See what visitors do with each variant.
The report tells you whether a variant wins. Session recordings and click maps help you understand why: where visitors click, what doesn’t respond, where they leave. Without any of the visitor’s text, images, or input, and only if you turn it on.
- Original12 points
Clicked 5 times in a row on “Appliquer”, then left from /panier without converting.
no order - Progress bar9 points
Clicked 4 times in a row on “Only €12.40 away from free shipping”, then converted on /commande.
ordered - Original7 points
Clicked “Free shipping from €60” with no effect, then left from /panier without converting.
no order
Read the transcript
Why does B win? Sessions and click maps show what visitors do with each variant.
Each session reads as timestamped events. What happened, page by page, to the second. Here, 1 rage click burst, 2 dead clicks, and left without converting after the tested element: 12 points.
The sessions to watch first. For “Free shipping progress bar”, 12 sessions kept out of 14: the ones with the most signals first, alternating between variants.
One click map per variant. On /panier, 12,201 visitors. In B, “Ambre candle · €14.90” goes from 4.8% to 8.4% of visitors: customers add an item rather than pay for shipping.
Probant receives only timestamped events. You turn it on yourself, and each visitor has to consent. Sessions are deleted 30 days after their last activity.
This test’s sessions are in the demo, no sign-up needed.
From the figure to the journey.
The visuals come from the demo store’s “Free shipping progress bar” test: its sessions and its cart clicks, run through the same calculations as in your workspace.
The sessions to watch first are picked for you.
Nobody has time to watch everything. Each session gets points for what deserves a look: rage clicks, dead clicks, JavaScript errors, leaving without converting after seeing the tested element. A session with no points isn’t suggested. For “Free shipping progress bar”, the rule keeps 12 sessions out of 14.
- The rule is shown under every list
- With equal points, the most recent comes first
- The list alternates variants, so each one gets its sessions
- Filters by variant and by outcome
Each session reads page by page.
A sequence of timestamped events from the moment each page opens: pages viewed, scrolling, clicks, the tested element appearing, errors, conversion. On the pages you choose, the replay also shows the page as the visitor saw it, cart and account included, rebuilt from its masked structure: text as bars of the same length, images as flat blocks, empty fields. A one-sentence summary recaps these events. At the top of the list, a visit to the original, on desktop, totaling 12 points.
- A session ends after 30 minutes without activity; until then, it isn’t scored
- Each page viewed links to that page’s click map
- 0:01The tested element appears on screen
- 0:06Scrolled to 100% of the page
- 0:18Click on a masked field
- 0:41Rage clicks: 5 rapid clicks on “Appliquer”, with no effect
- 1:10Dead click on “Appliquer”
A click and scroll map for each variant.
On each page of the test, the map shows where each variant’s visitors click and how far they read, by device. Nothing extra is collected: maps are counts drawn from sessions already recorded. On /panier, the demo brings together 12,201 visitors, on desktop and mobile.
- Shading by share of page clicks, on the same scale for all variants on a device
- Elements that get rage clicks or dead clicks are flagged
- A map from 30 visitors of one variant on the page
- Check out11 rage bursts47%
- Shipping calculated at the next step535 dead clicks8.2%
- Masked field8.2%
- Apply99 rage bursts, 144 dead clicks6.4%
- Shipping and returns4.9%
At least half of visitors scroll down to 90% of the page.
- Check out9 rage bursts42%
- Only €12.40 away from free shipping98 rage bursts, 718 dead clicks10%
- Ambre candle · €14.907.1%
- Masked field6.8%
- Apply94 rage bursts, 137 dead clicks5.2%
At least half of visitors scroll down to 90% of the page.
/panier · share of page clicks, all devices · 12,201 visitors
A and B are compared, without naming a winner.
For the most clicked elements, Probant compares the share of visitors who click at least once, with a range. The error risk is shared across all the comparisons shown: 19 here, hence 97.6% ranges. With “Progress bar”, the clearest difference is on “Ambre candle · €14.90”: +3.5 pts. Then “Filament bulb · €9.90”: +2.6 pts.
A map doesn’t name a winner: a difference in clicks describes behavior, and only the test’s metric decides.
No text, images, or input, and nothing without your consent.
Recording is off by default. Only the workspace owner can turn it on, in the settings or from a test page: it applies to all of the site’s visitors. Nothing to add to your site: the script already installed loads the sessions module only when the workspace records sessions, and after the visitor agrees when the workspace requires consent.
- A click in a form field keeps only the field type, with no label or position
- Areas marked data-probant-mask are treated as fields
- Email addresses and long numbers are replaced in labels
- Masked structure off by default, limited to chosen pages, recorded in the audit log
- Sessions deleted 30 days after their last activity
- 0:18Click on a masked field
- Keystrokes and field values
- Any text or image from the page viewed
- The page title, the referrer, the IP address
In detail.
- What is collected
- Per page viewed: path, screen type, time spent, maximum scroll, clicks, rage clicks, dead clicks, JavaScript errors, the tested element appearing, conversion.
- Rage click
- 3 or more clicks within 700 ms on the same element.
- Dead click
- Nothing changes on the page within 1 s, no scrolling and no URL change. Only judged for elements that should respond: link, button, image.
- Labels and errors
- Visible label of up to 60 characters; errors reduced to their message, up to 140 characters, with no stack trace or file name.
- Scrolling
- Share of visitors who reach each 10% step of the page height.
- A/B comparison
- Up to 40 elements per variant, plus each page depth, compared only with at least 5 visitors who click (or reach it) and 5 who don’t; differences that hold are kept with the Benjamini-Hochberg method, with at most 5% false differences among them on average.
- Retention
- 30 days after the last activity, then automatic deletion. Maps follow the same rule.
- Variants
- The module doesn’t send the variant: Probant recomputes it on receipt, as for assignment. A session can’t be attributed to a variant chosen by the page.
Frequently asked questions.
Do I need to install another script?
No. The script already installed loads the sessions module when the workspace records sessions. Without sessions turned on, nothing more is downloaded.
What do you see of a form?
That a field was clicked, nothing more: not its value, its label, or the click position. Keystrokes are not listened to.
Can a click map name a winner?
No. It describes behavior, with a range for each difference. The winner is decided on the test’s primary metric, at the checkpoint set in the test plan.
Next: Guardrails
See this feature in the demo, on the screen this page describes, or move on to the next one.
- Open demo, no sign-up
- Free to start, no overage charges
- Database and compute in Amsterdam, with Scaleway