Skip to content
  • There are no suggestions because the search field is empty.

How Outlier Detection Works in Convert Reports and Exports

Understanding how Convert handles extreme data points with outlier capping to ensure accurate, reliable A/B testing reports and exports.

IN THIS ARTICLE YOU WILL:

🎯 What Is Outlier Detection?

Outlier detection in Convert helps prevent unusually high or low order values from skewing your experiment metrics—like Revenue per Visitor (RPV), Average Order Value (AOV), and Product per Visitor (PPV).

You can configure outlier handling under:

Experience Summary > Stats & Settings > Order Outliers

You may read more about it here: Setting limits for Order Outliers

Supported outlier types include:

  • Minimum/Maximum thresholds (e.g., from $1 to $10,000)
  • Percentile thresholds (e.g., exclude orders below the 5th or above the 95th percentile)

Both Order Value and Number of Products Ordered can be filtered.

⚠️ Important: Outliers Are Capped, Not Removed. A common misconception is that Convert deletes outlier data. In reality:

➡️ Outliers are not removed—they are capped at the maximum or minimum values you define. For example, if you set a maximum order value of $10,000, and a visitor makes a purchase of $25,000, that value is capped at $10,000 in your reports and calculations.

🧮 Why Convert Uses Capping (Not Removal)

Capping outliers instead of removing them is a widely accepted practice in statistics. It provides several key benefits:

  • Preserves statistical integrity by keeping the full dataset intact
  • Reduces bias that might arise from excluding extreme values
  • Ensures consistency between reported numbers and raw data exports

Removing outliers entirely can reduce your sample size and lead to misleading metrics, especially in experiments with low traffic.

📊 Where Are Capped Values Used?

Capped values are used in:

  • All reporting dashboards
  • All statistical models behind metrics like RPV, AOV, and PPV

So, what you see in the Convert reports already reflects your configured outlier limits.

Example:

Let’s say your outlier setting caps Order Value at $10,000:

  • A purchase of $18,500 will be shown as $10,000 in the report.
  • This capped value is also what’s used in the RPV and other metric calculations.

📁 What About Raw Data Exports?

In exported raw data files:

  • You’ll see two columns:
    • The original value (e.g., $18,500)
    • The processed/capped value (e.g., $10,000)

This dual-column structure offers transparency while preserving the statistical treatment used in reporting.

💡 If your Excel shows “10,000.00” instead of “10000”, that’s just a formatting preference—it doesn’t affect the calculation.

✅ QA Test Case for Outlier Detection

Here’s how Convert verifies this feature internally:

Test Setup:

  • Order Value filter: Percentile 5–95
  • Product Count filter: Min 1, Max 50

Expected Behavior:

  • Orders outside these ranges are capped, not excluded.
  • Reports use the capped values.
  • Raw data exports show both original and capped values.

❓ Still Seeing Outliers in the Raw Export?

That’s expected!

✅ The original value is included for full visibility
✅ The capped value is what drives the reporting and statistical engine

If needed, verify that:

  • Your outlier settings are applied under Stats & Settings
  • You’re referencing the processed (not raw) column in Excel

Still unclear? Contact our support team with your experiment link, and we’ll assist you.