The best cold email benchmark for you is your own first 100 sends. Borrowed numbers describe other people's lists, other offers and other ways of counting, and they disagree with each other. Measure four things (bounces, replies, positive replies, meetings per 100 sent), skip opens, and after 100 sends write down a floor that tells you when to stop and a target that tells you what to aim for. This page shows how, with one sender's real arithmetic, including a part most guides leave out: a number from 100 emails is a range, not a point.
Why the numbers you find online don't transfer
Search for the average cold email reply rate and you'll find figures from about 2 percent to over 5, each from a different pool of senders. Some divide replies by emails sent, others by emails delivered. Some count "not interested" as a reply, some don't. Some pools are mostly software-sales campaigns to executives, which tells a plumber very little. None of them knows how good your list is. Use them as a rough sense of scale, never as a grade. The counting method behind your own numbers is in what is a good cold email reply rate, and the weekly routine is in cold email KPIs.
The four numbers worth measuring
Bounce rate (bounced / sent) tells you if the list is clean. Reply rate (real replies / delivered) tells you if anyone responds. Positive reply rate (interested replies / delivered) tells you if the right people respond. Meetings per 100 sent is the one that pays. If you're unsure about bounces, cold email bounce rate has the pause rule. There is a fifth number people love, the open rate, and it's the one to drop.
Why opens are the one to drop
An open is recorded when a tiny image loads. Mail apps that preload images count an open nobody made, and readers who block images are never counted. The figure can't tell you whether a person read your email, so it can't tell you what to change. Stop measuring opens explains it fully. Tracking images can also hurt delivery, so removing them costs you nothing.
Your first 100 sends, worked through
Gus owns Pellam Packaging and sells corrugated boxes to local food makers. He sends 100 emails in four batches of 25 over two weeks, same message, same kind of buyer. Results:
| Number | Count | Rate | Plausible range for the true rate |
|---|---|---|---|
| Bounced | 3 of 100 sent | 3.0% | 1.0% to 8.5% |
| Real replies | 6 of 97 delivered | 6.2% | 2.9% to 12.8% |
| Positive replies | 2 of 97 delivered | 2.1% | 0.6% to 7.2% |
| Meetings | 1 of 100 sent | 1.0 per 100 | 0.2 to 5.4 per 100 |
The last column is a standard 95 percent confidence range (the Wilson interval, if you want to look it up). Read the reply row this way: Gus saw 6.2 percent, but with only 97 delivered, the true rate could sit anywhere from about 3 percent to about 13 percent. Two replies more or less moves the number by two points. That is why one week of data should never make you rewrite a message. It is also why you shouldn't be crushed by a bad first batch.
Setting a floor and a target
A floor is the line below which you stop and look. A target is what you'll aim for next. Set both from your own first 100:
| Number | Gus's floor (act if worse) | Gus's target |
|---|---|---|
| Bounces | More than 5 in a batch of 100 (pause and verify the list) | Under 2 in 100 |
| Replies | Fewer than 3 in the next 97 delivered (half his baseline) | 8 in 100, after one change |
| Positive replies | 0 in the next 100 | 3 in 100 |
| Meetings | 0 in the next 200 | 2 per 100 |
The floors are about half the baseline, because ordinary luck can swing a count that far. The targets add one or two replies, which is a believable gain from one change, such as a sharper first line (see cold email opening lines). The bounce floor of 5 in 100 is a cautious rule of thumb, not a limit set by anyone else.
Reading the next batch
Gus changes one thing, the first sentence, and sends 100 more. He gets 8 replies from 96 delivered. Is that better than 6 of 97? Probably, but 8.3 percent against 6.2 percent isn't proof, and the ranges overlap almost entirely. He keeps the new line and sends 200 more. After 388 delivered and 24 replies, his rate is 6.2 percent with a range of about 4.2 to 9.0 percent. The range narrowed. At 400 sends the baseline is solid enough to reset, and that is what you do: reset the floor and target from the larger sample. The method for fair comparisons is in how to test a cold email with 50 prospects, and how long to wait for revenue is in how long does cold email take to work.
What would change your benchmark
A new list, a new offer, a new audience or a new season all reset the baseline. A list of restaurant owners and a list of dentists are different experiments. Keep a separate sheet for each, and don't blend them into one average that describes neither.
Rocketship finds the buyers, emails them from your own Gmail, follows up, and books the meeting on its own, so the numbers above come out of your own inbox and calendar. Plans that get you customers start at $24.99 a month (Launch), and both Launch and Frontier start with a 7-day free trial, card required. See the AI workers.