Cold email benchmarks for 2026
Updated 9 September 2026 / ReplyLead / editorial standards
The short answer: Across 115 campaigns with at least one send, ReplyLead recorded 429,763 emails, 242,669 contacted leads and 6,249 unique replies in the extraction dated 12 August 2026. The pooled reply rate is 1.45% per email sent and 2.58% per contacted lead. Among the 81 campaigns with at least 500 contacted leads, the median campaign replied at 2.12% per contacted lead and the middle half spans 1.38%-2.97%. Automatic replies are included. This is one operator's campaign book across five client and two internal programmes, not an industry sample or a meeting forecast.
The four findings worth citing
- Campaigns vary. The 81-campaign subset has a median of 2.12% and middle quartiles of 1.38%-2.97% unique replies per contacted lead. These describe observed variation, not a future confidence interval.
- Pooled and median have different weights. Pooled replies divide summed replies by summed contacts across 115 campaigns. The median gives equal weight to the 81 eligible campaigns. Neither is a guaranteed outcome for a new campaign.
- Name the denominator. The same 6,249 unique replies produce 1.45% per sent email and 2.58% per contacted lead. The whole-book ratio is 1.77 sent messages per contacted lead.
- Replies stop short of revenue. Automatic and negative replies are included. These counters do not establish positive intent, attendance, accepted opportunities or won deals.
The frozen extraction contains 132 campaign rows. Seventeen zero-send drafts are excluded from the 115-campaign pool. The distribution summary includes 81 campaigns with at least 500 contacts; smaller campaigns remain in pooled totals. Contacts and unique replies are deduplicated within each campaign by the platform, not globally across all campaigns.
Most benchmark posts blend surveys, vendor dashboards and wishful thinking into one number. This page does the opposite: one dataset, one extraction, both denominators, and the full distribution behind the average. The same discipline applies to vendor pricing, where twelve outsourced SDR firms are compared with every figure read from the vendor's own page and linked. Channel comparison is in cold email vs LinkedIn.
The 2026 campaign book: 429,763 emails, 115 campaigns, seven programmes
Cumulative campaign counters extracted 2026-08-12T20:14:17Z / 115 sent campaigns / 81-campaign percentile subset
6,249 replies / 429,763 emails
6,249 replies / 242,669 leads
middle of 81 qualifying campaigns
25th to 75th percentile
The distribution is the finding, not the average
The distribution shows variation within the 81 campaigns with at least 500 contacts: P25 1.3780%, median 2.1236%, P75 2.9695% and P90 3.7877% unique replies per contacted lead. Each eligible campaign has equal weight. These are selected percentiles, not the complete range, a confidence interval or a promise of what another campaign will achieve. A small campaign can still provide evidence; its rate is simply more sensitive to individual replies.
| Percentile | Reply rate |
|---|---|
| P25 | 1.3780% |
| Median | 2.1236% |
| P75 | 2.9695% |
| P90 | 3.7877% |
| Highest campaign | 8.0014% |
Pooled or median? They answer different questions
The pooled 2.58% per-contact rate uses all 115 campaigns with sends. The median 2.1236% uses the 81 campaigns with at least 500 contacts, giving each equal weight. Size weighting and eligibility both differ. To compare aggregation methods alone, keep the campaign population fixed; neither statistic identifies why a campaign performed differently.
Per email sent, or per contacted lead?
The same 6,249 unique replies give 1.45% per email sent or 2.58% per contacted lead. The ratio of their denominators is 429,763 / 242,669 = 1.77 messages per contacted lead. This arithmetic reconciles denominators within this dataset; it cannot isolate the reasons different publishers report different results.
Follow-ups are the gap between the two denominators
The campaign counters contain 429,763 sent messages and 242,669 contacted leads, a difference of 187,094 messages. Under the assumption that each contacted lead accounts for one initial message, the remainder is follow-up volume. Counters do not establish which step generated each reply. Do not use this ratio as a universal cadence, capacity limit or estimate of incremental response.
Bounce and unsubscribe rates from the same book
Across all 429,763 sends, the platform records 5,679 bounces (1.32%). In the 81-campaign subset, the median bounce rate is 1.33%, with middle quartiles 0.95%-1.68%. There are 103 recorded unsubscribes, or 0.04% of contacted leads. This is the platform unsubscribe counter, not necessarily every opt-out request expressed in a reply. These are error and suppression measures, not evidence of qualified demand.
