Cold Email Benchmarks From 2.5M Sends: Reply Rates, Positive-Reply Share, Inbox Placement, and Bounce

If you are evaluating cold email and you have read a few vendor "state of outbound" reports, you have probably noticed they all conclude that their own tool is the reason numbers go up. That is not data, that is a sales deck with a chart in it. You need actual operator numbers so you can tell whether a campaign is healthy, whether your agency is underperforming, and what to expect before you spend a quarter on it.

These are the benchmarks we see across more than 2.5 million cold emails sent and 950,000+ contacts enriched. No projections, no rounding up to make a point. Where our numbers are better than the industry, it is because of how the system is built, not magic, and we will tell you exactly which lever does the work.

Inbox placement: the number that gates everything else

Inbox placement is the percentage of sent emails that land in the primary inbox instead of spam or the promotions tab. It is the first metric to check because every downstream number is multiplied by it. A 5% reply rate on copy that only reaches 60% of inboxes is really a 3% reply rate on your list. The copy did not fail. The infrastructure did.

On shared sending infrastructure, the kind you get when an agency runs all its clients through one pool, real-world placement sits around 60%. Our average across active campaigns is 98.5%. The gap is not a trick. It comes from dedicated sending domains separate from your main domain, 52 warmed mailboxes spread across Google, Microsoft, and Azure so no single provider sees suspicious volume, correct SPF, DKIM, and DMARC, and sending volumes kept low per mailbox. If you want to understand how that setup is built, the deliverability page walks through the sending stack, and you can run your own subject lines and body copy through the free spam-words checker before you send anything.

Bounce rate: a data-quality signal, not a deliverability one

Bounce rate is the share of emails that come back undeliverable. People confuse it with deliverability, but a bounce usually means the address was wrong, not that you landed in spam. It matters because high bounces tell mailbox providers your list is dirty, and that reputation damage spills onto the messages that do reach real people.

We run between 0.15% and 0.9% bounce. The common cause of a bad bounce rate is decayed Apollo data: a single-source export where 15 to 30% of the addresses are stale, a job change away from invalid, or simply guessed. We solve it with waterfall enrichment, checking an email against multiple providers in sequence and verifying it before it ever enters a sequence, across more than 1,800 production Clay tables. If your current bounce rate is above 2 to 3%, the problem is almost always the data, not the sending. The Clay enrichment page covers how the waterfall is wired.

Reply rate and positive-reply share: the two are not the same

Reply rate counts every response, including "unsubscribe me" and "who gave you my email". It tells you little on its own. The metric that maps to revenue is positive-reply share: of the people who replied, what percentage are actually interested. A 10% reply rate that is mostly negative is worse than a 4% reply rate that is mostly qualified conversations.

Our target positive-reply share is 25 to 30%. That is driven less by clever copy and more by targeting the right account at the right moment. Signal-based triggers, a recent funding round, a relevant hire, a tech-stack change, a job change, mean you reach someone when the problem you solve is live for them. On the ATI campaign, 78,000 emails produced a 37% positive reply rate and over $300K CAD in pipeline. On the GearLocker campaign, a proprietary 66,000-school database produced 194 interested replies. You can read the full breakdowns on the results hub, and the mechanics of trigger-based timing on the signals page.

Timelines: what to expect across a 90-day window

The single most common reason campaigns get judged a failure is that they are judged too early. New mailboxes need warming. Domains need reputation. The first weeks of any honest cold email program are about building the asset, not harvesting from it.

A realistic shape: weeks 1 to 3 are infrastructure and warming, with low send volume and few replies by design. Weeks 4 to 8 are when volume ramps, placement stabilizes near the 98% range, and positive replies start landing consistently. By weeks 9 to 12 you have enough reply data to know your real positive-reply share and to cut the segments and messages that underperform. That is the structure of our fixed-scope 3-month pilot, and at day 90 you keep the entire system: the domains, the 52 mailboxes, the Clay tables, and the self-hosted n8n automation. If a provider tells you results in week one, they are running you through warm shared infrastructure that will eventually burn, or they are counting opens.