Destination URL Click Tracking in Email Tests
When you run an email experiment, opens tell you who showed up — but destination URL click tracking tells you who acted. This guide explains how marginal records clicks on the links in your emails and ties them back to each variant so your AI agent can pick the version that actually drives traffic.
What destination URL click tracking measures
Destination URL click tracking records every time a recipient clicks a link that leads somewhere — a landing page, a product, a docs article. Unlike an open, a click is an intentional action, which makes it the closest signal to revenue that lives inside the email itself.
marginal handles this tracking as part of its managed sends. Each link in your message body is instrumented so clicks are attributed to the recipient and to the variant they received, without you building a redirect service or a tracking pipeline yourself.
- Per-variant click counts so you can compare versions fairly
- Open and click tracking captured together in one result set
- Clicks rolled up into within-test lift metrics, not raw logs you have to parse
How marginal connects clicks to subject line A/B tests
Click data only matters in context. In a subject line A/B test, two recipients might both reach the same destination URL, but one came through a curiosity-driven subject and the other through a direct, benefit-led one. marginal keeps that attribution intact so you know which subject line earned the click — not just that a click happened.
The flow is simple: generate_variants drafts subject lines, launch_test sends them across your split, and get_results returns opens, clicks, and lift per variant. Because destination URL click tracking is built into the send, there's no separate setup step to enable it.
- generate_variants — produce subject line candidates
- launch_test — send and begin tracking opens and clicks
- get_results — read per-variant click rates and lift
- recommend_next — get a data-backed call on the winning variant
Letting an AI agent act on click data
The point of putting this inside an email MCP server is that an AI agent can read click results and decide what to do next — no dashboard babysitting. In AI agent email marketing workflows, click rate becomes the metric the agent optimizes toward, because it reflects engagement with your actual destination, not just inbox curiosity.
Connect any MCP client — Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, or OpenAI Codex — to https://marginal.sh/mcp with your Bearer API key. The agent calls get_results, sees which variant produced more destination URL clicks, and uses recommend_next to suggest the version to scale.
- Optimize for clicks-to-destination, not just opens
- Hosted endpoint at https://marginal.sh/mcp — no self-hosting
- Free tier covers 100 experiments per month to validate the workflow
Reading click metrics the right way
A high click rate on one variant only means something if the difference holds up across your test population. marginal reports within-test lift so you can see how a variant's click performance compares to its peers under the same conditions, which guards against chasing a small sample.
Pair click rate with the open rate from the same test to understand where a variant succeeds or fails. A strong open but weak click suggests the subject promised more than the body delivered — exactly the kind of insight that makes your next round of generate_variants sharper.
- Compare click lift across variants before declaring a winner
- Read open and click together to diagnose subject-vs-body gaps
- See the full tool reference at https://marginal.sh/docs/
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