Outbound Metrics and Benchmarks: The Metric Chain, Honest Ranges, and Your Own Baseline

By Laurel · Updated Oct 8, 2026

Outbound metrics trace a message from delivery to pipeline: delivered, bounced, spam-reported, replied, positive reply, meeting booked, meeting held, opportunity, and pipeline value. Each needs a stated denominator. Published benchmarks come from vendors describing their own users or clients, so treat them as rough ranges and build your own baseline per segment and channel.

A reply rate of 4% means one thing in your sequencing tool, another in a spreadsheet, and something else in a vendor's benchmark report. Until you know what was counted and what it was divided by, comparing it is guesswork.

This article covers measurement itself. The arithmetic that turns rates into a plan is in outbound pipeline math, per-rep targets are in BDR/SDR metrics and quota math, and the whole motion is in the outbound sales playbook.

Key Facts

• Instantly's 2026 benchmark of its own platform users' campaigns (1 January to 18 December 2025) reports a 3.43% average reply rate and 5.5% or more for the top quartile, counting all replies over emails sent (vendor data).
• Belkins, an outbound agency, reports a 0.45% average reply rate across 7,530,489 emails in its 2025 client campaigns, excluding auto-replies and bounce notices (vendor data).
• Apple says Mail Privacy Protection helps users prevent senders from knowing when they open an email, and Apple's clients were 62.26% of tracked opens in Litmus's July 2026 data.
• Google asks senders to keep user-reported spam below 0.10% and never reach 0.30%.
• The Bridge Group's 2025 survey of 351 B2B companies found 60% of SDRs at quota, the lowest in the study's history.

The Metric Chain From Delivery to Pipeline

Each link in the chain is a fraction of the one before. Deliverability metrics divide by messages, because mailbox providers count messages. Outcome metrics divide by contacts, because a four-step sequence sends four messages to the same person.

Open rate is missing on purpose; the section on what you can know explains why.

Read Each Link Two Ways

Stage-to-stage rates, such as held divided by booked, show where the chain leaks. Cumulative rates, such as held meetings per 1,000 contacts, show what the motion yields. Report both; conversion rate analysis covers the method.

Write the Definitions Down

Reply, positive reply, booked and held each need a one-sentence definition before anyone reports a number. Sort replies with one set of labels, as handling replies and booking meetings lays out, and decide in writing where "send me some info" goes. For opportunity, borrow the standard your pipeline already enforces, such as opportunity entry criteria. A definition that changes mid-quarter breaks every trend line.

Why Belkins' 0.45% and Instantly's 3.43% Can Both Be True

The two most-quoted recent figures sit 7.6 times apart, and both describe real 2025 campaigns. Definitions explain part of the gap, though the reports don't let you separate that from differences in audience. You can still see how far definitions alone move a number.

One Campaign, Six Reply Rates

Take an invented campaign: 1,000 contacts, 3,600 emails sent, 3,500 delivered. It drew 100 inbound messages, 30 of them auto-replies. The other 70 came from 50 different people, and 15 were interested.

Every row is arithmetically correct. The spread from 1.39% to 5.00% is a factor of 3.6 with no change to list, offer or copy. That factor is the sequence length: 3.6 emails per contact.

What the Two Reports Count

• Instantly divides all replies, follow-up responses included, by total emails sent. Its data comes from its platform's users.
• Belkins divides unique replies by emails sent, excluding auto-replies and bounce notices. Its data covers 7,530,489 emails in its clients' campaigns, over 5.6 million of them to the US. Belkins says its earlier studies divided by recipients who opened the email, which inflated the rates, so older Belkins figures aren't comparable.

Neither is "the" cold email reply rate. Before comparing a published rate with yours, ask what's in the numerator, what's the denominator, who was sent to, and in which period.

What You Can and Can't Know

Opens Are Unreliable

Apple's Mail Privacy Protection stops senders from knowing when a user opens an email, and Apple's clients made up 62.26% of tracked opens in Litmus's July 2026 data (over 1 billion opens in its customers' email, not cold outreach to your list). A logged open may be a person, a machine or neither. Belkins turned open tracking off across its 2025 campaigns and reports none. Judge copy and targeting on replies.

Delivered Isn't Inbox

A delivered message is one the receiving server accepted. It could still land in the inbox, a promotions tab or spam, and your sending tool can't see which. Google Postmaster Tools' domain spam rate shows the damage after the fact, and placement tests sample a few mailboxes; the cold email deliverability guide covers both. Report delivered as acceptance, not reach.

Replies and Meetings Are Only as Good as the Log

These are countable only if someone records them the same way every time. The person who replies isn't always the person you wrote to: Forrester puts 13 people in the average business buying decision, so a reply can come from a colleague who was forwarded the message. Log the original contact and the replier separately. And set the source on the opportunity when it's created, so "outbound-sourced" isn't reassigned later by whoever builds the report.

Published Benchmarks, With Their Samples

Everything below comes from a vendor or agency describing its own users or clients, apart from the mailbox-provider rules. None is an industry census.

One gap stands out. Neither the Instantly nor the Belkins report publishes a positive reply rate, and appointments appear only as a count. A positive-reply or meeting-rate figure with no stated definition, sample and period isn't a benchmark. Treat one as an unsourced claim.

Use published ranges as a smoke alarm, not a target. A rate a tenth of the range points at the list, the offer or deliverability, and one several times above it usually means auto-replies are in the numerator or the sample is tiny.

