an asset, and it will not appear as a problem in any of the six measures.

Making it repeatable

Run it quarterly with an identical method. The absolute numbers matter far less than the direction, and direction requires that the measurement itself does not change. Where you must change the method, restate the prior period on the new basis rather than creating a discontinuity that nobody can interpret a year later.