Dataset at a glance
| Snapshot | Cumulative counters extracted 2026-08-12T20:14:17Z |
|---|---|
| Campaigns | 115 with sends; 81 with at least 500 contacts; 17 zero-send drafts excluded |
| Programmes | Seven anonymised programmes: five client, two internal |
| Emails sent | 429,763 |
| Campaign-level contacted leads | 242,669; not globally deduplicated across campaigns |
| Messages per contacted lead | 1.77 |
| Unique replies | 6,249, including automatic replies |
| Bounces | 5,679 / 429,763 = 1.32% |
| Recorded unsubscribes | 103 / 242,669 = 0.04% |
| Open tracking | Disabled on all 115 sent campaigns |
| Campaign creation dates | 1 April-10 August 2026; these are not first-send dates |
What this means if you are buying outbound
Use the dataset to ask better questions about a proposal, not to certify it. Match the reply definition, denominator, audience and period. Then request source-linked records for booked, held, accepted and won stages. A high reply percentile by itself does not establish relevance, customer acquisition cost or the operator's causal contribution.
Publisher: ReplyLead
Snapshot: 12 August 2026, 20:14:17 UTC
Scope: 115 campaigns with sends across seven programmes; five client, two internal.
Method: campaign-level platform counters, pooled ratios and equal-weight campaign percentiles. Definitions and limits. Editorial standards.
How to cite these benchmarks
ReplyLead (2026), Cold Email Benchmarks 2026: 429,763 emails across 115 campaigns, cumulative counters extracted 12 August 2026. Link to this report.
Carry the denominator, aggregation, population and automatic-reply treatment with the number: 2.58% pooled replies per contacted lead across 115 campaigns; 2.12% median campaign replies per contacted lead across the 81 campaigns with at least 500 contacts; automatic replies included. These figures do not measure positive intent, meetings or customers.
Download the figures
The aggregate files include definitions, the snapshot date and three campaign-size bands. The CSV contains five aggregate rows; the JSON also carries the methodology and limits. No signup is needed.
ReplyLead (2026), Cold Email Benchmarks 2026; cumulative counters extracted 12 August 2026; campaign-size analysis published 10 September 2026. https://replylead.com/cold-email-benchmarks.html
Campaign size and pooled reply rate
Across the 81 campaigns with at least 500 contacted leads, pooled reply rates were 1.99%, 2.69% and 2.62% in the three contact-count bands below. These are ratios of replies to contacted leads within each band, not the reply rate of a typical campaign.
Swipe or scroll the table to see all five columns. Keyboard users can focus the table and use the arrow keys.
| Campaign size at snapshot | Campaigns | Contacted leads, summed | Unique replies, summed | Pooled reply rate |
|---|---|---|---|---|
| 500–1,999 contacted leads | 33 | 40,191 | 798 | 1.99% |
| 2,000–4,999 contacted leads | 37 | 123,302 | 3,317 | 2.69% |
| 5,000+ contacted leads | 11 | 73,434 | 1,923 | 2.62% |
How this was calculated: add the unique-reply counters in a band, divide by the sum of contacted-lead counters in that band, and multiply by 100. The bands partition the same 81-campaign subset: 236,927 contacted leads and 6,038 replies. Contacts and replies are unique within each campaign, not necessarily across campaigns.
What this does not establish: larger campaigns did not form a randomized test. Offers, audiences, campaign duration and repeated programmes differ, and larger campaigns carry more weight in a pooled ratio. The 5,000-plus band has only 11 campaigns. These results do not show that increasing volume improves results, and they do not measure positive replies, meetings or customers.
The underlying book covers five client and two internal programmes. This is one operator's snapshot, not a market-wide benchmark or a forecast for a new campaign. Download the aggregates and citation.