Segment Differences Dwarf Small Gains

Belkins' reply rate fell from 0.72% at companies with 0 to 10 employees to 0.22% at companies with 10,000 or more, and from 0.57% for founders and owners to 0.32% for VPs. A team selling to enterprise VPs shouldn't hold itself to 0.45%. Neither report breaks out Southeast Asia, so for Singapore, Malaysia, the Philippines, Indonesia or Vietnam, your own data is the benchmark.

Replies Come Early, Meetings Come Later

Instantly found 58% of replies arrive on the first email. Belkins' per-step reply rate is also highest at step one (0.59%, falling to 0.28% by step six), yet over 53% of its email-sourced meetings came from step 3 or later. A sequence judged on reply rate alone undervalues its follow-ups. Track meetings by the step that produced them, as cold email follow-ups explains.

Why Small Samples Mislead

A team sending to 400 contacts a week in one segment has single-digit counts for anything downstream. The ideal customer profile article's testing section gives the sample sizes needed to tell a 3% segment from a 6% one. Here is what happens below them.

Two Variants, 400 Contacts Each

Two versions of a first line each went to 400 contacts. Version A booked 4 meetings and version B booked 9.

B looks 2.25 times better. But the 95% intervals, the range of true rates consistent with each result, overlap from 1.19% to 2.54%. If both versions had the same true rate (13 meetings in 800 contacts is 1.63%), a gap this wide would still appear about 26% of the time by a two-sided Fisher exact test. That's roughly one time in four, too often to act on.

The Winner of Ten Is Flattered

A simulation shows the second trap. Send ten variants to 200 contacts each, every one with the same true 3% reply rate. Across 5,000 simulated runs, the best-looking variant averaged 5.0%, and in 80% of runs landed between 4.0% and 6.0%. Nothing differed, yet the "winner" looked two-thirds better than the truth. Test fewer variants, fix the sample size in advance, and confirm any winner on fresh contacts.

Intervals Narrow Slowly

The table holds the observed reply rate at 3% and changes only the contacts.

Quadrupling the contacts roughly halves the width. That's why metrics deep in the chain are noisiest: the denominator shrinks at every link. If 6 of 8 booked meetings were held, the show rate is 75%, with an interval from 41% to 93%. Judge upstream metrics weekly, downstream ones on pooled quarters.

Build Your Own Baseline

A baseline is your measured rate for one segment, one channel and one offer, with an interval and a period attached. Every new test, tool or hire gets compared against it.

Pick the Cuts

• Segment. ICP tier and persona, so a fit-plus-signal list isn't blended with a near-fit one.
• Channel. Email, calls, LinkedIn and chat apps have different rates. Never blend them.
• Message version. One row per version, so a copy change shows.
• Market. Country-level, since Belkins' data shows size and seniority move rates.
• Sending domain. For bounce and complaint rates, because reputation attaches to domains.

Every cut divides your volume and widens every interval, so cut no finer than your volume allows.

Group by Send Week and Let Cohorts Mature

A cohort is every contact first messaged in the same week. Measure each later outcome against that cohort, not against whatever happened this week, because outcomes lag. Wait until a cohort's last step has run, plus a week or two, before calling its rates. Opportunities follow the meeting, and closed revenue follows your sales cycle length. Label recent cohorts "immature" rather than reading a dip into them.

Pool Cohorts, Then Compare

Keep one row per cohort, segment and channel, with a column for each link in the chain plus contacts and emails sent. Add rows together for a pooled rate and interval, and keep pooling until the interval is narrow enough to change a decision.

Then compare each new cohort's own interval with the baseline's. If they don't overlap, something changed: the list, the offer, the domain or the market. One cohort outside the range is a prompt to look, not a verdict.

When reply status, meeting outcome and opportunity stage live on one record, this sheet is a report, not a reconciliation job. Rework's Lead Ops captures and routes leads with their source, and its Sales Ops CRM holds stages and amounts.

SDR Metrics: Activity Versus Outcome

The Bridge Group's 2025 report surveys 351 B2B companies. It calls its data observational, and its sample skews toward organizations with engaged sales development leadership.

The report uses its own stage labels, so define yours before comparing. Its note on pipeline is worth copying: the increase reflects higher average selling prices ($44K median in 2024 against $27K in 2022) rather than more meetings. Pipeline value mixes effort with deal size, so it can't be the only measure of an SDR.

How to Use Each Layer

• Activity is an input. It's controllable and fast, which suits diagnosis: low dials explain low conversations. But activity targets with no quality check produce volume, and volume produces complaints. Google's Postmaster Tools reports spam by domain, so one rep's spray counts against everyone's mailbox.
• Outcomes are the score. Count a meeting only when it's held and accepted under your entry criteria.
• Pair volume with quality. Sends with complaints, booked with held, held with accepted.
• Judge on a quarter, not a month. If meetings arrive at random with a mean of 10 a month, about one month in eight lands at 6 or fewer and one in eight at 14 or more, with no change in the rep. The same report's 3.0-month average ramp also means a new SDR's first quarter isn't a fair read.

The First Baseline Will Be Rough

Pick the most recent cohort that has finished its sequence, fill in one row of the sheet, and put an interval beside each rate. Pin the four definitions next to the dashboard. The interval will be wide. That's information too: it tells you how many more weeks of sending it takes before a number is allowed to change anyone's mind.