Data Sampling Issues

Diagnose and fix data sampling that affects accuracy and reliability in high-traffic analytics reports

Data Sampling Issues

What This Means

Data sampling occurs when analytics platforms process only a subset of your data instead of all available data, then extrapolate results. While this speeds up report generation, it can lead to inaccurate insights, especially for detailed segments or custom reports in high-traffic properties.

Impact on Your Business

Inaccurate Insights:

Reports based on estimates, not actual data
Segment analysis unreliable
Small segments may show incorrect trends
Confidence in data diminished

Decision-Making Problems:

Optimization based on sampled data
A/B test results potentially invalid
Budget allocation on incomplete information
Cannot trust detailed analysis

Analysis Limitations:

Cannot drill into specific segments
Custom reports show different results each time
Funnel analysis inaccurate
Time-series data inconsistent

How to Diagnose

Method 1: Check for Sampling Indicator in GA4

Look for sampling badge:
- Open any GA4 report
- Top right corner shows sampling status
- Green checkmark = unsampled
- Yellow badge = sampled data
Check sampling percentage:
- Click sampling badge
- Shows "Based on X% of sessions"
- Lower percentage = less accurate
Review data collection:
- Reports → Realtime
- Check daily session volume
- Note GA4 limits

What to Look For:

Sampling badge appears in reports
Percentage of sessions sampled
Which reports trigger sampling
Time periods affected

Method 2: Compare Standard vs Custom Reports

Check standard reports:
- GA4 → Reports → Acquisition
- Usually unsampled (up to limits)
- Note the numbers
Create custom exploration:
- GA4 → Explore → Free form
- Use same dimensions/metrics
- Compare numbers to standard report

Look for discrepancies:

Standard report: 10,000 sessions
Custom report: 9,500 sessions (95% sampled)
Difference: 500 sessions (5% variance)

What to Look For:

Different numbers between reports
Sampling indicator on explorations
Variance percentage
Inconsistent results on refresh

Method 3: Check Property Session Volume

Review daily session count:
- GA4 → Reports → Realtime
- Note daily session volume
- Compare to GA4 limits

GA4 sampling thresholds:

Standard property (free):
- 10M events per month
- Sampling may occur in explorations
- Standard reports generally unsampled

360 property (paid):
- Higher limits
- Less sampling
- Unsampled reporting available

Calculate if you're over limits:

Daily sessions: 50,000
Monthly sessions: 1,500,000
Average events per session: 10
Monthly events: 15,000,000 (exceeds 10M limit)
Result: Sampling likely

What to Look For:

Sessions approaching or exceeding limits
Event count near monthly threshold
Frequent sampling in explorations
Date range affecting sampling

Method 4: Test Date Range Impact

Create test exploration:
- GA4 → Explore → Free form
- Add dimension: Session source
- Add metric: Sessions

Try different date ranges:

Last 7 days: 100% data (unsampled)
Last 30 days: 50% data (sampled)
Last 90 days: 25% data (sampled)

Note sampling threshold:
- Identify date range where sampling starts
- Document session threshold
- Plan analyses accordingly

What to Look For:

Date range where sampling kicks in
Session count triggering sampling
Variance with shorter ranges
Consistent sampling patterns

General Fixes

Fix 1: Reduce Date Range in Reports

Analyze shorter time periods:

Use shorter date ranges:

Instead of: Last 90 days (sampled)
Use: Last 30 days (unsampled)
Or: Weekly reports combined manually

Break analysis into chunks:

// Pseudo-code approach
Week 1: Jan 1-7 (unsampled)
Week 2: Jan 8-14 (unsampled)
Week 3: Jan 15-21 (unsampled)
Week 4: Jan 22-28 (unsampled)

Combine results manually or via API

Schedule regular exports:
- Export weekly unsampled data
- Combine in spreadsheet/database
- Analyze complete dataset offline

Fix 2: Use Standard Reports Instead of Explorations

Leverage pre-calculated reports:

