Running a Product Name in Subject Line Test
Does putting the product name in your subject line lift opens, or does it feel like spam? Instead of guessing, run a controlled product name in subject line test through marginal — a hosted email MCP server your AI agent can drive directly.
Why test the product name placement
Leading with a literal product name can boost relevance for warm audiences but suppress curiosity for cold lists. The only reliable way to know which effect dominates for your audience is a subject line A/B test that measures real opens and clicks.
A product name in subject line test isolates one variable: the presence (or position) of the name. Hold everything else constant — preview text, send time, segment — so the lift you measure is attributable to the name itself.
- Variant A: "Your weekly summary is ready"
- Variant B: "Your Acme Dashboard summary is ready"
- Variant C: name early — "Acme: this week in 60 seconds"
- Keep audience, timing, and body identical across variants
How marginal runs the experiment
marginal exposes four MCP tools so an AI agent can run the whole loop without you wiring up an ESP. Generate candidate subject lines, launch the split, read open/click results, and ask for the next recommended move.
Because marginal is a hosted MCP server, there's nothing to deploy. You connect a client over the endpoint at https://marginal.sh/mcp with a Bearer API key and the tools become available to your agent.
- generate_variants — draft with and without the product name
- launch_test — start a managed subject line A/B test
- get_results — opens, clicks, and within-test lift
- recommend_next — suggest the winning variant or a follow-up test
Reading the lift correctly
marginal reports within-test lift metrics so you compare variants under the same send conditions rather than against a historical baseline. That matters for a product name test, where seasonality and list fatigue can otherwise masquerade as a name effect.
Look at both open rate and downstream clicks. A product name in the subject sometimes raises opens but attracts the wrong readers, flattening clicks — the recommend_next tool factors that into its suggestion.
Fit it into AI agent email marketing
If you're building AI agent email marketing workflows, this test becomes a reusable building block. Your agent can propose the product-name hypothesis, launch it, wait for significance, and promote the winner — all through marginal's MCP tools.
The free tier covers 100 experiments per month, enough to iterate on subject lines across several campaigns. See the docs at https://marginal.sh/docs/ for tool schemas and connecting your client.
- Works with Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Codex, and Zed
- No self-hosting — marginal runs the sends and tracking
- Authenticate with a marg_live_ Bearer key
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