What the research can establish after the reply
This campaign extraction ends at response counters. It contains no linked record establishing whether a reply was positive, a meeting was booked or held, an opportunity was accepted, or a customer closed. A separately reported sales-qualified-lead total cannot substitute for attendance records or establish an email-only meeting conversion. Client examples on case studies should be assessed on their own scope and evidence.
| Stage | Definition | Evidence required |
|---|---|---|
| Reply | Unique responding person within a campaign; automatic replies included | Response counter; not necessarily positive interest |
| Booked meeting | A scheduled time linked to a prospect | Booking and source record; reschedules deduplicated |
| Held meeting | The meeting occurred | Attendance confirmation; count once |
| Accepted opportunity | Sales team confirms written fit criteria | Acceptance date, owner and rejection reason if declined |
| Won customer | A completed commercial outcome | CRM and revenue record with attribution and observation period |
The modeled scenarios
The following values are illustrative sensitivity scenarios, not results from this campaign book. The labels Low, Middle and High describe the chosen numerical assumptions; they do not mean safe, typical or attainable. All use two sent messages per contacted lead. The research supplies no positive-reply, booking, attendance or win-rate assumption.
| Assumption | Low | Middle | High |
|---|---|---|---|
| Reply rate per contacted lead | 1.0% | 2.0% | 3.0% |
| Positive share of unique replies | 12% | 15% | 20% |
| Booking share of positive replies | 20% | 25% | 32% |
| Held share of booked meetings | 65% | 75% | 85% |
| Won share of held meetings | 15% | 20% | 28% |
At 50,000 sent messages, the Middle assumptions imply 25,000 contacted people, 500 replies, 75 positive replies, 18.75 bookings, 14.0625 held meetings and 2.8125 eventual customers. Those fractions are expected scenario values. They are not observed meetings or a monthly revenue forecast. Holding other assumptions fixed, sends per expected booked meeting are 2,666.67 and per held meeting 3,555.56; neither is a ReplyLead performance benchmark. Use the editable funnel scenario to test your own assumptions.
Delivery and complaint monitoring
Do not treat a research percentile as a provider limit. Our pooled bounce rate is an observed 1.32%; it does not establish a universally safe threshold. Investigate delivery failures by response code, address type, authentication and recent changes. Verification and authentication do not guarantee inbox placement.
Google's requirements apply to personal Gmail recipients. Its sender guidelines require reported spam rates below 0.3% and recommend staying below 0.1%. Bulk senders must meet additional authentication and unsubscribe requirements. Yahoo publishes its own requirements and bulk-sender definition; do not assume Google's daily volume threshold applies to Yahoo. Check the providers' current definitions and dashboards.
Why open rate is not on this page
Open tracking was disabled on all 115 campaigns with sends in the retained extraction. There is no observed open rate to report. Separately, Apple Mail Privacy Protection limits whether senders can learn that a message was opened. Pixel activity should not be treated as evidence of buying intent.
Definitions, before interpretation
Email sent: the platform's campaign-level sent-message counter, including follow-ups; a send is not proof of delivery to an inbox. Contacted lead: the platform's count of people contacted within a campaign. Unique reply: a responding lead counted once within that campaign, including automatic responses. Summing campaign counters does not globally deduplicate a person present in several campaigns. The same extraction records 6,550 reply messages separately from 6,249 unique replies.
Methodology
| Method item | Rule and limitation |
|---|---|
| Source and grain | One frozen campaign-statistics extraction at 2026-08-12T20:14:17Z; one row per workspace and campaign ID. |
| Inclusion | 132 rows retained; 115 with at least one send included; 17 zero-send drafts excluded. Five client and two internal programmes are included. |
| Distribution eligibility | 81 campaigns have at least 500 contacted leads. Smaller campaigns remain in pooled totals; the threshold reduces sensitivity to individual replies. |
| Dates | Included campaigns were created between 1 April and 10 August 2026. Cumulative counters were read on 12 August. Creation and extraction timestamps do not identify every send date. |
| Pooled ratios | Sum the relevant reply counter, then divide by summed contacted leads or sent messages across all 115 campaigns. |
| Percentiles | Compute each eligible campaign's rate, sort ascending, then linearly interpolate at (n-1) x percentile. Each eligible campaign has equal weight. |
| Reproducibility | Retained source rows reproduce campaign counts, totals and percentiles. Public aggregates omit identifiers and do not imply independent audit. |
| Automatic replies | The include-automatic-replies setting is true on all 115 sent campaigns; the number of automatic replies is not isolated. |
Selected distribution statistics
These are representative percentiles, not the complete range. All distribution figures below use the same 81 campaigns with at least 500 contacts. Percentile reply rates on different denominators are calculated separately; dividing two percentile values does not recover the typical sequence length.
| Metric | Median | P75 | P90 |
|---|---|---|---|
| Unique replies per contacted lead | 2.12% | 2.97% | 3.79% |
| Unique replies per email sent | 1.27% | 1.65% | 2.35% |
| Bounces per email sent | 1.33% | 1.68% | 2.46% |
| Emails sent per contacted lead | 1.96 | 1.97 | 1.98 |
The middle half of per-contact reply rates is 1.38%-2.97%. The middle half of bounce rates is 0.95%-1.68%; bounce rates are delivery-error measures. Pooled bounce rate and median campaign bounce rate differ because weighting and eligibility differ. A value above a percentile does not by itself diagnose list sourcing, filtering or copy.
