From the blog

Positive reply rate: what counts as positive

A reply is positive when it comes from the person you targeted, they have agreed to a next step of any size, and that step is available now. All three, or it is something else. Below: the eight replies that decide it, and why six tools publish six different versions of this number.

By Rahul Bageria, co-founder · Updated August 2026 · 12 min read

Share

The eight cases

The eight replies that decide the number

All eight land in a normal month, and almost every page on this topic stops before it reaches them. Verdict, failing gate, and what each one indicts.

The reply	Positive?	Gate that fails	What it indicts
Send me the info	Yes	None	Nothing. The smallest positive there is, and still a positive.
Not my area, talk to Priya	No here, yes for the account	Same person	Targeting, off by one role. Open a new record.
Not now, ask me in Q3	No	Still now	Timing. The usual way this number inflates.
Out of office naming a successor	No, and not a reply	Same person, nobody spoke	Your data. A free name change, verified.
What does it cost?	Yes	None	Nothing yet. The reply the tools split on.
Not for us, but we do have this other problem	No	Still willing, on this offer	Your wedge. Route it to the offer test.
An autoresponder	No, and not a reply	Nobody spoke	Your instrumentation, if it reached the numerator.
Take me off your list. Best of luck!	No, and action it today	Still willing	Nothing about the offer. The politeness is the trap.

Sources, each from the vendor's own material, August 2026: Saleshandy counts requesting more information as interested; lemlist lists asking for details as positive; Apollo's reply classifier (2024) files a pricing question under Follow up question; Woodpecker alone has a third bucket, Maybe later.

The framework

Three gates, asked in that order

Two people disagree about a reply because they are scoring warmth. Score what changed in the pipeline instead.

Definition

A positive reply clears three gates: it comes from the person you targeted, they have agreed to a next step of any size, and that step is available this quarter.

The one-sentence version · glossary entry

All three yes

•Count it once, against the person, never the message
•Size does not matter: a question, a document, a call
•Send me the info qualifies, and so does what does it cost
•Freeze the verdict when the campaign closes

Any one no

•Not positive, and the gate that failed is the diagnosis
•Same person fails: a targeting miss worth a new record
•Still willing fails: an offer problem, or a suppression job
•Still now fails: a timing problem with a date on it

The five reply buckets in our offer guide tell you what to fix. The three gates tell you what to count.

Reply rate counts everyone who answered. Positive reply rate counts everyone who is still coming.

One of our own guides files "Sounds interesting, send more info" under bad data in a nice suit, and it is right: as evidence about your market, a compliment predicts nothing.

This page asks the narrower question, whether it belongs on top of a pipeline metric, and there it counts. Both hold, and the first is worked through in what a reply can prove about your market.

The hard four

The four replies everyone gets wrong

Four cases produce nearly every argument about this number, and each has a documented tool behavior behind it.

Four cases, four gates

1

Same person

The referral is a win and not a positive

Positive for the account, not for this contact, so open a new record and start again. Woodpecker pauses the original prospect when a reply comes from a different address, then files that address as BLACKLISTED.

2

Still now

Ask me in Q3 is a date, not a yes

It fails one gate and only one, which is why it survives so long in a spreadsheet. Stamp the date and leave. Woodpecker alone gives it a bucket, Maybe later; lemlist offers two permanent labels.

3

Still willing

The no that opens a different door

The most flattering reply you will get, and still a no on this offer. It is also the most useful thing in the pile. Route it to the next offer test, not to the numerator.

4

Still willing

The polite unsubscribe

It reads warm and it is an opt-out, so action it the same day. Woodpecker gives opt-outs their own prospect status, separate from replies. Getting this wrong is a compliance problem, not a reporting one.

The comparison

Six tools, six different fractions

Every row comes from the tool's own product documentation rather than its benchmark posts, because on this metric those two do not agree. Where the bottom reads total replies, it inherits whatever that product counts as a reply.

Tool	What it publishes	Top of the fraction	Bottom of the fraction	Who decides
Saleshandy	Positive Reply Rate	Interested replies	Total emails delivered	Its AI, into five fixed classes
Instantly	Positive reply rate, on its blog	Positive replies	Total replies	Its AI, on labels you describe
Smartlead	Positive Reply Rate	Replies with a positive label	Total replies	Its AI, on a prompt you write
lemlist	Interested, as a count	Leads marked interested	Leads, each counted once	You or its AI, permanently
Apollo	Interested	Contacts marked interested	Total emails sent	Its own reply classifier
HubSpot	Nothing	Not published	Not published	Nobody. Reply rate, then meeting rate

Sources, all vendor product documentation, checked August 2026: Saleshandy's cold email statistics report (53.1 million emails, January to June 2026); Instantly on calculating reply rates (19 October 2025); Smartlead's dashboard analytics article (undated); lemlist on campaign stats; Apollo's Report on Sequences (12 August 2026, apollo.io help center); HubSpot on sequence performance.

