The "average" cold email reply rate is not what you think.

Published rates need comparable definitions.

Instantly's 2026 report, checked 9 September 2026, reports an overall average of 3.43%. It defines its numerator as all replies, including follow-up responses, divided by sent messages. Its tier table gives 5.5% for the top-quartile threshold and 10.7% for the top-decile threshold.

ReplyLead's numerator is unique repliers within each campaign, with automatic replies included. Even a per-send denominator therefore does not make the two metrics identical. Instantly describes active workspaces, but its methodology does not provide enough detail to recreate the headline from public campaign rows. Its exact averaging weights and automatic-reply treatment need confirmation for a strict comparison.

A vendor's published benchmark is evidence about its stated sample and definition. It is not evidence that your campaign should achieve that rate. Check the primary report and eligibility rules before comparing a headline with your own results.

Scene 02 // our complete book

Selected campaign observations, with full-subset summaries.

The analytic subset contains 81 campaigns with at least 500 contacted leads. The dot plot and table display only the 65 campaigns with per-contact reply rates at or above 1%; they are selected views, not the full distribution. Percentiles and outlier calculations retain all 81 eligible campaigns before each stated removal. Focus a displayed dot to read its value.

The 34 sending campaigns below 500 contacts remain in the 115-campaign pooled totals, but are excluded from the 81-campaign mean and percentile calculations. Seventeen zero-send rows are excluded from sending metrics. Publication selection is separate from analytical eligibility; see methodology.

Scene 03 // the signature problem

A few great campaigns can make an "average" look great.

Remove the one, four or eight highest per-contact reply-rate campaigns from the 81-campaign analytic subset. Watch the unweighted mean and median change. The selected dot view keeps its explicit 1% display rule; summary calculations include lower-rate observations.

The mean and median answer different questions, and both can move when observations are removed. This sample does not prove that every cold-email dataset has the same shape. Campaign size, reply classification and eligibility also affect comparability; the outlier exercise isolates only the stated removals.

Scene 04 // the definition trap

One dataset. Three honest "averages".

The two pooled rates use all 115 sending campaigns. The median uses the 81 campaigns meeting the 500-contact floor. Tap a definition to see the scope; the three figures are not interchangeable.

The 81-campaign median is 2.12% per contacted lead. It gives each eligible campaign equal rank. The 115-campaign pool and 81-campaign subset have different inclusion rules, so both weighting and scope must be stated.

These definitions explain our own arithmetic. They do not establish what caused a gap between different vendors' datasets, and they do not prove that either provider performs better.

A benchmark comparison worksheet.

Use the following checks before placing two rates on a shared performance chart. Unknown fields remain unknown; a similar-looking percentage does not fill the gap.

If one field differs, describe the difference instead of calculating a provider-performance multiple. A higher reply rate can coexist with fewer accepted opportunities. Use your own stage records to assess lead quality.

Scene 06 // try it yourself

Build a hypothetical campaign mix.

This educational model assigns the same rate to each campaign in a base group and another rate to an added group. Change group sizes and rates to see how the arithmetic mean and median respond. These are invented inputs, not ReplyLead performance or a recommended target.

Hypothetical default: (10 x 2.2 + 1 x 12) / 11 = 3.09% mean. The median of these 11 assigned rates is 2.20%. Their relative difference is +40%. Adding enough observations can also move the median.

What the send counters can establish.

The 115-campaign pool contains 429,763 sends and 242,669 summed campaign contacts, or 1.77 sends per campaign contact. The arithmetic excess is 187,094 sends. Treating every excess send as a follow-up requires the assumption that each contacted person corresponds to exactly one first-touch send; the aggregate counters alone do not verify each message's sequence position.

The pooled bounce rate is 1.32%: 5,679 bounces divided by 429,763 sends. A bounce is a delivery failure, not a reply or a spam complaint. This sample cannot establish a universally safe bounce threshold. Monitor delivery failures separately from provider spam-complaint requirements.

