Audit your own records

CRM data does not degrade dramatically. It degrades one blank field at a time, by people in a hurry, and the cost only appears when someone tries to do something with it: a segment that misses half its audience, an outreach list with three versions of the same company.

Decide what "complete" means for you

Not every field matters. A CRM audit that flags every empty box produces a report of four thousand problems and gets ignored.

Pick the handful of fields that actually block work. Usually: an owner, an email, an industry or segment, and a last-activity date. Those are the ones whose absence stops something.

Ask for what is broken, grouped by why

Every Monday, audit our CRM records.

Group what you find:
  MISSING     records without an owner, email, or segment
  STALE       open deals or active contacts with no activity in 90 days
  DUPLICATE   records that look like the same company or person
  IMPOSSIBLE  values that cannot be right, like a close date in the past
                on an open deal

For duplicates, show what makes you think they are the same and do not
merge anything.

Give counts per group and list at most ten examples each. If a group is
empty, leave it out.

Never merging automatically is the boundary worth keeping permanently. Two records that look identical may be two divisions of the same customer, a parent and a subsidiary, or two people who share a name. A merge is irreversible in practice, and the ten seconds a human spends confirming is cheap against losing the history of an account.

Fix the class, not the instances

The first audit gives you a list. The useful question is why each class exists.

Records with no owner usually come from one import or one form. Stale open deals usually mean a stage nobody closes. Fixing the cause removes the category rather than the instances, and the next audit is shorter for a reason.

Let the counts be the report

After the first month, stop reading the examples every week and read the four counts. Their direction is the signal: a count that stops falling means the cause is still live.

What good looks like

Four numbers on a Monday, falling for a month or two and then flat and small.

The measure is not a clean CRM, which no working CRM ever is. It is that somebody building a segment or a list can trust it without checking first.