Find out why your website traffic spiked

A sudden jump in visitors is only useful if you know where it came from. Was it the post you published, a mention you didn't know about, or a bot? The answer decides whether to repeat it.

How a spike is detected

Every hour is compared with the same hour of the week in earlier weeks, so a quiet Sunday night and a busy Monday morning each have their own baseline. An hour counts as a spike when it has at least 10 extra pageviews and is at least 50% above normal. Detection needs a few days of history before it has a baseline, so a brand-new site won't show spikes yet.

What explains it

For each spike you get the likely causes ranked by how much they rose above their own normal level:

• the source (a UTM source, else the referring site, else direct),
• the landing page,
• and the events, such as signups, that rose with it.

An AI summary then states what happened in a sentence or two, and you can ask your agent for the same thing with the explain_spike tool.

Give spikes context

Log what you do. When your agent creates a tracked link for a post, or you record a deploy or launch, it is saved as an annotation. Spikes list any annotation from the 24 hours before them, so "Show HN post" sits right next to the jump it caused.

Then filter to the spike

On the dashboard, click the bar for the spike's hour, then click a source or page to see how those visitors behaved. Everything on the page narrows to that traffic, and you can remove the filters one at a time.

Ask your agent: "Why did traffic jump on Tuesday, and should I repeat it?"

More guides

• A website analytics MCP server: ask your agent about your traffic
• Add website analytics to your project with Claude Code
• Track clicks and find where visitors drop off, without writing code
• Cookieless analytics: how it works, and what you give up
• A simpler Google Analytics alternative for developers