How to Run an NPS Survey Email Subject Test
Response rate on NPS surveys often hinges on a single thing: whether people open the email. This guide shows how to run an NPS survey email subject test with marginal, the hosted email MCP server, so your AI agent can test subject lines and pick the winner before the full send.
Why subject lines drive NPS response rates
An NPS survey only works if recipients see and engage with it. A subject like "Quick question — how are we doing?" may pull a different open rate than "Rate your experience (takes 30 seconds)." Small wording shifts compound across thousands of customers, and an unopened survey contributes zero data points to your score.
Because NPS sends usually go to a known, relatively engaged audience, you have a clean setting for a subject line A/B test: hold the audience and body constant, vary only the subject, and let real opens tell you which framing earns attention.
- Curiosity vs. clarity: "How did we do?" against "Share feedback in 1 click"
- Personalization: include first name or product name to lift relevance
- Urgency and effort cues: signal how short the survey actually is
Running the test with marginal's email MCP server
marginal exposes four MCP tools your agent can call directly: generate_variants, launch_test, get_results, and recommend_next. Connect once at https://marginal.sh/mcp with a Bearer API key, and an AI agent in Cursor, Claude Desktop, or Claude Code can drive the whole NPS survey email subject test conversationally.
A typical flow: ask the agent to generate several subject variants tuned to your survey tone, launch a managed send that splits traffic across them, then read back open and click results once the test has data.
- generate_variants — draft multiple NPS subject lines from one prompt
- launch_test — send the A/B split and track opens and clicks
- get_results — pull within-test lift metrics per variant
- recommend_next — get a data-backed suggestion for the rollout subject
Reading the results and choosing a winner
marginal tracks open and click rates per variant and reports within-test lift, so you can see how much better the leading subject performed against the others in the same send. For NPS specifically, watch clicks as well as opens — a click usually means the recipient reached the survey itself.
Once a variant clearly leads, use recommend_next to confirm the winner and apply it to the remaining audience. This keeps AI agent email marketing grounded in measured performance rather than guesswork.
Getting started
marginal is fully hosted — there is no self-hosted server to run and no Mailchimp or Klaviyo API key required to test subject lines. The free tier includes 100 experiments per month, which is plenty for ongoing NPS survey subject tests on most teams.
Point your MCP client at https://marginal.sh/mcp, add your marg_live_ key, and read the docs at https://marginal.sh/docs/ to wire up your first test.
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