There is no industry reply rate. There are vendor numbers built on different definitions. The two largest 2026 publications give an average of 3.43% (Instantly) and 3.7% (Saleshandy). An agency study covering the same year finds 0.45% (Belkins). lemlist puts the spread at "as much as 7.6x depending on methodology".
The more uncomfortable finding is that the benchmark ecosystem is circular. lemlist's 2026 benchmark post leads with Instantly's 3.43%, not lemlist data. Woodpecker's "2026 benchmarks" put 3.43% as the platform-wide average alongside its own 20M-email dataset. One vendor's number, defined loosely, has become the market's reference point.
The published numbers, side by side
| Publisher | Data | Period | Avg reply rate | Definition / notes |
|---|---|---|---|---|
| Instantly | "Billions of cold email interactions across thousands of active workspaces" | Jan 1–Dec 18, 2025 | 3.43%; top quartile 5.5%+; top tier 10.7%+ | "All replies received (including follow-up responses) divided by total emails sent." No statement on excluding auto-replies. |
| Saleshandy | 53.1M emails, 60,000 sequences, 23,000+ user profiles | Jan–Jun 2026 | 3.7%; top 5% 11–15%; top 1% 15–30% | Replies ÷ delivered; AI-classified into Interested / Not interested / Auto-reply / OOO / Do-not-contact |
| Belkins | 7,530,489 emails, 34,393 replies (agency's own client campaigns) | Jan–Dec 2025 | 0.45%; H1 0.50%; Dec 2025 0.35% | Replies ÷ total sent; auto-replies and bounces excluded; open tracking disabled |
| Woodpecker | "Well over 20 million" emails, 1,000+ customers | Updated June 2026 | Cites 3.43% | Mixes own data with external figures; claims a decline "from 5.1% in 2024 to 3.43% in 2026" |
| lemlist | Secondary (Instantly, Saleshandy, Belkins, Cleverly) | Updated Sept 2026 | Cites 3.43% | Aggregator post; no lemlist dataset disclosed |
| Gong | "Millions of cold emails" | Originally 2021, updated 2026 | Not published | Reports relative effects on meetings booked, not reply-rate baselines |
Why the numbers disagree
1. The denominator. Belkins explains its own drop: earlier studies "calculated against the number of unique recipients who opened an email". The 2026 study uses total sends. Its line: "A 5% reply rate against openers and a 0.45% reply rate against total sends can describe the same campaign." Since opens are unreliable (see Content and tracking), any opens-based reply rate is doubly unreliable.
2. What counts as a reply. Belkins excludes auto-replies. Instantly's definition counts "all replies received" and doesn't say whether out-of-office messages are filtered. lemlist attributes the Belkins gap to this, saying one source "counts every response that hits the inbox, the other counts only humans typing back". Belkins itself points to the denominator. Both effects are probably real, and neither has been quantified publicly.
3. Who is in the sample. Instantly and Saleshandy aggregate self-serve users: founders, agencies and SDR teams of every skill level, sending to whatever lists they bought. Belkins measures a single agency's managed campaigns, skewed to mid-market and enterprise targets, where its own data shows 10,000+-employee companies replying at 0.22% versus 0.72% for 0–10-employee companies. Woodpecker finds campaigns under 50 recipients average 5.8% versus 2.1% for 1,000+. List size and segment move the average more than copy does.
4. Averaging method. "Average" can mean total replies ÷ total sends (volume-weighted, dominated by the biggest senders) or the mean of campaign rates (dominated by many small campaigns). None of the reports says which it uses.
Saleshandy reports a 3.7% average reply rate and an "average positive reply rate across our dataset" of "3% to 5%". A positive reply rate cannot sit at or above the total reply rate on the same base. Either the bases differ (e.g. per campaign vs. pooled) or one figure is wrong. The same post describes a dataset of "67.4 million connected email accounts", which is hard to square with 23,000 user profiles.
What the data agrees on
Where several sources point the same way, the finding is more robust than any single average:
- Follow-ups matter, but the first email does most of the work. Instantly: 58% of replies come from step 1, 42% from follow-ups. Saleshandy: follow-ups generate 44% of positive replies. Both recommend about 4–7 touches.
- Smaller, tighter lists reply more. Woodpecker (<50 recipients: 5.8% vs 2.1%) and Saleshandy (<200 prospects: 2x reply and 4.4x positive reply vs >1,000).
- Senior people at small companies reply most. Belkins: founders/owners 0.57%, C-level 0.42%, VPs 0.32%.
- Short emails win. Instantly: elite campaigns under 80 words.
- Rates fell in 2025. Belkins shows a slide from 0.50% in H1 to 0.35% in December 2025 and blames tighter Google and Microsoft filtering. Woodpecker cites inbox saturation, enforcement and "low-effort AI-generated outreach".
Positive replies and meetings: the numbers that matter, and are barely measured
Reply rate is a vanity metric when out-of-office replies and "remove me" count. The outcome measures are thin:
- Saleshandy: positive replies 3–5% (inconsistent, as above), and top campaigns book "2 to 3 meetings for every 100 cold emails". The latter is a top-performer figure, not an average.
- lemlist cites Cleverly for a 0.5–2.5% meeting-booked rate (secondhand; the original was not checked).
- Gong reports only relative effects. One phrase that raises reply rates "decreases meetings booked by 14%". That shows reply rate and meeting rate can move in opposite directions.
No 2025–26 source found for this pass publishes a positive-reply or meeting rate with a clear denominator, a reply-classification method and a distribution (median, quartiles) rather than an average.
Benchmarks are marketing assets. Instantly's report sells the idea that 10%+ is attainable on Instantly; Belkins' low number sells the idea that you need an agency. Read every figure as "replies per send, as defined by someone who sells something", and anchor on your own baseline.
A defensible way to state a benchmark
If you publish or display benchmarks, state: (1) denominator: delivered sends; (2) numerator: human replies, with auto-reply/OOO removed by a stated classifier; (3) positive replies by a stated rubric; (4) median and quartiles, not just a mean; (5) cohort filters (segment, list size, sequence step, recipient provider); (6) period and sample size.
What this means for an entrant
- Publishing an honest benchmark is a cheap distribution wedge. The space is dominated by one loosely defined number that competitors repeat. A methodology-first annual report (human replies, positive replies, meetings, medians, by segment) would be cited, and would position you as the serious alternative. See Distribution.
- Put benchmarks inside the product. Show each customer their positive-reply rate against a cohort that matches their segment, list size and sequence step, with the denominator printed on screen. Incumbents show raw reply rates against a single global average.
- Classify every reply. Positive-reply measurement depends on reliable auto-reply, OOO, unsubscribe and objection detection. That is the base layer of AI reply agent and Campaign analytics. Saleshandy already does it, so it is table stakes heading into 2027.
- Track through to meetings. Gong's finding that reply-boosting copy can reduce meetings means optimising for replies alone is wrong. Calendar and CRM linkage (CRM sync) makes meeting rate the default north-star metric.
- Don't promise customers a number. With averages ranging 0.45–3.7% for the same year, any reply-rate promise is a churn trap. Sell measurement and improvement against their own baseline.