Cross-sell email subject line A/B test
Cross-sell campaigns live or die on the subject line — recipients already bought once, so your next message has to feel relevant, not pushy. This guide shows how to run a cross-sell email subject line A/B test with marginal, the hosted email MCP server your AI agent can drive directly.
Why subject lines matter for cross-sell
A cross-sell email reaches someone who has already converted, which means you have proof of intent but also a higher bar for relevance. The subject line decides whether the recommendation gets opened or filed as another upsell blast. Small wording changes — leading with the original purchase, naming a complementary product, or framing the offer as a tip — can move opens meaningfully.
Because cross-sell sits inside a lifecycle email flow, you can isolate the subject line as the single variable and let everything downstream stay fixed. That makes it an ideal candidate for a clean subject line A/B test.
- Reference the prior purchase: "Pairs perfectly with your new boots"
- Lead with benefit, not discount, to avoid promo fatigue
- Test specificity vs. curiosity in the same send
- Keep length short so mobile inboxes show the full line
How marginal runs the test
marginal is a hosted MCP server that exposes a handful of focused tools to your AI agent. You describe the cross-sell context, the agent calls generate_variants for subject line candidates, launch_test to send them across split audiences, and get_results to read open and click metrics with within-test lift. recommend_next then suggests the follow-up experiment.
The endpoint is https://marginal.sh/mcp and authentication is a Bearer API key (marg_live_...). There is nothing to self-host — connect a supported client and start your email subject line test.
- generate_variants — draft cross-sell subject line options
- launch_test — managed send with audience splits
- get_results — open/click tracking and lift metrics
- recommend_next — what to test on the next lifecycle email
A practical cross-sell workflow
Treat the experiment as a loop your AI agent can run end to end. Start from the customer's last order, generate three to five subject line variants that each take a different angle, and launch the subject line A/B test against a representative slice of recent buyers.
Once results land, let the agent read the lift and apply the winner to the full cross-sell segment. Then feed the learning into the next campaign so each lifecycle email compounds on the last.
- Anchor variants to the product the customer already owns
- Send to a holdout split, not the whole list, to measure cleanly
- Promote the winning line and log the result for future tests
- Reuse the framing that won across related cross-sell flows
Connect your AI agent
marginal works with the MCP clients developers already use, including Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, and OpenAI Codex. The free tier covers 100 experiments per month, which is plenty to iterate on cross-sell subject lines before you scale.
Point your client at https://marginal.sh/mcp, add your API key, and your AI agent email marketing setup can launch a cross-sell email subject line A/B test without leaving the editor. Full setup details are in the docs at https://marginal.sh/docs/.
- Hosted endpoint — no infrastructure to manage
- Bearer API key auth (marg_live_...)
- 100 experiments/month on the free tier
- Works across major MCP-capable editors and agents
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