Standard reports are less sampled:
- GA4 → Reports → Life cycle
- Pre-aggregated data
- Usually unsampled up to higher limits
Customize standard reports:
- Add secondary dimensions
- Apply filters
- Use comparison mode
- Still less sampling than explorations
When you must use explorations:
- Simplify dimensions (fewer breakdown)
- Reduce segments
- Limit filters
- Decrease date range

Fix 3: Upgrade to GA4 360

Consider paid version for high traffic:

GA4 360 benefits:

Standard GA4: 10M events/month
GA4 360: 1B events/month (100x more)

Standard: Sampling in explorations
GA4 360: Unsampled reports, unsampled explorations

Standard: Best effort support
GA4 360: Dedicated support, SLA

When to upgrade:

- Exceeding 10M events/month consistently
- Seeing frequent sampling
- Need accurate segment analysis
- Critical business decisions depend on data
- Multi-property roll-ups needed

Cost vs benefit:

GA4 360 pricing: $50,000 - $150,000/year
Consider if:
- Annual revenue > $10M
- Data accuracy critical
- Large marketing budget
- Enterprise analytics needs

Fix 4: Use BigQuery Export

Export raw, unsampled data:

Enable BigQuery export:
- GA4 → Admin → BigQuery Links
- Link to BigQuery project
- Choose daily or streaming export
- All events exported (unsampled)

Query unsampled data in BigQuery:

-- Example: Get unsampled session data
SELECT
  user_pseudo_id,
  event_name,
  event_timestamp,
  (SELECT value.string_value FROM UNNEST(event_params)
   WHERE key = 'session_id') AS session_id
FROM
  `project.analytics_PROPERTY_ID.events_*`
WHERE
  _TABLE_SUFFIX BETWEEN '20240101' AND '20240131'
  AND event_name = 'page_view'

Benefits of BigQuery:
- 100% unsampled data
- Raw event-level data
- Custom analysis without limits
- Join with other data sources
- Machine learning possible

BigQuery costs:

Storage: ~$0.02/GB per month
Queries: $5 per TB processed

Example site (1M sessions/month):
- Storage: ~$2-5/month
- Queries: ~$10-50/month
Total: $12-55/month (much less than 360)

Setup instructions:
- Create Google Cloud project
- Enable BigQuery API
- Link from GA4
- Wait 24 hours for first export
- Start querying

Fix 5: Reduce Event Volume

Optimize tracking to stay under limits:

Audit unnecessary events:

// Remove excessive scroll tracking
// Bad - tracks every 10%
window.addEventListener('scroll', function() {
  const scrollPercent = Math.round(window.scrollY / document.body.scrollHeight * 100);
  if (scrollPercent % 10 === 0) {
    gtag('event', 'scroll', { percent: scrollPercent });
  }
});

// Good - tracks only meaningful milestones
let tracked25 = false, tracked50 = false, tracked75 = false;
window.addEventListener('scroll', function() {
  const scrollPercent = window.scrollY / document.body.scrollHeight * 100;
  if (scrollPercent > 75 && !tracked75) {
    gtag('event', 'scroll', { percent: 75 });
    tracked75 = true;
  } else if (scrollPercent > 50 && !tracked50) {
    gtag('event', 'scroll', { percent: 50 });
    tracked50 = true;
  } else if (scrollPercent > 25 && !tracked25) {
    gtag('event', 'scroll', { percent: 25 });
    tracked25 = true;
  }
});

Consolidate similar events:

// Bad - separate event for each product view
gtag('event', 'view_product_1', {...});
gtag('event', 'view_product_2', {...});

// Good - single event with parameter
gtag('event', 'view_item', {
  item_id: 'product_1'
});

Remove debug events in production:

// Only fire debug events in development
if (window.location.hostname === 'localhost') {
  gtag('event', 'debug_checkpoint', {...});
}

Sample client-side events:

// Sample low-value events (track only 10%)
if (Math.random() < 0.1) {
  gtag('event', 'low_value_interaction', {...});
}

Fix 6: Use Data API for Unsampled Data

Extract data programmatically:

GA4 Data API:

