This is the health of the data feeding every play: coverage sets the ceiling on who you can reach, accuracy sets what reps hit on the dial. The mistake is trusting a provider's self-reported accuracy, which tests far below the claim.
Shape of it
5 signals
Share of queried records for which a provider returns the requested field.
records returned with value / records queriedBenchmarkBest single provider email ~65%; a 4-tool waterfall reaches ~68% enriched (62% valid), a +23% lift over the best single tool. Ignore the folklore "92%"
Coverage is the ceiling on how much of a list you can contact.
Share of found emails that verify as deliverable.
BenchmarkTarget >85% of found emails valid; a 4-tool waterfall lands ~62% valid of all records queried
The pre-send gate that protects bounce rate and domain reputation.
Rate at which records go stale from job changes, pivots, M&A.
BenchmarkEmail databases decay ~22% per year (ZoomInfo puts B2B data decay near ~30%/yr overall); contact data decays faster than firmographic
Dictates re-enrichment cadence; stale data is worse than none because it looks trustworthy.
Share of records that duplicate an existing entity.
duplicates identified / total recordsBenchmarkRule of thumb ~10–30%; measure your own on domain/email keys
Duplicates corrupt routing, double-count pipeline, and inflate enrichment spend.
Share of returned values that are factually correct (distinct from match rate).
BenchmarkVendor self-cited accuracy (Apollo 95%, Lusha 98%) tests closer to ~50% in the wild; treat vendor numbers as ceilings
Accuracy, not coverage, is what a rep hits on a bad dial. Separate match ≠ accuracy ≠ deliverability.