Win-back subject line tests in Continue
Lapsed subscribers are a hard audience: they already tuned you out once. This guide shows how to run win-back subject line A/B tests directly from Continue using marginal, the hosted email MCP server, so you can find the line that earns a second open.
Why win-back campaigns need their own subject line test
A win-back email isn't a regular newsletter. The recipient hasn't engaged in weeks or months, so the subject line carries more weight than the body — most people decide to re-engage or ignore you before they ever open. Reusing your standard subject formula tends to underperform here.
Running a deliberate subject line A/B test on a re-engagement segment tells you what actually pulls lapsed users back: curiosity, an incentive, a direct 'we miss you,' or a plain account-status nudge. Continue lets you orchestrate that test conversationally while you stay in your codebase.
- Win-back audiences open at lower baseline rates, so subject framing matters more
- Tone matters: apologetic, playful, and transactional lines pull different segments
- Measuring within-test lift keeps you honest about what's working
Generating win-back subject lines in Continue
With marginal connected as an MCP server, Continue's AI agent can call generate_variants to draft several distinct win-back angles in one pass. Describe the segment — for example, users inactive for 90 days — and ask for a spread of approaches rather than minor wording tweaks, so your test covers real strategic differences.
Good win-back subject lines in Continue typically span a few directions you can pit against each other:
- A direct re-engagement line: 'We saved your spot'
- An incentive line: 'Here's 20% to come back'
- A curiosity or update line: 'Here's what you missed'
- A loss-framing line: 'Your account goes quiet soon'
Launching and reading the subject line A/B test
Once you've picked your variants, have Continue call launch_test to send them across your win-back segment. marginal handles the managed send and splits the audience, then tracks opens and clicks so you don't wire up any tracking yourself. There are no Mailchimp or Klaviyo keys to manage — the server is fully hosted.
As results land, get_results returns open and click rates plus within-test lift, and recommend_next suggests the winning subject line or whether to keep testing. The free tier covers 100 experiments a month, which is ample for iterating on a single win-back campaign before you roll out the winner.
- launch_test — managed send with automatic audience split
- get_results — open/click metrics and within-test lift
- recommend_next — pick the winner or queue another round
- Endpoint: https://marginal.sh/mcp · Docs: https://marginal.sh/docs/
Fitting marginal into an AI agent email workflow
Because every step runs through MCP tools, Continue can drive the whole win-back loop as part of AI agent email marketing — drafting variants, launching the subject line test, and reporting lift back to you in chat. You stay in one window instead of bouncing between an ESP dashboard and your editor.
Connect marginal once with your Bearer API key (marg_live_...) and the tools become available to Continue. From there, win-back subject lines, transactional re-engagement nudges, and ongoing subject line A/B tests all run through the same hosted server.
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