Positive Reply Scoring

---
name: positive-reply-scoring
description: "Classify replies and compute positive reply rate, the metric that predicts revenue. Use to tell whether a campaign is actually working, not just getting replies."
---

Read [runtime rules](../operator-context/references/runtime.md) once per conversation before this skill. **Without an authorized local terminal and authenticated Apollo CLI, use [connector-only instructions](connector-only.md) instead of the local procedures below.** For outbound work, load [Operator Context](../operator-context/SKILL.md) first, using its connector-only path when appropriate.

# Positive reply scoring (iterate)

Reply rate tells you people are paying attention. Positive reply rate tells you they want what you are selling. This computes the second, because it is the number that actually predicts revenue.

A campaign can get a 5% reply rate and still be a disaster: if most of those replies are "unsubscribe" and "not a fit," you are burning domains for nothing.

```
positive reply rate = positive replies / total sent
```

## When to use

- After a campaign has run at least 14 days (before that, the sample is too small to trust).
- Weekly, as the Wednesday task in `weekly-rhythm`.
- When comparing two campaigns in an experiment (use the same cutoff for both).

## Getting the replies

For the counts around the replies (sent, bounced, and reply rate by sequence, step, and variant), run `campaign-reporting` first. This skill is the part a report cannot do: deciding which replies are actually positive.

Pull reply data from Apollo, which carries more than it used to (verified 2026-09-14):

- **Which emails got a reply, with Apollo's own read of it.** `apollo emails search --stats replied` (CLI) returns each replied message with a `reply_class`, such as `willing_to_meet` or `follow_up_question`, or no class at all. On MCP, `apollo_emailer_messages_search` gives per-sequence reply counts and per-message `replied` flags, and `apollo_analytics_sync_report` gives campaign-level numbers.
- **Treat `reply_class` as a first pass, not the verdict.** It is Apollo's classification, some replies carry none, and its classes do not map one to one onto the buckets below. Use it to sort the queue, then classify yourself.
- **The reply text itself is not reachable.** `apollo_emailer_messages_get_content` returns your own sent emails and explicitly excludes customer replies. So ask the operator to paste or export the reply bodies: the value of this skill is the classification and the metric, not the fetch.

## Classify every reply

One bucket each. Classify only the **first** reply per lead (later messages are the conversation, not the signal).

| Label | Meaning | Positive? |
|---|---|---|
| interested | "yes, tell me more," 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-fit | "not a fit" | no |
| hostile | angry, complaint, report | no (track as risk) |
| unsubscribe | opt-out | no |
| ooo / bounce | auto-reply / delivery fail | exclude from the denominator |

Positive reply rate = (interested + soft + referral) / total sent. Exclude OOO and bounces from the denominator, they are not real replies.

## Report

- **The number:** positive reply rate, and positive as a share of real replies.
- **Act now:** list the interested replies. Respond to these within minutes, not hours. A fast reply converts far better than a slow one.
- **Follow up:** referrals, reach the named person within a day and mention who sent you.
- **Watch:** hostile replies and the unsubscribe rate. Elevated hostility usually means bad targeting, not bad luck.

## Reading the number

No borrowed benchmarks. Use the reply-rate reality from `../operator-context/references/outbound-principles.md`: below 1% something is broken, 1 to 5% is normal, near 5% is good, above 5% is excellent. Positive reply rate is the stricter cut of that, so judge it against your own past campaigns, not an industry chart.

## Common mistakes

- Trusting reply rate alone. Spam-trap and unsubscribe replies inflate it while the campaign fails.
- Scoring later messages instead of the first reply.
- Sitting on interested replies. Speed is the whole game once someone raises a hand.