Investigate a website traffic drop with AI
Compare the suspected drop with a baseline in PostHog, find the dimension that explains most of it and identify one check that could confirm the cause.
To investigate a website traffic drop with AI, give Web analytics triage the affected metric and the date you first noticed it. A weekly report may show fewer visits, yet comparing an incomplete morning with a full day would make a healthy site look quiet. The agent starts by measuring the size and timing of the change against a suitable baseline.
It then checks explanations that would lead to different decisions. A decline concentrated in one referral source suggests a different follow-up from missing events on one browser after a release. The purpose of the report is to name the strongest hypothesis and a way to test it; a broad decline can remain unexplained when no single group accounts for it.
What it does
- Compare pageviews, sessions or conversion against a baseline.
- Look for changes coinciding with a deployment.
- Break down the difference by referrer, device, geography or path.
- Identify the segment explaining most of the movement.
- Return the leading explanation and one concrete follow-up.
Install and use
- Install the official skill from the PostHog community store. Connect an MCP-capable agent with query access to a project that already records the relevant metrics.
- Ask “Course-page visits fell on Tuesday. Compare the change with a matching period and check browser and referral sources.”
- Review the proposed cause and verify it against a release, link change or tracking check.
FAQ
Source
Made by PostHog. Original skill: Web analytics triage, published on GitHub under the MIT licence.
View on GitHub