Sender reputation is the asset an outbound programme actually runs on, and bounces are what spend it. A few percent of undeliverable addresses is enough to move a domain from the inbox to the spam folder — for every message it sends, not just the campaign.
Teams discover this in the wrong order. The campaign underperforms, someone raises volume to compensate, and three weeks later a colleague mentions that a customer said an invoice went to junk.
From the standard
“DMARC does not grant privileged delivery status to authenticated messages.”
The habit that causes it
Pattern guessing. A tool knows a name and a domain, infers first.last@company.com, and sends hopefully. Sometimes it is right. When it is wrong the message hard-bounces, and every hard bounce is recorded against the sending domain.
The practice persists because it is free and the cost is deferred and diffuse. Nobody attributes the invoice in the spam folder to the prospecting sequence from six weeks ago.
Three tiers, because they answer different questions
Tier one — syntax and disposability
Structural validation, plus rejection of known disposable domains, plus correction of obvious typos in common providers. It costs nothing and eliminates a surprising share of what a scraper produces. It cannot tell you whether a well-formed address exists.
Tier two — the mail-server probe
Ask the receiving mail server whether the mailbox exists, without sending anything. This is where most genuine verification happens, and it returns a clean verdict for the majority of addresses.
For a meaningful minority it returns neither yes nor no, and that third answer is where naive verification quietly fails.
Tier three — catch-all scoring
A catch-all domain accepts mail for any name at it, real or invented. The server says yes to everything, so the probe proves nothing. Large enterprises are disproportionately configured this way, which means the accounts most worth reaching are the ones verification struggles with.
Treating catch-all as valid reintroduces the bounce problem. Treating it as invalid discards your best accounts. The workable answer is a probabilistic score from other signals, with a threshold — above it, proceed; below it, route the lead to a different channel rather than gambling the domain on it.
Fail toward a different channel, not toward sending
The design decision that matters is what happens when verification cannot produce a verdict. Sending anyway spends reputation on a guess. Discarding the lead wastes the research already done. Routing it to a channel that does not depend on a verified address preserves both — and it means an unverifiable address never has to be tried.
Escalate, do not run everything through everything
Each tier costs more than the last, so each should only see what the previous one could not resolve. Syntax validation is free and runs on everything. The mail-server probe runs on what survives. Catch-all scoring runs only on the ambiguous remainder.
Run in that order, verification is a rounding error on the cost of the research that preceded it. Run in parallel, it is the largest line on the bill.
What to measure
Not reply rate — that measures the message. Measure bounce rate, spam complaint rate, and the share of leads that failed verification and routed elsewhere. The last one is the useful diagnostic: if it climbs, your account discovery is finding organisations whose contacts you cannot actually reach, and the fix is upstream in targeting rather than anywhere in the mail stack.
Axionalytics
Production agentic AI for enterprise engineering, data, and revenue teams.