Jump to content

Readers/Web/Metrics Platform Adoption/Sampling Rates

From mediawiki.org

Metrics Platform Transition

[edit]

Overview

[edit]

This document outlines the transition from the Event Platform to the Metrics Platform, including changes in sampling rates, event tracking, and schema structure.

Key Changes in Sampling Rates

[edit]

Standardization of Default Sampling Rates

[edit]

The default analytics sampling rate for both desktop and mobile is now 0.2 in the Metrics Platform.

Previously, mobile tracking was sampled at 0.1 → increased to align with desktop tracking.

Explicit Sampling Rate for OfficeWiki

[edit]

In the Metrics Platform, OfficeWiki's mobile sampling rate is explicitly set to 0, ensuring no mobile data is collected.

Configuration Files

[edit]

Event Platform (Legacy)

[edit]

Metrics Platform (Current)

[edit]

Before and After Comparison

[edit]

Event Platform (Legacy) Desktop Tracking

[edit]
'wgWMEDesktopWebUIActionsTracking' => [
    'default' => 0.2,
    'legacy-vector' => 0,
    'officewiki' => 0,
    'testwiki' => 1,   // 100% of events logged
    'enwiki' => 0.01,  // 1% of events logged
];

Metrics Platform (Current) Unified Sampling Rate

[edit]
'mediawiki.web_ui_actions' => [
    'schema_title' => 'analytics/mediawiki/product_metrics/web_ui_actions',
    'destination_event_service' => 'eventgate-analytics-external',
    'producers' => [
        'metrics_platform_client' => [...],
        'sample' => [
            'unit' => 'session',
            'rate' => 0.2,
        ],
    ],
];

Why Were These Changes Made?

[edit]
  • Combining desktop and mobile tracking → reduces redundancy and simplifies data analysis.
  • Increasing mobile sampling rate from 0.1 to 0.2 → ensures consistency with desktop tracking.
  • Explicitly setting OfficeWiki sampling to 0 → prevents accidental data collection.

Final Thoughts

[edit]
  • The transition to the Metrics Platform standardizes tracking and makes sampling rates more predictable.
  • Analysts should be aware of the changes when comparing past Event Platform data with new Metrics Platform data.