How to run a win back email A/B test
Lapsed subscribers respond to different language than your active list. A focused win back email A/B test tells you which subject line actually pulls dormant contacts back — and marginal lets an AI agent run that test end to end through a single MCP endpoint.
Why win back emails deserve their own test
A win back campaign targets people who stopped opening or clicking weeks or months ago. The tone, urgency, and incentive that work for engaged readers rarely move this segment, so reusing your standard subject lines wastes the send. Treat re-engagement as its own experiment.
The fastest signal is the open rate driven by the subject line, since a dormant contact has to choose to open before anything else matters. That makes a subject line A/B test the highest-leverage thing to run first on a win back flow.
- Test curiosity vs. direct "we miss you" framing
- Compare an incentive subject line against a no-discount version
- Try a question-style subject against a statement
- Measure open lift before optimizing body copy
Running the test through marginal's email MCP server
marginal is a hosted email MCP server that exposes its workflow as tools your AI agent can call. There is nothing to deploy — point your client at https://marginal.sh/mcp, add a Bearer API key, and the agent can draft, send, and read results without you leaving the editor.
A typical win back email A/B test maps directly onto four tools, so an agent can move from idea to recommendation in one conversation.
- generate_variants — produce subject line options tuned for lapsed subscribers
- launch_test — send the variants as a managed split with tracking
- get_results — pull open and click data plus within-test lift
- recommend_next — get a data-backed suggestion for the follow-up send
An AI agent workflow that picks the winner
AI agent email marketing works best when the model can both write and measure. With marginal connected, you can ask your agent to generate five re-engagement subject lines, launch the split, wait for results, and then call recommend_next to decide the second touch — all grounded in real open and click numbers rather than guesses.
Because lift is computed within the test, the agent compares variants against each other on the same audience and timeframe, which keeps the win back decision honest.
- Ask the agent to draft variants for a 90-day-inactive segment
- Launch a managed send and let tracking accrue opens and clicks
- Review within-test lift to identify the strongest subject line
- Use recommend_next to plan the second win back attempt
Getting started
marginal works with Cursor, Claude Desktop, Claude Code, OpenAI Codex, Windsurf, Cline, Continue, and Zed, so you can run a subject line A/B test from whatever environment you already use. The free tier covers 100 experiments per month — enough to validate a win back flow before scaling.
See the docs at https://marginal.sh/docs/ for tool schemas and client setup, then connect to the endpoint and run your first re-engagement test.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Free tier: 100 experiments/month
- Registry: sh.marginal/mcp
Get started with marginal
marginal is a hosted email marketing MCP server at marginal.sh. Sign up free, create an API key, and connect https://marginal.sh/mcp from Cursor, Codex, or Claude Desktop.
- 100 experiments/month on the free tier
- Four MCP tools: generate_variants, launch_test, get_results, recommend_next
- Listed in the MCP Registry as sh.marginal/mcp