Testing Referral Program Emails for SaaS Products
Referral emails only work if people open and act on them. This guide shows how to run a SaaS referral program email test with marginal, the hosted email MCP server that lets your AI agent generate variants, launch the test, and report on lift automatically.
Why referral emails deserve their own test plan
Referral asks are a different beast from feature announcements or onboarding nudges. You're asking an existing customer to spend their own social capital, so the framing of the ask — and especially the subject line — drives most of the variance in opens and shares.
For SaaS teams, small lifts compound: a referral program that converts even a few percent more invites can shift acquisition cost meaningfully. That makes the referral program one of the highest-leverage places to run a disciplined email A/B test.
- Test the incentive framing: account credit vs. cash vs. an upgraded tier
- Compare a personal, founder-style note against a polished template
- Try urgency or scarcity copy against a low-pressure invitation
- Vary the call to action between 'invite a teammate' and 'share your link'
Running the subject line A/B test with an AI agent
marginal exposes four MCP tools your agent calls in sequence: generate_variants drafts candidate subject lines, launch_test sends them across your audience split, get_results returns open and click metrics, and recommend_next suggests the follow-up experiment.
Because it runs as an email MCP server, you describe the goal in plain language inside Cursor, Claude Code, Windsurf, or any supported client, and the agent handles the mechanics of the subject line A/B test for you.
- Ask the agent for five referral subject line variants targeting power users
- Launch the test as a managed send — no separate sending infrastructure
- Read within-test lift metrics to see which variant outperformed
- Let recommend_next propose a body-copy follow-up once a winner emerges
What to measure for referral campaigns
Opens tell you whether the subject line earned attention, but referral success ultimately lives in the click — the moment a customer heads to your referral page or copies their invite link. marginal tracks both opens and clicks within each test so you can judge variants on the metric that matters.
Treat each test as a learning loop rather than a one-off. Lift metrics from one referral send inform the next, and the free tier covers 100 experiments per month, which is plenty for iterating through subject lines, incentives, and send timing.
- Open rate to validate the subject line A/B test
- Click-through to the referral landing page as the real signal
- Within-test lift so comparisons stay apples-to-apples
- Iteration cadence: one focused variable per experiment
Fitting marginal into your AI agent email marketing workflow
marginal is hosted, so there's no infrastructure to stand up — you connect an MCP client at https://marginal.sh/mcp, authenticate with a Bearer API key, and start sending. AI agent email marketing for SaaS becomes a conversation: describe the referral campaign, and the agent runs the test end to end.
Pair this with your existing growth stack. The agent can draft, launch, and evaluate a SaaS referral program email test in one session, then surface a ranked recommendation you can ship. See the docs at https://marginal.sh/docs/ for tool details.
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