Run enough addresses through an SMTP check and the same shapes appear regardless of whose list it is. We have not run that volume ourselves yet, and say so on every page that touches accuracy; what follows is the mechanism, which is well established and does not depend on our own numbers. The provider mix explains most of the variation, and it explains it before anything about acquisition source does.
A list of consumer Gmail addresses resolves well, because Google rejects unknown recipients clearly when it lets you ask. A list of business domains resolves much worse, because catch-all configuration and security gateways sit in the way. Same verifier, same day, completely different coverage.
That is why a single accuracy figure is meaningless without knowing what was in the file. A vendor quoting one number is quoting an average over a list composition they chose, and yours is different.
The second pattern is that the hard providers are hard for structural reasons rather than technical ones. Yahoo's policy is a decision. A catch-all is an administrator's setting. Proton's refusal is a product feature. None of those yield to a better algorithm.
The patterns that hold
Domain age predicts decay better than address age does. A free webmail domain that peaked twenty years ago carries mostly abandoned accounts whatever the list says about when they signed up.
Role addresses cluster at small organizations, where they are frequently the correct contact rather than a problem to remove.
ISP addresses decay faster than webmail addresses, because the mailbox dies with the broadband contract. That failure is invisible until you send.
B2B lists decay fastest of all, because people change employers and the address goes with the job. A two-year-old B2B list is a different asset from a two-year-old consumer list.
Duplicates are more common than anyone expects, particularly on Gmail where dots and plus tags disguise one mailbox as several.
Where this argument costs us something
The short version
- Expect coverage to track your provider mix rather than your list source.
- Treat a B2B list as decaying faster than a consumer one.
- Normalize Gmail addresses before verifying, because the duplicate rate is higher than it looks.
Questions people ask
Which providers resolve best?
Ones that reject unknown recipients cleanly and do not throttle heavily: Fastmail, Zoho, Seznam, the Polish portals and Web.de among them. The largest providers are not the easiest.
Does list source matter at all?
It predicts consent and spam trap risk, which matter enormously. It predicts coverage much less well than the provider mix does.