Does cold email still work?

The evidence, with its denominators showing

Numbers first, definitions attached - because reply-rate claims without denominators are the category's favourite trick. Our 2026 campaign book covers 402,477 emails across 89 campaigns for 5 client programmes: 230,740 contacted leads, 5,844 unique replies. Pooled, that is 2.53% of contacted leads replying; taken campaign by campaign, the median is 2.12%. Per contacted lead, not per email sent - a 2-touch cadence means sends outnumber leads, and quoting per-email rates flatters nobody honestly. The full book, updated as campaigns close, lives at cold email benchmarks.

One distribution fact reframes most disappointment: most campaigns reply below the pooled average, because strong campaigns contribute disproportionate volume to the pool. A first campaign landing under the headline number is the expected case, not a failure verdict - what matters is which side of the median it lands on and why. What a realistic target looks like for a specific list and market is worked through in what is a realistic cold email reply rate.

What "working" means, concretely

A reply is the unit of evidence, not the goal. The chain runs replies, then qualified interest, then meetings held, then revenue closed - and each link is measurable. We track the seven numbers that predict revenue rather than vanity counts; open rates in particular have been unreliable since privacy proxies began prefetching pixels, and any programme still steering by them is navigating with a broken compass. When someone asks whether cold email works, the operational translation is: does it produce qualified meetings at a cost per meeting that your deal size supports? On our measured funnel, run properly, it does.

The four ways programmes actually fail

"Cold email stopped working" almost always decomposes into one of four named failures:

• The list. Unverified data bounces, decayed data misses, and even verified lists carry traps - catch-alls, retired addresses, gateway rejections counted as bad data. Past a 2% bounce rate the campaign is damaging its own infrastructure with every send.
• The infrastructure. Sending from an unwarmed mailbox, a single domain, or at inflated per-mailbox volume puts mail in spam before copy gets a vote. Infrastructure design, the warmup ramp and volume discipline are the unglamorous majority of deliverability.
• The copy. Long, vague, template-scented messages fail against any list. The rules that decide replies are mechanical enough that our grader scores drafts against them.
• The fit. The one no vendor mentions: a market that does not buy from outbound, an ACV that cannot carry acquisition cost, an ICP unreachable by email. No execution quality fixes a fit problem - which is why the honest answer to "will it work for us" is sometimes no, below.

Is it getting harder? Yes - asymmetrically

Two real headwinds, neither fatal. Providers keep raising the sending bar - authentication requirements and bulk-sender thresholds that formalised in 2024-2025 - which punishes casual senders and, in doing so, quietly clears the field for disciplined ones. And AI has made mediocre volume free: inboxes now drown in fluent, interchangeable, obviously generated mail. Our grader maintains a 580-term AI-tell vocabulary because recipients pattern-match that voice instantly and discount it. Both trends transfer results from undisciplined senders to disciplined ones rather than shrinking the channel - concentration, not decline.

Who should not use cold email

The disqualifiers, stated plainly: deal sizes too small to carry the cost of a programme run properly; markets where buyers demonstrably live somewhere else - some are phone-first, some are network-native; offerings whose addressable list is a few dozen accounts, where account-based touch beats fleet economics; and anyone unwilling to run the system - verification, warmup, cadence discipline, staffed replies - because a half-run programme spends the money and skips the results. One caveat on that last group: unwillingness to run the system is not unwillingness to use the channel. Having it run for you is a normal buying decision, and lead generation services is the guide to making it well.

Common questions

Is cold email dead in 2026?

The measured opposite: 402,477 sends this year produced replies from 2.53% of contacted leads pooled, median campaign 2.12%, converting to qualified meetings. What is dying is undisciplined volume - the rules tightened, and the results concentrated among senders who run the system properly.

What reply rate should I expect?

Plan around the median, not the pool: 2.12% of contacted leads on our campaign book, with wide spread by list quality and market. The distribution and what moves a campaign up it are on the reply-rate distribution.

How long does cold email take to show results?

The mechanical lead time is infrastructure: mailboxes need their two to two and a half week warmup before carrying campaigns. Once live, replies arrive with the sends themselves - a campaign's early sends are already evidence of where it sits.

Does AI-written email still work?

Generated sameness is now the strongest tell in the inbox - fluent, interchangeable and discounted on sight. AI assistance survives only under editing discipline that removes the voice; wholesale generation is how a campaign joins the noise it competes with.