Referral Program Email Subject Line A/B Test
Referral emails live or die on the open. If recipients never see your invite-a-friend ask, your reward incentive never gets a chance to work. This guide shows how to set up a referral program email subject line A/B test with marginal, so you can find the line that drives shares — driven entirely from your AI agent.
Why subject lines matter in referral campaigns
A referral program asks customers to do something for you — recommend your product to people they know. That favor only happens after the open, which makes the subject line the single highest-leverage thing to test in this lifecycle email.
Referral subject lines tend to cluster around a few angles: the reward ("Give $20, get $20"), the social proof ("Your friends will thank you"), and the urgency ("Your referral bonus expires Friday"). An email subject line test removes the guesswork about which framing your audience actually responds to.
- Reward-forward: leads with the dollar amount or credit
- Friendship-forward: emphasizes helping people the recipient cares about
- Status-forward: positions the referrer as an insider or VIP
- Curiosity-forward: hints at the offer without revealing it
Running the test with marginal's email MCP server
marginal is a hosted email MCP server that your AI agent connects to over the Model Context Protocol. Instead of clicking through a dashboard, you describe the campaign in plain language and the agent calls marginal's tools to build and run the experiment.
For a referral program email subject line A/B test, the flow uses four MCP tools in sequence. Each maps to a concrete step, so the agent can move from idea to results without you leaving your editor.
- generate_variants — draft several referral subject lines across reward, friendship, and urgency angles
- launch_test — split your audience and send the managed test
- get_results — pull open and click tracking with within-test lift metrics
- recommend_next — get a data-backed suggestion for the next round
Designing a referral test that produces a clear winner
Keep each variant testing one idea. If one subject line changes the reward amount and the tone at the same time, you won't know which lever moved opens. Hold the body content constant so the subject line is the only difference across the split.
Because referrals are a lifecycle email rather than a one-off blast, the goal is a repeatable winner you can apply to every future cohort. Read both opens and clicks: a curiosity line might win opens but lose clicks if it overpromises the reward.
- Test 3–4 subject line variants per round to keep splits statistically meaningful
- Match send timing across variants to avoid time-of-day bias
- Let recommend_next inform the next angle instead of restarting from scratch
- Watch click lift, not just open lift, to confirm the line attracts real referrers
Connect once, iterate from any agent
marginal works with the MCP clients developers already use — Cursor, Claude Code, Claude Desktop, Windsurf, Codex, Cline, Continue, and Zed. Point any of them at https://marginal.sh/mcp, authenticate with your Bearer API key, and your AI agent email marketing workflow can launch a subject line A/B test on demand.
The free tier covers 100 experiments per month, which is plenty for iterating on a referral program across multiple cohorts. See the docs at https://marginal.sh/docs/ for tool schemas and examples.
- Hosted — no infrastructure to run or maintain
- Bearer API key auth with keys like marg_live_...
- Same MCP tools available across every supported client
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