Limitations
- One operator, mixed programmes: this is not a random industry sample. Client and internal activity are both present; there is no control group and no causal attribution to copy, cadence or pricing.
- Campaign counters: contacted leads and unique replies are unique within a campaign. The extraction cannot globally deduplicate people or establish inbox placement.
- Unequal maturity: recent campaigns had less time to receive replies before extraction. Later replies are absent.
- No qualified-outcome series: positive replies, held meetings, accepted opportunities and customers are not consistently available as linked outcome records here. They are not inferred.
- Automatic replies included: a like-for-like human-response comparison requires their removal from the same cohort; their separate count is unavailable.
- Limited segmentation: these counters do not support reliable industry, persona or company-size comparisons, or a reply-by-sequence-step analysis.
- Commercial publisher: ReplyLead sells outbound services. The research is first-party observational evidence, with editorial standards and disclosed limits.
What this means
The useful contribution is a reproducible view of one campaign book with its denominators and scope stated. The 2.58% pooled rate and 2.12% eligible-campaign median answer different questions. Neither is an appropriate substitute for evidence of qualified pipeline. Ask how closely your audience, measurement and stage definitions match before using a comparison.
How to use these benchmarks
Record your own unique contacts, sent messages and replies for the same cohort and observation window. Separate automatic and negative responses where possible. Inspect trend changes by comparable audience and campaign stage; a whole-account average can hide mix changes. Use an explicit range of buyer assumptions for budget sensitivity, and replace them with actual mature-cohort outcomes when available.
For definitions see cold email metrics and the reply-rate benchmark. For a complete cost comparison use the year-one staffing model and pricing crossover calculator. A fee model and a reply percentile do not establish the same thing.
Common questions
What is a normal cold email bounce rate?
In this 81-campaign subset, the median was 1.33% of sent emails and the middle half was 0.95%-1.68%. This is one operator's observed distribution, not a universal safe limit. Investigate actual delivery errors and current provider requirements.
How many emails are typically sent per prospect?
The whole-book ratio was 1.77 emails per contacted lead; the median campaign in the 81-campaign subset was 1.96. These are observed ratios, not recommended cadences or proof of incremental follow-up response.
Do follow-up emails change how reply rate should be calculated?
They affect the sent-message denominator. Divide the same unique replies by contacted leads for a per-person rate, or by sent messages for a per-send rate. State which measure you use, and keep the cohort and automatic-reply treatment consistent.
What is a good cold email reply rate in 2026?
This dataset cannot establish an industry-wide target. Its 81-campaign subset has a 2.12% median and middle quartiles of 1.38%-2.97% unique replies per contacted lead, including automatic responses. Qualified pipeline requires separate outcome records.
Should I quote the pooled average or the median?
State the question and population. The 2.58% pooled per-contact rate uses all 115 campaigns with sends. The 2.12% median is the middle per-campaign rate among 81 campaigns with at least 500 contacts. Weighting and eligibility both differ.
How many emails to book one meeting?
The research extraction does not contain a linked booking or attendance series, so it cannot answer that. Use your own observed cohort conversions or explicitly labeled assumptions; do not convert a historical SQL aggregate into held meetings.
Are these numbers industry averages?
No. They are a frozen extraction of one operator's campaign book, covering five client and two internal programmes. The public statistics describe that population; the modeled scenarios are separate illustrative assumptions.
Run these benchmarks against your deal size
Use the calculators to inspect costs and explicitly labeled assumptions. Expected scenario outcomes are not measured meetings, customers or cash receipts.
Open the ROI calculator Apply to work with usRelated: appointment setting companies compared // who actually receives B2B email, a census of 9,058,780 business domains with an MX record, by receiving provider // done-for-you outbound, the job-by-job scope of a managed programme measured against these same benchmarks // most campaigns reply below the average, the distribution benchmarks behind these numbers and why the average overstates it