A cold email campaign can get a 5% reply rate and still be a disaster. If most of those replies are unsubscribes and "not a fit", the campaign is burning sending domains for nothing, and the dashboard still looks healthy.

Reply rate tells you people are paying attention. Positive reply rate tells you they want what you sell, which makes it the number that predicts revenue. Working it out means reading every reply, and that is the step most teams skip.

What it does

You give it the send, bounce, and reply counts plus the text of the replies, and it answers in one reply. If you have not got the data together, it tells you exactly which numbers to export.

It sorts the first reply from each lead into one of nine buckets, calculates positive reply rate, and hands you a report in a fixed order: the number, the interested replies to answer now, the referrals to follow up, the risk signals to watch, and the full breakdown. Any reply it was unsure about comes with its reasoning, so you can overrule it.

What it catches

• A reply rate inflated by bad replies. Unsubscribes and spam-trap replies push reply rate up while the campaign fails.

• Scoring too early. The campaign should have run at least 14 days before the sample is worth trusting.

• Unfair comparisons. Two campaigns compared on different cutoff dates are not comparable. The prompt uses one date for both.

• Counting the wrong message. Only the first reply per lead is scored. Later messages are the conversation, not the signal.

• Auto-replies in the maths. Out-of-office replies and bounces are removed from the count, because they are not real replies.

• Trusting the platform's labels. Tags like "interested" are a first pass. Some replies carry none, and the tags do not match the buckets one to one.

• Borrowed benchmarks. Below 1% overall reply rate means something is broken, 1 to 5% is normal, and above 5% is excellent. Positive reply rate is judged against your own past campaigns.

• Slow follow-up. Interested replies should get an answer within minutes, not hours, and referrals within a day.

How to use it

Paste it into any chat. Copy the prompt below into ChatGPT, Claude, or any other assistant, then paste in your counts and reply text.

Add it to a Claude Project. Upload the prompt as a project file. If you connect Apollo, it pulls the send, bounce, and reply counts itself, and asks you only for reply text it cannot reach. Setup: the Claude setup guide.

The prompt

Copy the whole prompt below, from the first line to the last.

You are scoring the replies to my cold outbound campaign. Reply rate tells me people are paying attention. Positive reply rate tells me they want what I sell, and it is the number that predicts revenue. A campaign can get a 5% reply rate and still be a disaster if most of those replies are unsubscribes and "not a fit", because I am burning domains for nothing. Answer in one reply once you have the data.

If my sending platform is connected as a tool, pull the send, bounce, and reply counts yourself, and the replies if you can reach them. Reading is fine without asking. If nothing is connected, tell me exactly which numbers to export and ask me to paste the reply text.

Before scoring, check the sample. The campaign should have run at least 14 days; before that the sample is too small to trust. When comparing two campaigns, use the same cutoff date for both.

The platform's own reply labels are a first pass, not the verdict. Some platforms tag replies as "interested" or "question". Some replies carry no tag, and the tags do not map one to one onto the buckets below. Use them to sort the queue, then classify every reply yourself. Many platforms return my sent emails but not the prospect's reply text, so if you cannot see the words, ask me for them. The value here is the classification, not the fetch.

Classify every reply into one bucket. Classify only the first reply per lead. Later messages are the conversation, not the signal.

Bucket	Meaning	Positive?
Interested	"Yes, tell me more", or booked	Yes
Soft	"Send info", "reach out in Q3"	Yes
Referral	"Not me, talk to X"	Yes, high value
Neutral	A question, no commitment	No
Not now	"Not right now"	No
Not a fit	"Not a fit"	No
Hostile	Angry, a complaint, a report	No, track as risk
Unsubscribe	Opt-out	No
Out of office or bounce	Auto-reply or delivery failure	Remove from the denominator

Positive reply rate = (interested + soft + referral) / total sent, with out-of-office replies and bounces excluded from the count. They are not real replies.

Reading the number. Do not borrow an industry benchmark. For overall reply rate: below 1% something is broken, 1 to 5% is normal, near 5% is good, above 5% is excellent. Positive reply rate is a stricter cut of that, so judge it against my own past campaigns. Hostile replies and a high unsubscribe rate usually mean bad targeting, not bad luck.

Mistakes to catch:

- Trusting reply rate alone. Unsubscribes and spam-trap replies inflate it while the campaign fails.

- Scoring later messages in a thread instead of the first reply.

- Sitting on interested replies. Once someone raises a hand, speed is the whole game.

What to hand me at the end, in this order:

The number: positive reply rate, positive replies as a share of real replies, and the send count behind both. Act now: every interested reply, with the lead and a one-line summary. Tell me to respond within minutes, not hours, because a fast reply converts far better than a slow one. Follow up: every referral, with the named person. Reach them within a day and mention who sent me. Watch: the hostile count, the unsubscribe rate, and what they suggest about the targeting. Full breakdown: the count in every bucket, and any reply you were unsure how to classify, with your reasoning, so I can overrule it.

Method adapted from the positive reply scoring skill in Apollo Operator, a free, open-source headless GTM toolkit by Creatop: github.com/creatop-gtm/apollo-operator

Part of the Apollo Operator prompt pack

This is one of 19 free prompts from Apollo Operator, the open-source headless GTM toolkit Creatop builds and runs on its own campaigns. Get the full pack here: the Apollo Operator prompt pack.