How to Run an All Caps Subject Line Test
Does writing your subject line IN ALL CAPS actually win more opens, or does it just trip spam filters and annoy subscribers? Instead of guessing, run an all caps subject line test and let the open rates settle the argument. This guide shows how to set one up through marginal, a hosted email MCP server your AI agent can drive directly.
Why test all caps subject lines at all
All caps can read as urgent and attention-grabbing, or as loud and spammy — and the difference usually depends on your audience, your sending reputation, and the rest of the inbox they're scanning. There is no universal rule, which is exactly why an all caps subject line test beats a hunch.
A proper subject line A/B test isolates the variable: same audience, same send window, same body content, with only the casing changing between variants. That way the lift you see is attributable to the caps, not to a different offer or a better send time.
- Compare full caps vs. sentence case vs. selective caps on a key word
- Watch open rate first, then click-through to confirm the lift carries
- Keep an eye on spam-flag signals if your ESP exposes them
- Test on a representative sample before rolling out to the full list
Setting up the test with an email MCP server
marginal is a hosted email MCP server, so your AI agent connects to a single endpoint and gets a small set of tools for running experiments. There's nothing to deploy and no server to maintain — you authenticate with a Bearer API key and the agent handles the rest.
For an all caps subject line test, the typical flow uses four tools: generate_variants to draft casing options, launch_test to send them against a sample, get_results to pull open and click metrics, and recommend_next to suggest the winning treatment.
- generate_variants — produce all caps, sentence case, and hybrid options
- launch_test — run a managed send as a subject line A/B test
- get_results — read open/click tracking and within-test lift metrics
- recommend_next — get a data-backed call on which casing to ship
Letting an AI agent run the experiment
Because marginal speaks MCP, AI agent email marketing workflows become conversational. Tell your agent in Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, or OpenAI Codex to 'run an all caps subject line test for the spring promo' and it can call generate_variants, launch the test, and report the lift once results land.
Connect to the endpoint at https://marginal.sh/mcp and the free tier covers 100 experiments per month — enough to test casing across several campaigns before you commit. Setup details live in the docs at https://marginal.sh/docs/.
Reading the results without fooling yourself
All caps often produces a short-term open bump that doesn't survive scrutiny. Look past the headline open rate and check whether clicks and downstream conversions hold up — a louder subject that draws curiosity but not intent can hurt long-term engagement.
marginal reports within-test lift so you're comparing variants under identical conditions rather than against last month's campaign. Run the test, read the lift, and only escalate the caps to your full audience if the data earns it.
- Prefer click and conversion lift over raw opens for the final call
- Re-test occasionally — audience tolerance for caps shifts over time
- Use selective caps as a middle ground if full caps underperforms
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