Since we're sending scraper output data through EventGate, it will be subjected to the standard Wikimedia 90-day data retention window unless we configure a custom sanitization or sanitization allowlist mapping. This task is to identify whether any sensitive data is present, and to decide how we handle the analytic data. The main alternatives are either:

• that we keep only the aggregated result;
• or preserve several (3) months of detailed, per-page analysis.

Outcome

• Freeform text such as template names: 30 days: each month, we'll drop previous results and write the new data to dedicated tables. Any oversighted content will disappear with the next scrape.
• Per-page summaries: 90 days: this will follow the default EventGate data retention window.
• Per-wiki summaries: Indefinite: no plans to delete. These are purely numeric aggregations of public information and have no sensitive data.