Major Improvements to SegmentStream Attribution

What’s Changed?

Previously, SegmentStream’s Visit Scoring Attribution analyzed user journeys based on visit sequences and interactions. While this approach provided significantly greater accuracy compared to traditional multi-touch attribution models, we identified a key area for improvement: not all visits contribute equally to conversions, even if they contain certain events.

With this update, our model now incorporates active time spent on key pages as a stronger indicator of user intent.

The model now tracks:

• Mouse movements – Distinguishes active engagement from passive page loading.
• Clicks and scroll depth – Identifies meaningful interactions beyond just landing on a page.
• Focus changes – Differentiates between real user activity and background tabs or inactive sessions.

By tracking actions like mouse movements, clicks, scrolls, and focus changes, we measure the energy users invest in researching your product. We found that energy better predicts future conversion rather than just the sequence of events, and we proved this through experiments. Using a geo holdout test, we confirmed that this new approach delivers more precise attribution results, significantly improves bot detection, and better reflects user behaviour.

We validated this approach through geo holdout experiments, confirming that incorporating active engagement:

• Increases the precision of conversion prediction – Better differentiation between engaged and passive users.
• Improves bot detection — Reduces the impact of low-quality traffic.
• Provides a clearer picture of marketing effectiveness — Helps marketers allocate budget toward traffic that truly matters.

How to Use It

This update has been automatically applied to existing SegmentStream customers, so no manual action is required. Attribution results in reports now reflect these improvements, ensuring that decision-making is based on a more precise understanding of user behaviour.

For a deeper dive into how this affects your specific use case, reach out to our team. We’ll continue refining our models to provide the most reliable insights for marketing attribution.

Stay tuned for further updates!