Benchmarks

Last verified: 2026-08-05

TL;DR

• Published reply-rate averages for 2026 run from 0.45% to 3.43%. Both are honest; they divide by different things ^{[1, 2]}.
• Before adopting any benchmark, establish three things: the corpus, the year, and the denominator. Without all three the number is decoration.
• Your own last campaign is a better benchmark than anyone's published average.

The current numbers, with their owners

Why 0.45% and 3.43% are both true

The agency corpus divides replies by every email sent, counting each follow-up in a sequence as its own send ^[2]. The platform corpus reports replies against contacts in a campaign ^[1]. A four-touch sequence therefore produces a denominator roughly four times larger under the first definition — and a number roughly four times smaller for identical performance.

This is the single most useful thing on this page. A team that adopts 3.43% as its target while measuring the way the agency does will conclude it is failing by a factor of seven, and will start rewriting copy to fix a definition.

Benchmarks by segment, not in aggregate

Aggregate averages hide the effect that actually drives results. From the same 7.5M-email corpus: companies of 0-10 employees reply at 0.72%, 11-50 at 0.49%, and enterprises above 10,000 at 0.22%; by role, founders lead at 0.57%, C-level 0.42%, VPs 0.32% ^[2]. A separate corpus reports lists under 50 contacts at 5.8% against 2.1% for lists over 1,000 ^[1].

So "is 1.5% good?" has no answer without knowing who you emailed. Compare like with like or do not compare.

Contradictions are normal here

Two reputable corpora disagree about where replies come from: one reports 58% of replies arriving on the first email and 42% on follow-ups (2026) ^[1], another that 55% of replies come from a follow-up (2023) ^[3]. Both are large samples honestly reported.

Possible reasons — different sequence lengths, different eras, different customer bases — are unresolvable from outside. The honest reading is that follow-ups produce a large share of replies, somewhere near half, and that any more precise claim is borrowed confidence. Sequence design implications are in Sequence architecture.

Benchmarks decay

Reply rates have declined industry-wide as volume rose and filtering tightened, so an undated benchmark is worse than no benchmark — it is an old number presented as a current one. Two rules follow: never quote a rate without its year, and treat anything older than about two years as history rather than as a target. The 2019 studies still widely cited for cold-email performance were measured on a link-building corpus in a different filtering era ^[4].

Use yourself as the benchmark

Published averages are useful once, to tell you whether you are in the right order of magnitude. After that the only comparison that controls for your product, list and market is your own previous campaign, measured the same way. Keep the definitions stable across campaigns even if you decide they were the wrong definitions — a consistent wrong denominator is more useful than a corrected one that breaks the series.

References

1. Woodpecker — Cold email statistics (20M+ email corpus, updated Jun 2026)
2. Belkins — Cold email response rates (7.5M emails sent in 2025, updated Jun 2026)
3. QuickMail — Cold email statistics (65M journeys and emails, Dec 2023)
4. Brian Dean — Email outreach study (12M-email study with Pitchbox, 2019)

• Up: Playbook contents › Measurement and iteration
• Previous: Core metrics
• Next: Diagnosing underperformance