appconfig-experiment-backfill

Generate mock Application Insights telemetry for testing Experimentation on Azure App Configuration.

This CLI backfills realistic custom events, page views, browser timings, and exception telemetry into an App Insights resource so you can exercise the experimentation analysis pipeline without real user traffic.

Note — This is a personal tool by @rossgrambo, not an official Azure App Configuration product.

Quick Start

npx appconfig-experiment-backfill \
  --flag=button-display \
  --feature-flag-reference="https://..." \
  --variants=get-started,sign-up \
  --connectionString="InstrumentationKey=..." \
  --users=10 \
  --days=1

Parameters

| Parameter | Required | Default | Description | |---|---|---|---| | --flag | Yes | — | Feature flag name | | --feature-flag-reference | Yes | — | Feature flag reference string | | --variants | Yes | — | Comma-separated variant names (e.g. control,treatment) | | --connectionString | Yes | — | Application Insights connection string | | --users | No | 100 | Number of simulated users per hour | | --days | No | 7 | Number of days to backfill | | --allocation-id | No | 123e4567-e89b-12d3-a456-426655440000 | Allocation ID GUID |

What Gets Generated

For each simulated user the tool emits:

1. FeatureEvaluation event — required by the experimentation pipeline, includes FeatureName, Variant, TargetingId, FeatureFlagReference, and AllocationId.
2. Custom events (~50 event types) — user interactions like UserClickedButton, CompletedPurchase, SearchedContent, etc., each with realistic measurements. Every event has a 90% chance of firing per user.
3. Page views (8 pages) — with browser timing measurements (totalDuration, networkDuration, etc.). Each page has a 70% chance of being visited per user.
4. Exceptions — simulated at a low rate (1% for the first variant, 3% for others).

Variant adjustments are applied to measurements so different variants produce statistically distinguishable data, letting you verify that the experimentation analysis pipeline detects meaningful differences.

License

MIT