Apollo's is not a fraction of one thing: contacts on top, emails sent underneath, so a longer sequence lowers it with nothing changing in the market. All six publish something, which is more than the category usually does. No two publish the same thing.

The receipts

Three of them cannot agree with themselves

The harder version of the finding is not six tools differing from each other. Three of the six publish two definitions of their own metric, on two properties. Here are the two clearest.

What they publish in one place

✕Instantly: positive replies ÷ total replies × 100, three times on one page
✕Smartlead: (positive replies ÷ total replies) × 100, in its dashboard docs

What the same company publishes elsewhere

✓Instantly: no rate in the product at all, only Opportunities, a lifetime count
✓Smartlead: no positive reply rate anywhere in its report on more than 850 million emails

Instantly's Opportunities counter keeps every lead ever marked positive, even after the status turns negative. A count that cannot go down is not a metric.

None of this makes them careless. A number's name does not tell you what was measured. Smartlead's own report stops there: only a reply shows genuine interest.

Sources: Instantly, Opportunities in campaign analytics; Smartlead, The State of Cold Email 2026, H1 (more than 850 million emails, January to June 2026).

Want the replies worked and the rule written down, by the two of us?

Book a Fit Check

The denominator

What goes on the bottom of the fraction

Per contact, counted once, best verdict they ever gave you, frozen when the campaign closes. A person is the unit you are trying to reach.

One campaign, three conventions

1,000 contacts, four steps, 4,000 emails, 30 replies, 9 of them clearing all three gates. Per contact that is 0.9%. Per email sent, 0.23%. As a share of replies, the formula Instantly and Smartlead both publish, 30%. The largest reads about 133 times the smallest.

Which denominator to use in general, and when a rate is a finding rather than noise, is worked through in our outbound market learning guide.

The limits

What it predicts, and what it does not

It beats reply rate by exactly one step, and that step is smaller than it looks.

Four honest limits

Closer

One step nearer a meeting, not a promise

Saleshandy puts the interested share among its strongest campaigns between 25% and 50% of all replies. Even at the top of a 53.1 million email sample, the split moves about twofold.

Not fit

It says nothing about whether they can buy

A person willing to take a next step is not a person with a budget, a problem and a mandate. Read it next to the qualified meeting, never instead of it.

Not stable

It does not survive a change in your own rule

The boundary moves when the offer moves. Against a demo, send me the info is a soft yes; against a benchmark offer it is the whole conversion. Keep the rule beside the offer.

Not portable

Comparable to your own number and nothing else

Six tools, six fractions, three holding two positions at once. A benchmark from one cannot be set against a number from another, or against yours.

Which kind of reply arrives at all is decided mostly by the ask at the end of the email.

The scoring

Score the gates, not the warmth

Both replies below read friendly. One is an opt-out, one is a targeting correction, and neither is a positive.

Don't

Score how it made you feel

thanks, this is really interesting. please take me off the list though

✕Reads positive to anything scoring sentiment
✕Fails still willing on the second clause
✕Filing it wrong is a compliance problem

Do

Ask the three questions in order

not me, priya owns this now. copying her in

✓Fails same person, so not a positive here
✓Positive for the account: open a new record
✓The gate that failed names the fix

An illustrative walkthrough of the mechanism, not a specific client result. We report real numbers only when they are real. The example replies are invented, and no reply rate is attached to them.

The ops

Counting it by hand, with two people and no dashboard

At twenty replies a month the replies are the evidence and the rate is mostly noise. Ten minutes on a Friday is the whole system.

The Friday ten minutes

Three habits and one argument

Write the rule at the top of the sheet

Date it, and read it again before you touch a reply. It is there to keep your week out of the verdict.

One row per contact, not per message

Best verdict they ever gave you, frozen when the campaign closes, so last quarter stays readable.

Two people, the same twenty replies

Score separately, compare monthly. You will disagree on three or four, and that argument beats the number.

