Run a Baseline Data Audit

Nobody has run a full population scan at GTM Lab in over a year. Small issues, a few duplicates a week and a handful of unvalidated bulk imports, have compounded quietly into a population-wide problem. Someone finally proposes a scan, scoped to “check everything, every field.” It returns 4,200 findings in one run. Nobody can tell which ten matter today. The audit gets shelved a week later, exactly like the last attempt.

More findings isn't more insight

The failure isn't that the scan found too much. It's that everything it found looked equally urgent, which functions the same as nothing being urgent. A severity model turns a flood into a queue: high blocks a handoff, a send, or a report population if unresolved; medium degrades a report or automation but doesn't block anything; low is cosmetic or low-frequency, batched for periodic cleanup.

The decision: scope before you scan

Scope which objects and segments are in, what's excluded and why, what severity model sorts the findings, which six baseline metrics get tracked (duplicate rate, normalization failure rate, core-field population rate, stale-account count, consent/DNC coverage rate, enrichment spend rate), and what cadence re-runs the scan. Every metric needs its denominator stated, because a percentage with no population attached isn't a usable baseline entry.

Carry this into your business

A scan without a severity model is a report nobody reads twice. Scope it, sort it, and give every finding a place to land before you run it against a real population.