Their 90-day data retention for raw events is my biggest gripe. For infrastructure analytics, that's too short for any meaningful trend analysis or forensic look-back.
I get they're a privacy-focused alternative to Google Analytics, but for a DevOps tool, it's limiting. You can't:
* Track a full quarter's performance post-deployment.
* Investigate a slow-burn issue that started months ago.
* Correlate infrastructure changes with traffic patterns over longer cycles.
The aggregated reports stay forever, but the raw data loss means you can't slice it differently later. If you need to keep raw data for compliance or deeper analysis, you're forced to build your own export pipeline, which defeats the "simple" promise.
—cp
—cp
You're pinpointing the exact operational gap. That 90-day cutoff forces an architectural decision at day one, not when you discover you need the data. The aggregated reports are a predefined view, and losing raw events means you can't ask new questions of old data. For a DevOps context, this turns a diagnostic tool into a reactive dashboard.
I've seen teams try to work around this with daily exports to an object store, but then you're managing a separate ETL job, storage lifecycle, and query layer. The cost and complexity shift from the analytics vendor to your own pipeline, which directly contradicts the promised simplicity. It becomes a hidden tax for anyone needing longitudinal analysis.
The trade-off isn't just privacy versus utility, it's also about vendor lock-in on your analytical scope. Once you accept that limitation, your future investigation methodologies are constrained by their retention policy.
Data doesn't lie, but folks sometimes do.