Operator note

How we run it

We classify by hand when the angle is new or the run is a controlled experiment, custom messages for every prospect. When an angle or an ICP is expanding on a process we already have, everything we know goes into one source-of-truth document, an agent does the first-level classification against it, and we keep reviewing and improving that document.

RB

Rahul Bageria

Co-founder, Real Good GTM

That is this page's argument one step further: the rule has to be written precisely enough that an agent reaches the verdict a person would, with people reviewing it continuously, and the reason we work that way is speed, on the classification and on the first reply back.

Key takeaways

4 points

1 Write the rule down before you touch the replies.
2 One row per contact, best verdict, frozen at campaign close.
3 The gate that fails is the diagnosis, not the label.
4 Auto-tag for triage in month one, not for the metric.

Once the sorting is past a Friday habit, the reply layer we run for teams is the same rule with a human on the hot ones.

The trap

The one way to cheat this number

Any metric that decides something gets gamed, and this one is unusually cheap to game.

A subject line implying a conversation you never had, or an ask so vague that agreeing costs nothing. Both lift the count, and the meetings are with people answering something else.

!

Caution

A misleading email scores beautifully

Hunter, a vendor in this category, puts it plainly in its own cold email guide: you will not do anyone a favor by maximizing your positive reply rate through misleading your recipients. The number goes up, the domain and the market both pay for it, and neither shows up in the same report.

Do this instead

Read the positive rate next to meetings that were actually held. A gamed number separates from that one inside a month.

Failure modes

Four ways the count goes wrong

None of these is exotic, and none of them announces itself.

Politeness scored as interest

The warmest reply in the pile is often an unsubscribe. Any rule that reads tone rather than commitment files it as a win, every month, in the same direction.

The rule that quietly drifted

Nobody wrote it down, so it moved with the month and the mood. Two quarters that were never scored the same way are not a trend.

Counting messages, not people

One person who answered step two and step four arrives as two positives. lemlist counts each lead once whatever the sequence does, and that is the shape to copy.

Trusting the auto-tag in month one

Apollo's classifier reports 90% accuracy across all categories. Nobody has shown how often an interested reply becomes a meeting, so treat the label as a hint, not a forecast.

Operator note

Learned the hard way

The real enemy of this number is the person classifying on a Friday afternoon after a bad week. All four of these are that person with a tool attached, and the rule on the sheet is the only thing that has ever fixed it for me.

KM

Kshitij Maheshwari

Co-founder, Real Good GTM

FAQ

Questions founders ask

Does send me more info count as a positive reply?

Yes. It is the smallest positive there is and it is still a positive: same person, a next step agreed, available now. Saleshandy's definition counts requesting more information as interested, and lemlist lists asking for details among its positive examples. Whether it is good evidence about your market is a different question.

Should I count a reply that asks me to come back next quarter?

No. It fails the third gate, still now, so it is a dated re-entry rather than a win. Stamp the date, leave, and count it if the date produces something. Woodpecker is the only major tool with a bucket for it, Maybe later; on lemlist you choose between two permanent labels.

Should a referral count as a positive reply?

Not for that contact. It fails the first gate, same person, and it is genuinely positive for the account, so open a new record and start again. It also arrives from a different address, which most tools cannot attribute. Woodpecker pauses the original prospect and files the new address as blacklisted.

What is a good positive reply rate?

There is no answer that transfers, and that is the honest one. Six tools publish six different fractions and three publish two definitions each, so a benchmark from one cannot be set against a number from another. Set your own baseline over four weeks with one written rule, then compare only against yourself.

Can I let the tool classify replies for me?

For triage, yes. For a metric you will make decisions on, not in the first month. Apollo's classifier reports 90% accuracy across all categories, but nobody has shown how often an interested reply becomes a meeting, so treat the label as a hint, not a forecast. Hunter, a vendor, recommends going through replies by hand for that reason.

Why do two tools show different positive reply rates for one campaign?

Because they are computing different fractions. One divides interested replies by delivered emails, two divide positive replies by all replies, one counts each lead once on a permanent flag, and one puts contacts on top of emails sent. The sixth has no positive metric at all.

About the author

Rahul Bageria

Co-founder of Real Good GTM. He has been the first business hire and Chief of Staff at seed-stage B2B startups, building outbound pipeline before any playbook existed. This post comes from sorting replies at seed volume, where twenty a month means the rule you wrote down matters more than the rate it produces.

Connect on LinkedIn

Keep going