B2B Software Referral Program Email Test
Referral programs only work if the invite email actually gets opened and acted on. This guide shows B2B software teams how to run a referral program email test with marginal — subject line A/B tests, managed sends, and lift metrics driven straight from your AI agent.
Why referral emails need their own A/B test
In B2B software, your referral program competes with a crowded inbox of vendor pitches, product updates, and internal threads. The difference between a 12% and 22% open rate is often a single subject line decision — and that decision shouldn't be a guess.
Referral invites also carry unusual context: the sender is a trusted customer, the incentive matters, and the recipient is frequently a peer at another company. A dedicated email A/B test lets you isolate what actually drives shares versus what you assume drives them.
- Test incentive framing: 'Get $50 credit' vs. 'Give your team a head start'
- Test personalization: referrer's name vs. company name in the subject
- Test urgency vs. value-first phrasing for the referral ask
- Compare plain-text invites against branded HTML for trust signals
Running the test with marginal's email MCP server
marginal is a hosted email MCP server that plugs into AI agents like Cursor, Claude Code, and Windsurf. Instead of switching tools, you describe your referral campaign in chat and the agent calls marginal's tools to generate variants, launch the test, and report results.
The workflow maps to four MCP tools. Your agent uses generate_variants to draft subject line options for the referral invite, launch_test to start a managed send and split traffic, get_results to pull open and click data, and recommend_next to decide which variant to scale.
- generate_variants — produce multiple referral subject lines to A/B test
- launch_test — send and split the campaign across recipients
- get_results — track opens, clicks, and within-test lift
- recommend_next — get the statistically backed winner to roll out
Subject line ideas for B2B referral invites
Use the subject line A/B test to compare distinct strategies rather than minor word swaps — that gives you cleaner, faster signal. Below are angles worth pitting against each other for a B2B software referral program.
- Social proof: 'Sarah at Acme thinks you'd like this'
- Outcome-led: 'Cut onboarding time by half — here's an invite'
- Incentive-forward: 'You and a colleague both get a month free'
- Curiosity: 'A tool your team keeps asking for'
Connect your agent and start free
Point your AI agent at the marginal endpoint at https://marginal.sh/mcp and authenticate with your Bearer API key. The free tier includes 100 experiments per month — enough to iterate through several rounds of referral subject lines before you commit budget to a full send.
Because marginal is hosted, there's nothing to self-host or maintain. Your AI agent email marketing loop stays inside the editor, and you get repeatable, measurable referral tests. See the docs at https://marginal.sh/docs/ to wire up your first test.
- Endpoint: https://marginal.sh/mcp
- Works with Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, and Codex
- Free tier: 100 experiments/month
- Hosted — no infrastructure to run
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