Five things watched
Swings in sessions are looked for across the five dimensions that most often explain one.
- Channels and referrers
- Countries and devices
- Entry pages
Features·Product analytics
The page that tells you what moved.
This page checks your numbers for you. When one moves, it tells you which one, by how much and how unusual that is.
What it does
Changes found for you.
Five detectors check your own history, using a baseline that allows for the day of the week.
A baseline that knows your week
A day is compared with the median of the same weekday in your history, so a normal quiet Sunday is not flagged. A week or month is compared with the one before.
- Median of matching weekdays
- Yesterday, rolling 7 and rolling 30 day windows
- Low, moderate or severe severity
Ranked and explained
Sorted by impact so the biggest thing is first, each with a plain-English summary of what moved. On Growth, one click hands it to the AI assistant to find out why; on any plan, you can pin it to your charts as a marker.
- Biggest change first
- What moved and by how much
- Ask why, or pin it to your charts
How it works
Use it in three steps.
01
Collect a couple of weeks
Detection compares against your own past, so it needs some history first.
02
Open Insights
Everything that moved unusually is already listed, worst first.
03
Act on the top one
Each insight names the dimension that moved, so the next report to open is obvious.
Read next
Usually read beside this one.
Product analytics
Boards you build once and read every day.
Build boards from saved reports and share them with a link.
Read morePlatform
Analytics your AI can operate.
Ask about your data in plain English, in the app or over MCP.
Read moreWeb analytics
The numbers you check every morning.
Break down visitors and revenue by where they came from.
Read moreFAQ
Questions about this one.
What people ask before they switch it on.
How is this different from an alert?
An alert fires on a threshold you had to guess in advance. This compares what happened with what normally happens, and finds changes you would not have thought to set a threshold for.
Why does a weekend dip not show up?
Because the baseline is the median of the same weekday in your own history. A quiet Sunday is only flagged when it is quiet compared to your other Sundays.
How much history do I need?
A couple of weeks, so each weekday has a baseline. Before that, the detectors have nothing to compare against.
Is the explanation written by AI?
No. The title and summary are built from the numbers, so they read the same way every time. They say what moved, by how much and against which baseline. “Why did this happen?” passes the insight to the AI assistant, which looks into your data.
See your own numbers today.
Free up to 100k events a month · no credit card