Frequently Asked Questions¶

Questions are tagged by audience: [Buyer] for media buyers and campaign managers, [DevOps] for platform engineers, [Both] for shared questions.

[Buyer] Why is my coverage percentage below 100%?¶

Coverage measures how many date x report-type cells have data vs. how many are expected. Common reasons for gaps:

Google didn't send a report for that date (public holiday, export lag).
Gmail import missed the email (check Gmail status).
A specific report type is not available for your seat (e.g., quality data may not exist for all buyers).

Check the data freshness grid on /import to see exactly which cells are missing. See Data Import.

[Buyer] What's the difference between "waste" and "low win rate"?¶

Waste = bid requests your bidder rejected without bidding. This is QPS you paid for but couldn't use at all. Fix it with pretargeting.

Low win rate = bid requests your bidder bid on but lost the auction. This means your bids are not competitive enough. Fix it with bid strategy, not pretargeting.

Both show up in the funnel but require different actions. See Understanding Your QPS Funnel.

[Buyer] Can I undo a pretargeting change?¶

Yes. Go to /history, find the change, click "Preview rollback" to see what will revert, then confirm. The rollback itself is recorded. See Pretargeting Configuration.

[Buyer] How often should I re-import data?¶

Daily. Gmail auto-import handles this automatically. If you're importing manually, do it once per day after reports arrive. Stale data means stale decisions.

[Buyer] What does the optimizer actually change?¶

The optimizer proposes changes to your pretargeting configs: adding or removing geos, sizes, publishers, etc. It never applies changes automatically. You review and approve each proposal. See The Optimizer.

[DevOps] Why did the runtime health strict gate fail?¶

Check the workflow logs: gh run view <id> --log-failed. Look for FAIL vs. BLOCKED:

FAIL = something broke. The data-freshness timeout and SET statement_timeout issues are common culprits. See Troubleshooting.
BLOCKED = a dependency is missing, not necessarily a code bug. Examples: no quality data for this buyer, proposal has no billing_id. Compare against previous runs to distinguish regressions from pre-existing gaps.

[DevOps] Why is the data-freshness endpoint slow?¶

The query scans rtb_daily (~84M rows) and rtb_bidstream (~21M rows). If the query plan degrades to a sequential scan instead of using the (buyer_account_id, metric_date DESC) indexes, it will take minutes.

Fix: ensure queries use the generate_series + EXISTS pattern (14 index lookups instead of full table scan). See Database Operations.

[DevOps] How do I check what version is deployed?¶

curl -sS https://scan.rtb.cat/api/health | jq -r '.git_sha,.version'

This returns the git SHA and image tag. Compare against your commit log.

[DevOps] How do I deploy a fix?¶

Push to unified-platform
Wait for build-and-push.yml to succeed
Trigger deploy.yml via gh workflow run with confirm=DEPLOY
Verify with /api/health

See Deployment for the full procedure.

Check Cloud SQL Proxy: sudo docker ps | grep cloudsql. If it's down, restart it, wait 10 seconds, then restart the API container. See Troubleshooting for the full procedure.

[Both] Where does Cat-Scan's data come from?¶

Google Authorized Buyers CSV exports. There is no Reporting API. Data arrives either by manual CSV upload or automatic Gmail ingestion. See Data Import.

[Both] Is it safe to re-import the same CSV?¶

Yes. Every row is hashed and deduplicated. Re-importing never double-counts.

[Both] What languages does the UI support?¶

English, Dutch, and Chinese (Simplified). The language selector is in the sidebar.