Dedicated IP vs Shared Email Test: A Practical Guide
If you're weighing a dedicated sending IP against a shared pool, the only honest answer comes from testing your own list and content. Here's how to design that comparison cleanly — and where subject line testing fits into the picture.
What a dedicated IP vs shared email test actually measures
A dedicated IP gives you full control over a sending reputation, but it only pays off at consistent, high volume. A shared IP pool spreads reputation across many senders, which can help low-volume programs warm faster but ties your fate to other people's behavior.
When you run a dedicated IP vs shared email test, you're isolating one variable: the sending infrastructure. To keep the comparison meaningful, hold everything else steady — same audience segment, same send time, same content, same authentication (SPF, DKIM, DMARC).
- Inbox placement rate per IP type, segmented by mailbox provider
- Open and click rates as downstream proxies for placement
- Bounce and complaint rates over the warm-up window
- Time-to-stable-reputation for a freshly assigned dedicated IP
Why subject lines confound deliverability tests
Subject lines move open rates more than most infrastructure changes do. If you compare a dedicated IP send with one subject line against a shared send with another, you can't tell what caused the difference. That's why a subject line A/B test should run inside each IP arm — not across them.
The clean design is a nested one: split each IP cohort into the same subject variants. Now you can read deliverability effects and creative effects separately, instead of guessing which one drove the lift.
- Use identical subject variants in both the dedicated and shared cohorts
- Randomize within each cohort to avoid time-of-day bias
- Report within-test lift so creative wins don't masquerade as IP wins
Running the test with marginal's email MCP server
marginal is a hosted email MCP server that lets an AI agent design, launch, and read these experiments through a small set of tools. You drive it from clients like Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, or OpenAI Codex — the agent calls the tools and you get structured results back.
For a dedicated IP vs shared comparison, generate matching subject variants, launch a managed send for each cohort, then pull open/click tracking and within-test lift metrics to compare arms. Because it's hosted at https://marginal.sh/mcp, there's no infrastructure for you to stand up.
- generate_variants — produce identical subject lines for both IP arms
- launch_test — run the managed send and subject line A/B test per cohort
- get_results — fetch open/click tracking and lift figures
- recommend_next — let the AI agent email marketing loop suggest the follow-up
Reading the results without fooling yourself
Give a dedicated IP time to warm — early sends will underperform a settled shared pool simply because no reputation exists yet. Compare stable windows, not day-one numbers, and weight by mailbox provider since Gmail and Outlook can respond very differently to the same IP.
marginal's free tier covers 100 experiments per month, which is plenty to run a properly nested IP-by-subject comparison a few times before committing to infrastructure. The docs at https://marginal.sh/docs/ walk through the tool calls and result schema.
- Don't declare a winner during IP warm-up
- Segment placement by provider before averaging
- Confirm subject-line lift is consistent across both IP arms
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