No per-step reply attribution or reconciled attendance data is present here. A sequence-length distribution cannot establish the best number of follow-ups. Review recipient response and opt-out records when deciding whether to continue a sequence.

Scene 08 // where do you stand

Locate your rate against this sample.

Enter a unique-replier rate per contacted lead to locate it against the selected percentile reference points for all 81 eligible campaigns. This is a descriptive comparison, not a grade for campaign quality, deliverability or commercial return.

Enter a rate from 0% to 100%. No outcome prediction is made.

More numbers from the same programmes: cold email statistics collects ReplyLead's reply, bounce and volume figures with their populations and dates.

When this page does not apply.

• You want a forecast for your own campaign. Every figure here is descriptive of one operator's sample; your list, market and copy set your rate. Use the ROI calculator with your own assumptions.
• You are comparing with a per-send benchmark. Per-send and per-contact rates differ by the number of touches (1.77 sends per contact on this book); run the comparison worksheet above first.
• You need meetings or revenue, not replies. This extraction has no reconciled attendance or won-deal data; the metrics guide covers what to record after the reply.
• You are reading this months from now. The counters are a frozen 12 August 2026 extraction; the current benchmark and method are on the cold email benchmarks page.

Methodology and limits.

The approved extraction was taken on 12 August 2026 at 20:14 UTC. It contains 132 campaign rows across seven programmes: five client and two internal. Seventeen zero-send rows are excluded from sending metrics. All 115 sending campaigns remain in the pooled totals; 81 with at least 500 contacted leads enter campaign mean/median/percentile calculations. The threshold is an editorial inclusion rule, not proof of statistical confidence.

Totals: 429,763 sends, 242,669 summed campaign contacts, 6,249 unique replies, 5,679 bounces and 103 unsubscribes. Automatic replies are included in all sending campaigns. Within-campaign unique-reply counters do not deduplicate people across different campaigns. No held-meeting, accepted-opportunity or won-deal records are supplied by this extraction.

Counters are cumulative at extraction. Campaign creation dates span 1 April to 10 August 2026; they do not identify first-send dates or define a clean activity window. The mean gives campaigns equal weight; pooled rates divide summed counts; percentiles use linear interpolation between ordered campaign rates. The outlier scenarios remove exactly the highest one, four or eight per-contact rates from the full eligible subset.

Public dot plots and the table display only the 65 eligible campaigns at or above 1% per-contact reply rate. Summary calculations include all 81; the full distribution is not published. The selected table cannot reproduce every pooled or subset statistic. This operator-specific sample cannot prove industry performance, causal drivers or the economics of a future campaign.

ReplyLead sells outbound services and publishes this analysis of its own records. See editorial standards, research methodology and publisher identity.

Sources and calculation scope.

Instantly's 2026 benchmark report: primary methodology and reported rates, checked 9 September 2026. Values describe that publisher's stated sample and are not normalised into a ReplyLead performance comparison.

ReplyLead canonical benchmark study: frozen 12 August extraction, with the 115-campaign pool and 81-campaign eligibility distinction. Distribution and threshold tool: pooled versus median interpretation and retrospective threshold counts. Public selections and private source data are distinguished in the methodology above.

Outside primary sources for the definitions this page relies on, each read on the date shown:

1. RFC 5321, Simple Mail Transfer Protocol (IETF): what a 'send' is: one message transferred to a receiving host, the denominator of a per-send rate. checked 24 September 2026.
2. RFC 5322, Internet Message Format, section 3.6.4 (IETF): the In-Reply-To and References headers that technically identify a message as a reply. checked 24 September 2026.
3. RFC 3834, Recommendations for Automatic Responses to Electronic Mail (IETF): what an automatic response is; this page's numerator includes them and says so. checked 24 September 2026.
4. NIST/SEMATECH e-Handbook of Statistical Methods: percentiles: the definition of the 25th, 50th and 75th percentiles used for the 1.38% / 2.12% / 2.97% figures. checked 24 September 2026.

Go deeper in the research