Customer stories
Real results.
Real workloads.
How engineering teams at some of the world’s most data-intensive companies use PhoenixAI to power their most demanding workloads.
5×
avg. query speedup
<1s
latency
Intuit
100K+
events per second
2s
total latency for real-time ML predictions
35–60×
faster marketing aggregations
How Intuit Achieved Sub-4-Second Real-Time Analytics at 100K+ Events Per Second
2-second total latency for real-time ML predictions at 100K+ events per second, and marketing aggregations over 10B+ records cut from 5 hours to 5 minutes after replacing Apache Druid.
3×
cost-performance gain
Delivering Faster Analytics at Pinterest
Partner Insights migrated from Druid to StarRocks: 50% lower p90 latency on 32% of the instances, with 10-second data freshness for advertiser dashboards.
Shopee
3–10x
faster join queries than Presto
How Shopee 3xed Their Query Performance With StarRocks
Compared with Presto on Hive, StarRocks ran Shopee's API queries 10 to 2,000x faster, join queries 3 to 10x faster on 60% less CPU, and ad-hoc queries 2 to 3x faster in tests.
Roblox
100%
of LLM traffic captured
How Roblox Built LLM Observability with OpenTelemetry, Kafka and StarRocks
Roblox stores LLM logs and traces in StarRocks with 100% sampling, across 50+ LLM services, 12 coding harnesses and 50 models.
Fanatics
~90%
Snowflake cost savings
From Snowflake to StarRocks+Apache Iceberg: How Fanatics Cut 90% of Analytics Cost at 6PB Scale
About 90% Snowflake cost savings, up to 95% less Snowflake usage, and ad-hoc queries on Iceberg 10x faster than in Athena, with petabytes of Iceberg data behind one StarRocks query layer.
Tencent Games
50%
higher development efficiency
How Tencent Games unified real-time and batch game analytics with StarRocks
Tencent Games replaced a Lambda architecture with StarRocks over Apache Iceberg: 50% higher development efficiency, and a 14-billion-row by 1-billion-row JOIN in 44 seconds on a 300-node cluster.
Coinbase
30K/s
Kafka ingestion into PK tables
How Coinbase Powers Blockchain Analytics at Scale with StarRocks
573 billion rows across 300+ tables and 10 blockchains. Sub-second analytics with 30K Kafka messages/sec into primary key tables — for fraud detection and compliance.
Tencent
2x
faster similarity search
How Tencent Runs Vector Search for RAG and AIGC Image Management on StarRocks
Tencent moved OLAP and vector search for its AIGC image library onto StarRocks: 2x faster similarity search at half the total cost, plus RAG retrieval for its chatbots.
Grab
1
store for real-time and historical Spark metrics
How Grab Rebuilt Its Spark Observability Platform on StarRocks, Replacing InfluxDB
Grab replaced InfluxDB with StarRocks for Spark job monitoring. Kafka data is queryable within seconds, and 33 days of job runs are analyzed in StarRocks instead of the data lake.
Airtable
2B rows/min
average export ingestion rate
How Airtable Loaded Nearly 2 Trillion Rows into StarRocks for Archive Validation
Airtable loaded its RDS export into StarRocks in under 10 hours at about 2 billion rows per minute, and about 1 trillion archive rows in about 3 days, to validate its archive.
Celonis
100+
table joins per query
How Celonis Built a Process Mining Platform on StarRocks to Handle 20+ Billion Row Queries
PQL queries that join 100+ tables across 20–30 billion-row datasets, with sub-2-second dashboard responsiveness and hundreds of custom native functions.
Demandbase
PB-scale
lakehouse on Apache Iceberg
How Demandbase Built a Scalable Data Lakehouse with StarRocks and Iceberg
Petabyte-scale data on Apache Iceberg, near real-time freshness through partial routine loads, and disaster recovery and blue-green upgrades that load new StarRocks clusters from Iceberg.
