Is a 2% cold email reply rate good?

There is no single verified figure. Belkins measured 0.45% replies per total send across 7,530,489 emails in 2025. Instantly measured 3.43% per send the same year. Different denominators, different populations, both real.

What counts as a good cold email reply rate?

Belkins' 0.45% blended (US 0.51%, founder-level 0.57%) is below 1%, but only under a strict per-send denominator it adopted deliberately. Instantly's 3.43% describes the same category measured differently. Both channels still produce replies at scale.

Is cold email illegal under GDPR or CAN-SPAM?

Regulated, not banned. CAN-SPAM never required prior consent; it prohibits deceptive headers and requires an opt-out and a physical address. GDPR Recital 47 allows direct marketing on legitimate-interest grounds. UK PECR permits emailing corporate bodies without prior consent. This is not legal advice.

Denominator Decoder · Evidence system

See what your cold email reply rate actually means.

Paste any reply-rate number, yours or a vendor's, say what it was divided by, and get it re-expressed on a stated denominator against dated, sourced benchmarks. No email gate and no AI interpretation.

Evidence version 2026-08-08 · Last audited August 8, 2026

The star mechanic

Decode a reply rate.

A reply-rate figure without a stated denominator is not a metric. Enter the number and what it was divided by. The decoder normalizes it against sourced benchmark bands, or tells you honestly when the comparison cannot be made.

Claim checker

What have you heard?

Or try one of these reviewed sample claims:

Ungated strategy check

How many of your outbound rules rest on unverified claims?

Select the claims that currently drive your cold-email process. This is an evidence-exposure score, not a deliverability or performance score.

Structured feedback

Did the evidence record help?

One evidence layer

Every verdict and every benchmark comes from the same versioned dataset as the Claims Index.

The decoder, the checker, and the Index are generated from one claims file. Each record has a stable ID, aliases, verdict, finding, primary source, action, category, evidence status, first and last review dates, research version, and change note.

Questions about the decoder

Why does it sometimes refuse to compare my number?

An opens-based or unique-opener reply rate cannot be honestly converted to a per-send figure, because Apple Mail Privacy Protection inflates opens. When the denominators cannot be reconciled, the decoder says so rather than inventing a conversion.

Are the benchmarks or verdicts produced by AI?

No. Every band and verdict is a human-reviewed record from the same versioned dataset as the Claims Index. The normalization is deterministic arithmetic and taxonomy lookup, not a model call.

How often is the evidence reviewed?

The active dataset is reviewed at least quarterly and whenever Belkins, Instantly, Pin, or Saleshandy publishes a newer benchmark. Each record shows its last review date and change note.

Browse all 14 reviewed claims · Inspect the full method · Read the evidence answer · Watch the demonstration · Read the Tool Trust Charter