Behavioural analytics, reimagined

Understand how people actually browse

TraceTray turns low-level cursor signals into interpretable behavioural clusters. No session recordings, no guesswork.

0
cursor moves logged
0
pauses detected
0
clicks tracked
0s
session duration

From raw signals to real insight

Four steps from page load to behavioural cluster. No configuration needed.

01 / collect

Lightweight tracking

A single script tag. Logs cursor movement, pauses, clicks, and scrolls in real time. Batches and sends without blocking the page.

02 / store

Session assembly

Batches are merged server-side into complete session documents. Each session gets a unique ID generated at page load.

03 / extract

Feature engineering

Raw events are converted into interpretable features: pause rate, cursor velocity, spatial entropy, dwell before click, and more.

04 / analyze

Behavioural clustering

KMeans and agglomerative clustering group sessions into archetypes. Visualisations surface what separates each group.

Features grounded in research

Every signal is drawn from published work on cursor tracking and web usage mining. Click any card to learn more.

Pause Rate

Cursor pauses per second, a proxy for reading depth and attention.

Based on Huang (2011) and Schoemann (2021), pauses correlate with fixation and cognitive engagement. A high pause rate suggests the user is reading or deliberating, not just scanning.

click to expand
🖱

Cursor Velocity

Average speed of movement in pixels per millisecond.

Velocity distinguishes purposeful navigation from erratic scanning. Slow, deliberate movement often precedes a click. Fast movement across the page suggests the user is searching rather than reading.

click to expand
🎯

Dwell Before Click

Average time between the last pause and a click event.

Drawn from Schoemann (2021), pre-click dwell captures hesitation and decision friction. Long dwell times before clicking suggest the user is uncertain or reading surrounding content carefully before committing.

click to expand
🔀

Spatial Entropy

Shannon entropy of cursor distribution across the page.

Inspired by Rehse (2024), spatial entropy quantifies how broadly the cursor explored the page. High entropy means the user covered many regions; low entropy suggests focused attention in one area like header fixation.

click to expand
📐

Zone Ratios

Proportion of cursor activity in top, centre, and bottom page regions.

The page is divided into a 3×3 grid and cursor density is measured per cell. Zone ratios summarise whether the user focused on the header/navigation area, the main content, or scrolled to the bottom of the page.

click to expand
🔍

Move Density

Cursor moves per second, normalised for session length.

Raw move counts are meaningless without normalisation. A long session will always score higher. Move density controls for duration, making it a genuine signal of how actively the cursor was used regardless of how long the user stayed.

click to expand

Try it yourself

Your session is being tracked right now. The monitor on the right reflects your live signals. The quiz on the left generates click and decision data.

Behavioural Signals Quiz

Test your knowledge. Each answer generates a tracked click event.

Your Live Session

Real-time signals from your current browsing session on this page.

Cursor moves
0
Pauses
0
Clicks
0
Scrolls
0
Avg velocity
n/a
Duration
0s

These are the exact same signals TraceTray extracts and feeds into the ML pipeline. Your session will be included in the next analysis run.

Simple, transparent pricing

TraceTray is currently in early development. Join the waitlist to be notified at launch.

Starter
$19/mo

For indie developers and small projects getting started with behavioural analytics.

  • Up to 5,000 sessions/month
  • Core feature extraction
  • 3 behavioural clusters
  • CSV export
Enterprise
$149/mo

For larger sites that need custom clustering, data access, and priority support.

  • Unlimited sessions
  • Custom feature modules
  • Raw data API access
  • Dedicated support
  • Self-hosted option

Join the waitlist

TraceTray is currently in active development. Leave your details and we'll reach out when it's ready.

You're on the list.

We'll be in touch when TraceTray is ready. Thanks for your interest.

FAQ

How is TraceTray different from session recording tools like Hotjar?
Session recorders replay what happened visually. TraceTray doesn't record video. It extracts structural features from cursor signals and groups sessions by behavioural pattern. The output is a cluster label and a feature profile, not a replay. This means you get a summary of how people browse, not an archive of individual sessions.
What does the tracker actually collect?
Mouse movement coordinates (sampled at 50ms intervals), cursor pause events, click events with basic element context (tag, ID, text snippet), and scroll events. No keystrokes, no form values, no screenshots. The raw event log is only ever used to compute session-level features. It isn't stored long term.
How many sessions do I need before clustering is meaningful?
Practically, you want at least a few hundred sessions before trusting the cluster boundaries. With smaller datasets the pipeline still runs and produces output, but clusters may shift significantly as more data comes in. The silhouette score gives you a quantitative sense of how confident you should be in the current grouping.
Can I run TraceTray on my own infrastructure?
Yes, the Enterprise plan includes a self-hosted option. The full stack is a Node.js collection server, MongoDB, and a Python analysis pipeline. All components are open and composable, so you can adapt them to your own infrastructure without being locked into a SaaS model.
Is this project research-backed?
Yes. TraceTray grew out of an undergraduate research project on web usage mining and behavioural session analysis. The feature set is grounded in published work from Huang (2011), Schoemann (2021), Fernández-Fontelo (2020), and Rehse (2024). The pipeline is designed to be site-agnostic rather than optimised for a single controlled study context.