# Cta Ab Test Tracker

> Run a fair A/B test on the final call-to-action of cold emails and report which version gets more positive replies. Use when the user wants to improve reply rates or asks which email version works better.

- Skill: `syp1xd/cta-ab-test-tracker` (Agent Skill)
- Install (CLI): `npx skillmds@latest add syp1xd/cta-ab-test-tracker`
- Raw SKILL.md: https://api.skillmd.com/api/skills/syp1xd/cta-ab-test-tracker/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: syp1xd (https://skillmd.com/u/syp1xd)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/syp1xd/cta-ab-test-tracker

---


# CTA A/B test tracker

## Design
- Change **only the last ask**. Keep subject, body, offer, sender, and targeting the same.
- Assign A or B by a stable hash of the recipient's company email domain, so the split is about 50/50 and one company always gets one version.
- One prospect gets one version. Never resend to test the other version.
- Example: A = "Could you introduce me to the team that owns this?" B = "Are you the right person to speak to about this?"
- Save `experiment`, `variant`, exact body, and send confirmation with each record.

## Classify replies
`positive` (interest in a call, proposal, profile, or a useful introduction), `negative`, `opt_out`, `bounce`, `automatic` (out-of-office, read receipts), `other`.
Link replies by In-Reply-To / References. Dedupe by reply message ID. Hold uncertain matches.

## Report
Per version: confirmed sends, unique human replies, positive replies, meetings, bounces, opt-outs.
- Denominator = confirmed sends, not attempts.
- Compare after the same 14-day window for every email.
- Report hot and warm leads separately.
- Always show sample sizes. Below about 30 completed sends per version, call results preliminary. Do not declare a winner early.