Skip to main content

Custom metric, built and refined in place

Metric building is the highest-volume thing customers do with the MCP. The first answer is almost never the final one. The MCP is designed to refine in place: change the time range, swap a dimension, add a filter, and the same metric updates. This shows the iteration loop.

Turn 1: Build it

Fullstory MCP
Build me a metric: of users who hit /signup in the last 30 days, what percent triggered a custom event "account_created"?
Fullstory MCP
AI Response

Built a draft metric: "Signup-to-account-created conversion (last 30 days)". View and save it in Fullstory: [metric link]

Note: your org doesn't have a defined event named "account_created". The closest match is "signup_complete", which fires on the /welcome page after signup. Should I use that instead?

Turn 2: Compute it

Fullstory MCP
Yes, use signup_complete. Run it.
Fullstory MCP
AI Response

Signup-to-signup_complete conversion (last 30 days): 62.4%

  • 12,418 users hit /signup
  • 7,749 of them fired signup_complete

Turn 3: Refine the grouping

Fullstory MCP
Now break it down by traffic source.
Fullstory MCP
AI Response

Conversion by traffic source (last 30 days):

Traffic sourceConversionUsers
Organic search78.2%4,102
Direct71.5%3,228
Email68.9%1,634
Referral59.7%540
Paid social41.3%2,914

Turn 4: Tighten the cohort

Fullstory MCP
Now show me only mobile users.
Fullstory MCP
AI Response

Conversion by traffic source, mobile only (last 30 days):

Traffic sourceConversionUsers
Organic search71.8%1,883
Direct64.2%1,541
Email58.3%749
Referral51.1%224
Paid social28.7%1,612

Turn 5: Watch it over time

Fullstory MCP
Change it to a daily trend over the last 60 days. Keep the mobile filter and the paid social cohort only.
Fullstory MCP
AI Response

Daily trend: paid social mobile conversion (last 60 days)

Values range from 22% to 41%.

Notable shifts:

  • Dropped from 38% to 24% the week of April 12
  • Returned to 33% by April 26
  • Currently trending at 29%

What you walk out with

A draft metric, refined through five turns, with a link to view it in Fullstory. If you want to keep it for a dashboard or ongoing monitoring, click through the link and save it. The MCP builds and refines the definition for you, but the final save is a deliberate user action so nothing clutters your org by accident.

Behind the scenes

The MCP routes this workflow through: build_segment, build_metric, compute_metric, and update_metric. Turns 3, 4, and 5 are all update_metric calls — the same draft definition is refined in place across the chain.