UberEats — Merchant-Facing Ad Analytics
UberEats uses Apache Pinot's upsert capability to serve accurate real-time ad analytics to merchants, updating impressions, clicks, and conversions from a Kafka changelog.
Scale
Real-time impression, click, and conversion data for merchant dashboards across the UberEats platform
Before
Append-only OLAP systems unable to reflect real-time budget updates — dashboards were fast but inaccurate
After
Apache Pinot with upsert support: Kafka changelog updates existing records in real time → accurate live view
Key Insight
Upserts in a real-time OLAP system are non-trivial — most columnar engines are append-only. Pinot's upsert support is what enables 'accurate' dashboards (not just fast ones).
In a Snowflake Conversation
Upserts in a real-time OLAP system are non-trivial — most columnar engines are append-only. Pinot's upsert support is what enables 'accurate' dashboards (not just fast ones).
My Read
Practitioner commentary coming soon.
Relevant Conversations