Cold Email Benchmarks: How to Build Your Own Outreach Baseline

A benchmark helps only when it measures the same thing you measure. In cold outreach that rarely holds, and the reasons stack up.

Start with definitions. One report divides replies by emails sent, another by people contacted, a third by people delivered. Some count out-of-office notices and unsubscribes as replies. Run one campaign through each formula and you get three different “reply rates”.

Then the audience. Writing to founders of ten-person agencies and writing to procurement heads at large manufacturers are different jobs. An average across both describes neither.

The rules of sending change too. Google, for example, introduced requirements for everyone sending to Gmail accounts from February 2024, as set out in its email sender guidelines. A figure collected before a change like that may describe a market that no longer exists.

Survivorship bends the numbers upward. Campaigns that flop get stopped early and never reach a case study. Averages built from whatever kept running, or whatever made a good blog post, lean optimistic.

And the sample is often opaque. Many figures come from one vendor’s customers, people who chose that vendor and used it in particular ways. Without the method, you can’t tell what population the number stands for.

None of this makes the publishers dishonest. Their numbers answer a different question from yours. Yours is narrower: how your outreach, to your buyers, is performing this quarter, and whether it is getting better.

There is no single average worth managing against. Reply rates move with your offer, your list, your market and your definitions far more than with any industry norm.

If a published figure does land on your desk, put these questions to it before it shapes a decision:

• Exactly how a reply was defined, and whether automatic replies were excluded
• Whether the denominator was emails sent, people contacted or people delivered
• Which industries, company sizes, regions and buyer titles were included
• When the data was collected, and over how long
• How many campaigns and people sit behind the figure
• Whether poorly performing campaigns were included or filtered out

Even a well-documented number describes someone else’s offer sent to someone else’s list. Two companies writing to the same CFOs can see very different results because one fixes a problem the CFO has this quarter and the other doesn’t. Offer and targeting usually move the number far more than any average would suggest.

A figure that fails most of these checks is a conversation starter at best. Manage against your own trend, measured the same way every month.

A baseline is your recent performance, measured consistently, with enough context written down that you can explain it later. It takes six steps.

1. 1Fix your definitions, one formula per metric, and write them down. The guide to outreach metrics that matter sets out a full chain from contacted to opportunity.
2. 2Count people, not messages. Someone who receives four steps and replies once is one contact and one reply.
3. 3Group people into cohorts by the week or month they were first contacted, then measure each cohort’s replies and meetings over a fixed window afterwards.
4. 4For every cohort, record the segment, list source, message version, channel mix and sending inboxes.
5. 5Work out each metric for every segment you target, as well as the total.
6. 6Write the baseline down as a dated table with definitions, sample sizes and results. Every later test is read against it.

One row per segment and cohort is enough, with these fields:

The first version will be rough. Its job is to give next month something to be compared with. Don’t skip the notes column: six months on, a line like “holiday week, two inboxes paused” is the only explanation anyone will have for the dip.

Keep the table where everyone who changes outreach can open it, next to the campaign calendar the team already checks. A baseline only its owner can find stops being used the first week that person is on leave.

Small samples are noisy. Contact fifty people and a single extra reply moves your reply rate by two percentage points. Before reading anything into a number, estimate how far chance alone could move it.

A standard rough approximation for a rate measured on a sample:

Margin ≈ 2 × √( rate × (1 − rate) ÷ people )

The result is roughly the range either side of your measured rate where the true rate probably sits. An illustrative example: you contact 400 people and 20 reply, a 5% reply rate. The margin is 2 × √(0.05 × 0.95 ÷ 400), about 2.2 percentage points. The true rate for that audience could plausibly be anywhere from about 3% to 7%.

Three things follow. Halving the margin takes four times the people: going from 400 to 1,600 narrows the range from about ±2.2 points to about ±1.1. Meetings are rarer than replies, and a meeting rate needs far more people behind it before it settles. And counts belong next to rates. “5% (20 of 400)” shows the reader how fragile the number is, where “5%” on its own hides it.

The approximation is rough and works poorly with only a handful of replies. In that case, wait for more data. A more precise formula won’t rescue a sample of six.

The comparison that pays off is between your own segments, measured the same way over the same period. It tells you where next month’s effort should go. An illustrative example:

Operations in manufacturing looks stronger than finance, and the two ranges overlap only a little. That makes it a reasonable segment to lean into while you keep confirming it. Logistics shows the highest rate, but on just 100 people, and its range is wide enough to cover both of the others. Give it another round before drawing a conclusion.

Test one change at a time. If a new segment, a new message and a new sending domain all start on the same Monday, you won’t know which of them moved the number. The segments themselves should come from your ideal customer profile, which lets each result feed straight back into who you target next.

A baseline describes the conditions it was measured under. When those change, start a fresh one instead of comparing across the break. Re-baseline when you:

• Change your ideal customer profile or move into a new market
• Change your offer, pricing model or the main problem you lead with
• Add or replace sending domains or inboxes
• Add a channel, such as LinkedIn or WhatsApp, to an email-only sequence
• Change a metric definition or the tool that records it

Keep a change log beside the baseline with three columns: date, what changed, why. After a few quarters it explains your numbers better than any published benchmark can, because it is built from your list and your offer.

Don’t let the baseline turn into a ceiling. It exists to make improvement visible. For most teams the next place to look is the gap between a positive reply and a held meeting, which the guide to reply-to-meeting conversion takes apart stage by stage.