Agentic analytics examples: 12 questions to ask about your users
Twelve questions a SaaS team can ask an analytics agent, from “what happened this week” to “who got stuck on checkout”, with what Behavly Agent reads to answer each one and the step it can take next.
The easiest way to understand agentic analytics is to see the questions it answers. Below are twelve a SaaS team asks every week, grouped by what they are about. For each one: what Behavly Agent reads to answer it, and what it can do next.
Every answer comes from your own workspace at the moment you ask. Every next step that would change something waits for you to choose Confirm.
Traffic: who came and what they did
1. “What did people do on my site this week?”
It reads: site behavior across your tracked sites for the last seven days: visitors, visits, page views, clicks, rage clicks, bounce rate, how many visitors are identified, the top pages, referrers, countries, devices and a trend.
Next: ask what changed against last week, or for a chart of it.
2. “Where did this month’s visitors come from?”
It reads: referrers, countries, devices, browsers and UTM campaigns for the period, for one site or all of them.
Next: ask for one site only, or for the same numbers against the month before.
3. “Which pages do people interact with most?”
It reads: the pages that have heatmap data, ranked by how much there is to draw (clicks count most, then pointer movement, then scrolling), with click and scroll counts per page.
Next: ask what people click on the top page. That is question 7.
Signups and activation
4. “Where do signups drop off?”
It reads: the signup funnel step by step (visitors, engaged, saw signup, started the form, submitted, newly identified), where people dropped off and which form fields failed.
Next: open the heatmap of the step that lost the most people. More in AI funnel analysis.
5. “Who visited pricing this week?”
It reads: identified users, the signed-in people the snippet recognized, one row per person with browsers and sites merged by email, filtered to the page. Each comes with last activity, visits, pages, device and country.
Next: ask for them as a pattern a campaign can start from. The Find High-Intent Visitors prompt on the Agent home does exactly this for repeat pricing visitors.
6. “How many new users did we get, and how many have gone quiet?”
It reads: audience health for the identified people the snippet knows: active (seen in the last 7 days), at risk (8 to 30 days), inactive (over 30 days) and new (first seen in the last 7 days), with how those counts moved over the period.
Next: ask for the at-risk users as a pattern and a campaign that brings them back.
Friction: where people got stuck
7. “What frustrates people on checkout?”
It reads: that page’s heatmap as numbers: the most-clicked elements with their share of clicks, where people rage-clicked (3 or more fast clicks in one spot), clicks that did nothing, clicks by device and how far down the page people scroll.
Next: open the painted map from Heatmaps on the live page to see it for yourself. More in Rage clicks and dead clicks.
8. “Who got stuck this week?”
It reads: identified users filtered by friction: the people who rage-clicked or dead-clicked in the period, with the pages they were on.
Next: look at one person’s own heatmap, or ask for a follow-up that offers help with the step they were on.
People: one person, or a group
9. “What did this person do on our site?”
It reads: that one person’s activity, found by part of their email address: visits, pages, time on site, rage and dead clicks, device, country and recent pages.
Next: in the Inbox, every reply already arrives with the sender’s site history beside it, and Behavly Agent can draft the answer.
10. “Find the people who saw pricing but never bought.”
It reads: your words, turned into a pattern Behavly can check. It answers with the pattern’s name, its rules in plain words, the exact number of people who match right now and a sample of up to five. Nothing is saved yet.
Next: save it, and an always-on campaign takes in each new person who matches. More in Find user behavior patterns with AI.
Results: what your follow-ups did
11. “How did my last campaign do?”
It reads: that campaign’s results from your workspace. The Campaign Review prompt on the Agent home asks for rates, drop-off and three fixes.
Next: ask it to make one of the fixes. It shows the change on a confirmation card first.
12. “Show me visitors this month as a chart.”
It reads: the numbers for the period, then draws them in the chat as a line, bar or donut chart, and comments on what it shows.
Next: ask what explains the peak.
All twelve at a glance
| Question | What Behavly Agent reads | Answer is |
|---|---|---|
What did people do this week? | Site behavior, last 7 days | Numbers and a trend |
Where did visitors come from? | Referrers, countries, devices, UTM | Ranked sources |
Which pages get the most interaction? | Pages with heatmap data | Ranked pages |
Where do signups drop off? | The signup funnel | Steps and failed fields |
Who visited pricing? | Identified users by page | People |
New and quiet users? | Audience health | Active, at risk, inactive, new |
What frustrates people on checkout? | The page’s heatmap | Elements, rage and dead clicks |
Who got stuck? | Identified users by friction | People |
What did this person do? | One person’s activity | Their pages and clicks |
Saw pricing, never bought? | Your words as a pattern | Rules, count, sample |
How did my campaign do? | Campaign results | Rates and fixes |
Visitors as a chart | The period’s numbers | A chart in the chat |
Questions about signed-in users work on every plan. Anonymous visitors, people the snippet saw but could not tie to a person, are a paid-plan feature from Starter up.
What makes these questions agentic
None of them names a report. The agent decides what to read, reads it live, and answers with the people as well as the number. And most end in a step it can take for you: a saved pattern, a campaign, a reply draft. That last part, from an answer to a follow-up email, is in From insight to action.

