Referral Program Email Tests for Finance Apps
Referral loops are the cheapest growth channel for finance apps — but the invite email decides whether anyone forwards it. This guide shows how to run a finance apps referral program email test using marginal, the hosted email MCP server your AI agent can call directly.
Why referral emails are worth A/B testing
In finance apps, referral programs ride on trust and incentive clarity. A subject line that buries the reward, or copy that sounds too pitchy, kills the forward. Small wording changes move the needle on opens, link clicks, and ultimately invites sent.
An email A/B test removes the guesswork. Instead of debating whether "Get $25 when a friend joins" beats "Your friends should see this," you ship both and let real opens and clicks decide.
- Reward framing: dollar amount vs. fee waiver vs. social proof
- Urgency vs. evergreen invite language
- Personal sender name vs. brand sender for trust
- CTA wording: "Invite a friend" vs. "Share your link"
Running the test with marginal
marginal exposes four MCP tools your AI agent uses end to end: generate_variants drafts subject-line and copy options, launch_test sends them as a managed split, get_results returns open and click metrics with within-test lift, and recommend_next suggests the follow-up experiment.
Because marginal is hosted at https://marginal.sh/mcp, there's no infrastructure to stand up. Authenticate with a Bearer API key (marg_live_...) and your agent can launch a subject line A/B test for your referral campaign in one conversation.
- generate_variants — produce referral subject line and body options
- launch_test — managed send with automatic split assignment
- get_results — open/click tracking plus within-test lift
- recommend_next — data-driven next experiment to try
A practical test plan for finance apps
Start narrow. For a referral program, test one variable at a time so you can attribute the lift cleanly. Subject lines are the highest-leverage place to begin because they gate every downstream metric.
Once you find a winning subject, let recommend_next move you to body copy and CTA tests. Across a quarter, this compounds: a stronger invite email means more shares per active user.
- Round 1: subject line A/B test on reward framing
- Round 2: CTA copy and button placement
- Round 3: sender identity and preview text
- Track click-through to the invite link as your primary metric
AI agent email marketing without leaving your editor
marginal works as an email MCP server inside the tools developers already use — Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, and OpenAI Codex. Your AI agent drives the whole loop, so referral tests live next to the code that powers your finance app.
The free tier covers 100 experiments per month, which is plenty to iterate on a referral campaign before you commit. Read the full tool reference in the docs at https://marginal.sh/docs/ and point your client at https://marginal.sh/mcp to start.
- Hosted — no self-hosted server to maintain
- 100 experiments/month on the free tier
- Works across major AI coding clients
- Endpoint: https://marginal.sh/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