If you've ever tried to benchmark your cold email reply rate against "industry average" and found numbers ranging from under 1% to over 8% depending on which report you opened, you're not losing your mind. The reports really do disagree that much, and the reason has almost nothing to do with your campaign quality.

The denominator problem

Take two widely cited 2026 benchmark reports. Instantly's Cold Email Benchmark Report, drawn from billions of interactions across its user base, puts average reply rate at 3.43%. Belkins' 2026 study, based on 7,530,489 managed client sends, reports just 0.45%. Neither number is wrong. They're measuring different things.

Software platforms like Instantly and Woodpecker typically calculate reply rate as unique replies divided by delivered emails or unique opens, filtering out hard bounces and unauthenticated sends before the percentage gets calculated. Full-service agencies like Belkins use total emails sent as the denominator, gross volume, no filtering. Belkins recorded 34,393 unique replies out of 7,530,489 total sends for that 0.45% headline figure. Recalculated against opened or delivered mail instead of gross sends, that same dataset lands in the 3-5% range other platforms report. Same campaigns, same results, wildly different-looking number depending on what you divide by.

The practical takeaway: before you compare your reply rate to any published benchmark, check what's in the denominator. A 3% reply rate against delivered mail and a 3% reply rate against total sends are not describing the same performance.

What the numbers actually say, once you account for that

With the caveat above in mind, a few reasonably consistent figures show up across independently collected 2026 datasets:

  • Instantly's 2026 report puts average reply rate at 3.43% platform-wide, with 5.5%+ counted as top-quartile and 10.7%+ as elite (top 10%).
  • Woodpecker's dataset, aggregated across more than 20 million sent emails, frames 2-5% reply rate as "average," 5-8% as "good," and 8-15%+ as "excellent."
  • Belkins' 2026 study puts the agency-methodology (total-sends) figure at 0.45%, and this is the single best explanation for why benchmark numbers you'll find online disagree by 5-10x. If you see a very low published reply rate, check whether it's counting total sends before assuming your campaign is underperforming.
  • Lavender's Cold Email Benchmark Report, based on 231,818 emails across roughly 50,000 inboxes, found that shorter emails and higher-graded copy meaningfully move reply rates, from around 3.4% up to 4.3%, a roughly 27% lift, tied to copy quality rather than volume.

What actually moves the number: sequence structure and channel pairing

Two findings show up consistently enough across independent vendor datasets to be worth building process around, even with the caveat that exact multipliers vary by source.

First, where replies come from in a sequence. Both Instantly's 2026 report and Astra GTM's 2026 benchmarks independently converge on the same pattern: roughly 58% of all campaign replies come from the first email, with the first follow-up adding another ~20%. That means the first two touches in a sequence are doing the overwhelming majority of the work. If your first email isn't earning a reply, more follow-ups aren't going to compensate proportionally.

Second, channel pairing beats channel purity. Gong Labs' analysis of more than 85 million cold emails found that pairing cold calling with an email cadence nearly doubles the email reply rate on its own, from 1.81% email-alone to 3.44% when a call is part of the same sequence, even when the prospect doesn't pick up. The call itself doesn't need to connect to move the email number.

What to ignore

You'll see benchmark tables online broken out by industry vertical (SaaS vs. healthcare vs. legal, etc.) and by company size or buyer seniority, often with very precise-looking percentages down to the decimal. Treat those specific numbers skeptically. The directional idea, that less digitally saturated verticals and smaller target companies tend to see higher engagement than saturated ones like SaaS or cybersecurity, shows up consistently enough to be a reasonable planning assumption. But the exact percentages circulating in these tables often can't be traced back to a single verifiable dataset, and different vendors' "same" vertical numbers frequently contradict each other. Use the direction, not the decimal point.

Open rate is worth a similar caution, for a different reason. Reported open rates in current benchmark reports vary widely, largely because Apple Mail Privacy Protection pre-loads tracking pixels via proxy servers regardless of whether a human actually opened the message. That inflates the number in ways that are hard to back out cleanly, which is part of why open rate has become a weaker signal than it used to be. Reply rate and meeting-booked rate, both of which require an actual human action rather than a pixel load, are the more trustworthy numbers to optimize against.

How to actually use these benchmarks

The honest way to benchmark your own campaigns is to first identify which denominator you're using internally, delivered, opened, or total sent, and then only compare yourself against reports using that same denominator. If you can't tell what a published number is counting, don't anchor to it.

Beyond that, the two most actionable, cross-verified levers are sequence design (your first two touches carry roughly 78% of total reply volume, so that's where editing time belongs) and channel pairing (adding a call to an email cadence roughly doubles email reply performance on its own). Both of those are things you control regardless of which vendor's benchmark table you're staring at.

Given how much of the outcome rides on the first email and how differently "reply rate" gets defined across tools, it's worth having a second check on that first send before it goes out, something that scores relevance and clarity against what actually correlates with replies rather than against a vanity benchmark that isn't measuring what you think it is.