// Node.js example using GA4 Data API
const {BetaAnalyticsDataClient} = require('@google-analytics/data');

const analyticsDataClient = new BetaAnalyticsDataClient();

async function runReport() {
  const [response] = await analyticsDataClient.runReport({
    property: `properties/${propertyId}`,
    dateRanges: [
      {
        startDate: '30daysAgo',
        endDate: 'today',
      },
    ],
    dimensions: [
      { name: 'sessionSource' },
    ],
    metrics: [
      { name: 'sessions' },
    ],
  });

  // Process unsampled data
  console.log('Report result:');
  response.rows.forEach(row => {
    console.log(row.dimensionValues[0].value, row.metricValues[0].value);
  });
}

runReport();

Benefits:
- Unsampled for most reports
- Automated data extraction
- Custom dashboards with fresh data
- Integration with other systems

API limitations:

- 10,000 rows per request (pagination needed)
- Complex explorations may still sample
- Rate limits apply
- Requires coding knowledge

Fix 7: Create Filtered Views with Subproperties

Divide traffic across properties:

Create multiple GA4 properties:

Main property: All traffic
Property A: North America traffic only
Property B: Europe traffic only
Property C: Mobile app only

Conditional tracking:

// Send to different properties based on criteria
let propertyId;

if (userLocation === 'north_america') {
  propertyId = 'G-AAAAAAAAAA';
} else if (userLocation === 'europe') {
  propertyId = 'G-BBBBBBBBBB';
} else {
  propertyId = 'G-CCCCCCCCCC';
}

gtag('config', propertyId);

Benefits:
- Each property has lower volume
- Reduced sampling
- Focused analysis per region/segment
Drawbacks:
- More complex setup
- Cannot easily compare across properties
- More properties to manage
- Consider carefully before implementing

Platform-Specific Guides

Detailed implementation instructions for your specific platform:

Platform	Troubleshooting Guide
Shopify	Shopify Sampling Issues Guide
WordPress	WordPress Sampling Issues Guide
Wix	Wix Sampling Issues Guide
Squarespace	Squarespace Sampling Issues Guide
Webflow	Webflow Sampling Issues Guide

Verification

After implementing fixes:

Check sampling status:
- Create test exploration
- Check for sampling badge
- Note percentage improvement
- Document which fixes worked
Verify BigQuery export:
- Check BigQuery console
- Verify daily tables created
- Run test query
- Confirm event counts match
Monitor event volume:
- GA4 → Configure → DebugView
- Check events per session
- Calculate monthly projection
- Verify under 10M limit
Compare report accuracy:
- Run same report multiple times
- Results should be consistent
- No sampling indicator
- Confidence in data restored

Common Mistakes

Not checking for sampling indicator - Unaware data is sampled
Using long date ranges unnecessarily - Triggers sampling
Over-tracking low-value events - Exceeds event limits
Not considering BigQuery - Free/low-cost unsampled access
Complex explorations on large datasets - Always sampled
Not using API for large exports - Missing unsampled option
Ignoring event volume optimization - Wasteful tracking
Not understanding GA4 limits - Surprised by sampling
Assuming all reports unsampled - Explorations more likely sampled
Not documenting sampling patterns - Cannot optimize

Troubleshooting Checklist

Sampling indicator checked in reports
Monthly event volume calculated
Under 10M events/month (or 360 considered)
Unnecessary events removed
Date ranges optimized
Standard reports used when possible
BigQuery export enabled
Data API utilized for automation
Event sampling implemented for low-value events
Regular data exports scheduled
Sampling patterns documented
Team trained on sampling awareness

Sampling Severity Levels

No Sampling: 100% of data

Ideal state
Full accuracy
Complete confidence in analysis

Light Sampling: 90-99% of data

Minimal impact
Generally acceptable
Small variance in results

Moderate Sampling: 50-89% of data

Noticeable impact
Use caution with decisions
Consider alternative approaches

Heavy Sampling: < 50% of data

Significant accuracy concerns
Do not trust for critical decisions
Implement fixes immediately
Consider BigQuery or 360

Additional Resources

↑ Back to top