Audience Segmentation

Audience segmentation is the practice of dividing website visitors into groups based on shared characteristics — such as behavior, demographics, or traffic source — to deliver targeted experiences and analyze experiment results.

Audience segmentation is the process of dividing your visitors into distinct groups based on shared attributes. Instead of treating all traffic as one homogeneous audience, segmentation lets you understand how different groups behave and optimize for each one.

Common Segmentation Criteria

Behavioral Segments

  • New vs. returning visitors — First-time visitors need different messaging than people who've been to your site before
  • Engagement level — Visitors who've read 5 articles behave differently from those who bounced after one page
  • Funnel stage — Someone on the pricing page is further along than someone reading a blog post

Demographic & Firmographic Segments

  • Company size — Enterprise buyers have different needs than startups
  • Industry — A SaaS company and an e-commerce brand care about different features
  • Job role — Marketers, product managers, and developers respond to different messaging

Traffic Source Segments

  • Organic search — High intent, looking for specific answers
  • Paid ads — Targeted but may need more convincing
  • Social media — Often exploratory, lower immediate intent
  • Referral — Arrives with context and trust from the referring source

Technical Segments

  • Device type — Mobile vs. desktop visitors convert at very different rates
  • Browser/OS — Useful for identifying technical issues affecting specific platforms
  • Geographic location — Language, currency, and cultural expectations vary by region

Segmentation in Experimentation

Segmentation is critical for understanding experiment results beyond the top-line number:

Post-Test Segmentation

After a test concludes, break down results by segment. A variant might show a modest 5% overall lift, but segmented analysis could reveal:

  • +15% lift for mobile visitors
  • -2% lift for desktop visitors
  • +20% lift for new visitors
  • No change for returning visitors

This tells you the variant is working for specific audiences, not universally.

Pre-Test Segmentation (Targeting)

You can also target experiments to specific segments. For example, show a simplified checkout flow only to mobile visitors, or test enterprise-focused messaging only for visitors from companies with 500+ employees.

Segmentation Pitfalls

  • Too many segments — Slicing data too thin reduces sample sizes and produces unreliable results
  • Segment fishing — Looking at dozens of segments until you find one that "won" is a form of p-hacking. Segments should be defined before the test, not discovered after.
  • Ignoring interactions — Segments can overlap (a visitor can be both "mobile" and "new"), which complicates analysis