Skip to content
Marketing Analytics LMS

Case 5: Investigating Funnel Drop-offs at User Level

Case 5: Investigating Funnel Drop-offs at User Level

Section titled “Case 5: Investigating Funnel Drop-offs at User Level”

Starting Point

Section titled “Starting Point”

Funnel showing 90% drop-off (50,266 people) between landing page and achieve screen

The Investigation Process

Section titled “The Investigation Process”

1. Don’t accept the number as abstract

Section titled “1. Don’t accept the number as abstract”
  • This isn’t just “55,000 views dropped to 5,000”
  • These are 50,266 SPECIFIC individuals you can investigate
  • Each one has a complete profile and session data

2. Click through to see the dropped-off users

Section titled “2. Click through to see the dropped-off users”
  • PostHog allows clicking on the drop-off segment
  • Opens a list of people who viewed landing page but didn’t reach achieve screen
  • Can scroll through hundreds/thousands of actual user profiles

3. For each user in the drop-off, you can examine

Section titled “3. For each user in the drop-off, you can examine”

Session Behavior:

  • Did they bounce immediately? (session recordings show this)
  • How long did they stay on landing page?
  • Did they interact with anything?
  • Where did they drop off?

Traffic Source Context:

  • utm_source: fb, google, direct, etc.
  • utm_medium: cpc, social, email, etc.
  • utm_campaign: Specific campaign ID
  • utm_term: Keywords or targeting parameters
  • utm_content: Ad variation identifier
  • ad_id: Exact ad that drove them
  • adset_id: Ad set configuration
  • campaign_id: Campaign identifier
  • aff_sub: Affiliate/channel identifier

Landing Page Context:

  • landing_page_source: Full URL with all parameters
  • $initial_host: Domain they landed on
  • Which landing page variant did they see?

User Context:

  • Geography (subdivision, timezone)
  • Device (viewport width/height)
  • Browser properties
  • When they first visited (first_seen_at)

4. Pattern Discovery Through Segmentation

Section titled “4. Pattern Discovery Through Segmentation”

With all this context, you can discover patterns:

  • “All users from campaign X are bouncing” → Bad ad creative or targeting
  • “Users from utm_term=Y convert better” → That keyword/targeting works
  • “Mobile users (viewport width < 500) drop off more” → Mobile experience issue
  • “Traffic from specific ad set bounces immediately” → Ad/landing page mismatch
  • “Users who land on variant A bounce, variant B converts” → A/B test winner identified

5. Segmented Analysis

Section titled “5. Segmented Analysis”

Instead of just “90% drop-off,” you can segment:

  • Drop-off rate by campaign
  • Drop-off rate by landing page
  • Drop-off rate by traffic source
  • Drop-off rate by device type
  • Drop-off rate by geography

This reveals WHERE the problem is, not just THAT there’s a problem

Example Investigation Questions You Can Answer

Section titled “Example Investigation Questions You Can Answer”

❓ “Is this drop-off consistent across all campaigns?” → Segment by campaign_id, check conversion rates

❓ “Are we buying bad traffic from specific sources?” → Filter by utm_source and ad_id, compare bounce rates

❓ “Is the landing page broken on mobile?” → Segment by device properties, watch mobile session recordings

❓ “Did a specific ad drive poor-quality traffic?” → Look at users with that ad_id, watch their sessions

❓ “Is one landing page variant performing worse?” → Compare landing_page_source variants

Key Insights

Section titled “Key Insights”
  1. Numbers represent real, investigable people: The 50,266 drop-offs aren’t a black box
  2. Context enables pattern discovery: Rich metadata lets you find “all X have problem Y”
  3. From aggregate to actionable: You go from “90% drop-off” to “Campaign 12023 with utm_term=discount is driving bounces”
  4. No blind spots: You can investigate any angle you think of - if the data exists, you can segment by it
  5. Hypothesis testing without waiting: Don’t need to ask for reports - just filter and see

Principles Applied

Section titled “Principles Applied”

Tool Features Enabling This

Section titled “Tool Features Enabling This”
  • Click-through from funnel drop-offs to person list
  • Full property data retained at person level
  • Ability to filter and segment by any property
  • Session recordings linked to every person
  • UTM and campaign tracking automatically captured
  • Multiple ways to slice the same data (by campaign, source, term, content, etc.)
Section titled “Navigation”