Running a SaaS Cross-Sell Email Test
Cross-sell is where a lot of SaaS revenue hides — existing accounts already trust you, so the question is which message gets them to add a second product. This guide walks through structuring a SaaS cross-sell email test with marginal, the hosted email marketing MCP server your AI agent can drive directly.
Why cross-sell emails need disciplined testing
In SaaS, cross-sell sends go to engaged, paying users — which means small lift differences translate into meaningful expansion revenue at scale. A vague 'check out our new module' email rarely beats a sharp value-first angle, but you won't know without measuring it.
The trap is shipping one variant and calling the open rate 'fine.' A proper email A/B test isolates one lever at a time so you can attribute the result. Subject line is usually the highest-leverage place to start for cross-sell, since it gates whether the offer is even seen.
- Segment by adopted product, not by plan tier alone
- Test the angle: time saved, workflow gap, or bundle pricing
- Start with a subject line A/B test before reworking the body
- Watch click-through, since cross-sell intent shows in the click
Structuring the test in marginal
marginal exposes four MCP tools your agent calls in sequence: generate_variants, launch_test, get_results, and recommend_next. For a SaaS cross-sell email test, you describe the target product, the audience, and the cross-sell offer, then let the agent produce subject line variants to launch.
Because marginal is hosted at https://marginal.sh/mcp, there's no infrastructure to run. The managed send and open/click tracking happen on marginal's side, and within-test lift metrics come back so your agent can judge which variant won.
- generate_variants — draft competing cross-sell subject lines
- launch_test — start the email A/B test with a managed send
- get_results — pull open/click and within-test lift
- recommend_next — get the suggested follow-up variant
Letting an AI agent run the loop
AI agent email marketing works well for cross-sell because the workflow is repetitive across segments. Connect marginal as an MCP server in Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Codex, or Zed, and the agent can generate variants, launch, read results, and propose the next test without you leaving the editor.
A typical iteration: the agent generates four subject lines for a CRM-to-analytics cross-sell, launches the subject line A/B test, waits for get_results, and uses recommend_next to draft the follow-up for non-clickers — all from a single conversation.
- Auth uses a Bearer API key (marg_live_...)
- Free tier covers 100 experiments per month
- Run multiple segment tests in parallel from one agent session
- Docs at https://marginal.sh/docs/ cover each tool
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