What is a catch-all email address?

Companies switch catch-all routing on so a misspelled name or a former employee's address still lands in a shared inbox instead of bouncing. For the company, that's a safety net. For anyone checking a list, it hides the one answer that matters.

Here's why. Ask a normal server about a missing mailbox with RCPT TO and it answers 550, no such user. Ask a catch-all server and it answers 250 to every name you try, because it has agreed to take everything and sort it out later.

The test for it is a control probe. You ask the same server about a random address nobody could have registered, something like x7q2k9 at the same domain. If that is accepted too, the domain takes anything, and a 250 for the real address proves nothing about the person.

Catch-all routing is common on business domains. In one 10,000-address B2B benchmark, about 28% of the addresses sat on catch-all domains. How a verifier labels that share moves its headline accuracy figure more than the quality of its engine does.

Yahoo and AOL act like catch-alls on purpose. They return 250 or 252 to every recipient to stop address harvesting, then bounce the message later, so Yahoo-hosted domains almost always come back catch-all.

A catch-all verdict means unproven, not bad. Some of those addresses belong to real people and some are typos or departed staff. Your own engagement data and a small test send sort one from the other.

Suppose you run a 10,000-address B2B list through a verifier and 2,800 of those addresses sit on catch-all domains, which is close to the usual share. Of the remaining 7,200, say 6,300 return a clear 250 after the control probe was refused, and 900 return a clear 550.

An honest report says coverage was 72%, because that's the share of the list that got a real verdict. Accuracy is then measured inside that 72%. A vendor who quotes you 99% is nearly always describing this inner group and staying quiet about the 2,800. With us those 2,800 cost a credit each like any other result, and the label is what you're buying: proof that the domain accepts everything, so a valid on those rows from anyone else is a guess.

Now watch what happens if those 2,800 get stamped valid instead. Your report shows 9,100 good addresses, the summary looks wonderful, and the first campaign decides how many were real. If a fifth of them are dead, that's 560 hard bounces from addresses you were told were fine.

Keep the catch-all segment out of your main send. Start with a batch of a few hundred, from the same domain and with the same content, and read the bounces over the next three days. Some accept-all servers reject unknown users only after the handshake, so the failure arrives later as a bounce message instead of a refusal on the wire.

If that first batch bounces under 2%, widen it. At 10%, you've learned something about where the data came from, and it cost you a few dozen bounces to find out.

Age matters more here than anywhere else on your list. A catch-all address you collected last month from a signed-up customer is probably fine. One that came from a purchased file three years ago has nothing vouching for it at all, because the single check that could have caught it doesn't work on that domain.