TL;DR
- The median cold email bounce rate is 1.54%, across 850M+ sends in Smartlead's 2026 data.
- The top quarter stays below 0.73%. The 75th percentile is already at 3.21%.
- Every major platform puts "good" at 2-3% or less, and starts pausing or warning around 5%.
The benchmark: Smartlead's 850M sends
The largest published dataset we found is Smartlead's State of Cold Email 2026, summarised on its bounce rate post (checked 2026-10-17). It covers more than 850 million sends.
| Where you sit | Bounce rate |
|---|---|
| Top quarter of senders | Below 0.73% |
| Median sender | 1.54% |
| 75th percentile | 3.21% |
So a quarter of senders bounce more than 3%, which is above where most platforms say trouble starts. The gap between the top quarter and the bottom is mostly list quality: how fresh the data is, whether it was verified, and what happened to the catch-alls.
How Smartlead grades a bounce rate
The same post sets bands measured over a rolling 14 days:
| Band | Bounce rate (14 days) | Smartlead's description |
|---|---|---|
| Exceptional | Below 1% | Usually heavy double verification |
| Normal | 1-2% | Single-pass verification and reasonable hygiene |
| Warning | 2-4% | Inbox providers start scoring the sender down |
| Too high | Above 4% | On any mailbox with 100+ sends |
| Effectively burned | Above 8% | Treat the mailbox as burned |
Smartlead also cites 25-30% annual data decay, and notes that catch-all-heavy lists can produce bounce rates of 8% or more.
What each platform publishes
| Platform | Recommended | Enforced brake |
|---|---|---|
| Instantly | 2% or lower | Auto-pause above 5% after 200 sends (on by default) |
| Smartlead | Under 2% over 14 days | Opt-in auto protection at a threshold you set |
| lemlist | Under 3%, pause above 5% | Alert at 10%, auto-pause at 50% after 100 leads |
| Salesloft | 2% benchmark, 5% or less "healthy" | None |
| HubSpot Sequences | Under 3% "good" | None for sequences; marketing email can be suspended above 5% |
| Outreach | n/a | Mailbox locked 24h at 125 bounces, or above 5% after 125+ sends |
| Reply.io | n/a | Warning at 4%; step 1 stops at 8% in 24h with 200+ sends |
| Snov.io | n/a | Pause at 15% of the last 100 sends |
From each platform's help center or blog, checked 2026-10-17: Instantly, Smartlead, lemlist, Salesloft, HubSpot, Outreach, Reply.io (Reply flags these newer docs as not yet confirmed against the live product), Snov.io.
What Google and Microsoft say
Neither publishes a bounce-rate threshold. Gmail's bulk sender rules publish a spam complaint limit (stay under 0.3%), not a bounce limit. That doesn't mean bounces are free: sending to addresses that don't exist is one of the clearest signs of a scraped or bought list, and it feeds into how your domain is rated. More in how many bounces burn a domain.
Catch-alls: the benchmark inside the benchmark
One segment bounces far more than the rest. In Hunter's test, accept-all (catch-all) addresses bounced 27% versus 1% for valid ones, and 38% of domains in its sample were accept-all. If your overall rate is in the warning band, check how many catch-alls went out before you blame anything else. See should you send to catch-all emails?
How to get into the top quarter
- Verify every list before import, including lead-database exports marked verified. The full workflow.
- Separate catch-alls into their own slower campaign, or skip them on new domains.
- Re-verify lists older than about 30 days, and before the later steps of long sequences.
- Measure per mailbox, not just per campaign. A healthy average can hide one mailbox that's bouncing badly.
Verification starts at $29 for 5,000 checks, and credits never expire. See pricing or estimate your cost with the cost calculator.