Conductor
<1s
query latency on terabytes
How Conductor Built Hallucination-Free Agentic Analytics on PhoenixAI Cloud
Sub-second query performance on terabytes of data, idea to production in under two months, and decimal-level accuracy for AI agents — Conductor's split-reasoning architecture for hallucination-free agentic analytics.
iQIYI
33x
faster P50 ad-hoc query latency
How iQIYI Runs Ad Analytics and Ad-hoc Lakehouse Queries on StarRocks
iQIYI moved ad analytics and ad-hoc lakehouse queries to StarRocks: 400% better ad performance, 33x faster P50 latency for migrated Spark SQL queries, 4.6 person-days of analyst time freed.
NAVER
6x
faster JOIN query with a materialized view
How JOIN Changed How We Approach Data Infra At NAVER
On NAVER's own data, StarRocks completed every JOIN benchmark query while ClickHouse failed four of five. A materialized view then made a JOIN query 6x faster with no SQL changes.
Alibaba Group
T+1→min
data freshness
How Alibaba Runs Near-Real-Time Flash Sale Analytics on StarRocks and Paimon
Taobao Flash Sales went from T+1 offline reports to minutes-level analytics on Apache Paimon and StarRocks, with materialized views refreshed every 10 minutes for BI dashboards.
Cisco Webex
50%
lower average query latency vs. Trino
How Cisco Webex Unified Pinot and Trino Workloads on StarRocks
Cisco Webex moved Pinot and Trino workloads to StarRocks: about 50% lower average query latency than Trino, over 10× faster common queries with materialized views, about 80% less disk with Flat JSON.
Demandbase
49→1
cluster consolidation
Demandbase Ditches Denormalization By Switching off ClickHouse
49 ClickHouse clusters replaced with a single 45-node StarRocks cluster — 60% fewer nodes, 90% less storage, and ETL pipelines for denormalization gone.
RedNote
10,000+
QPS per cluster
RedNote Delivers Integrated Advertising Analytics with StarRocks
More than 10,000 QPS per cluster with TP99 latency under 100 ms in stress tests. RedNote built an integrated advertising analytics platform on StarRocks.
DiDi
80%
lower overall cost
How DiDi Replaced Apache Druid, Presto and ClickHouse with StarRocks
DiDi's monitoring and alarm system went from 60+ nodes to fewer than 10 on StarRocks, with overall cost down more than 80% and P90 query time from 500 ms to 150 ms.
Zepto
30M+
rows ingested per day
How Zepto Built Real-Time Brand Analytics for Quick Commerce with StarRocks
Zepto serves partner-brand dashboards from StarRocks: sub-second queries for external users, a 300M+ row sales table, and 30M+ rows a day streamed in through Kafka Routine Load.
Amazon
<5s
standard queries at 1,000 concurrent users
How Amazon Stores FinTech runs financial OLAP on StarRocks and Amazon EKS
Amazon Stores FinTech runs StarRocks on Amazon EKS: under 5 s for standard and under 20 s for complex queries at 1,000 concurrent users, 3X–5X ClickHouse's multi-join throughput.
SmartNews
3.6×
faster read queries than Trino
How SmartNews Is Replacing Trino and ClickHouse with PhoenixAI Cloud
In SmartNews' tests, StarRocks served 800 QPS on an advertiser workload where ClickHouse failed, and ran replayed Trino read queries 3.6× faster at the same cluster size.
CTW
<10s
audience data freshness
How CTW Powers Real-Time Ad Targeting, BI, and AI Analytics with PhoenixAI
Audience segment refresh from hours to seconds, 3× faster BI dashboards, and Doraemon — CTW's internal AI agent — answering business questions with 90%+ accuracy.
Yuno
~5s
data freshness, down from 1 hour
From Hours to Seconds: How Yuno Accelerated Customer-Facing Analytics By Switching Off Snowflake and Athena
Data freshness from 1 hour to ~5 seconds, most dashboards from ~3 seconds to under 1 second, and over 40% lower compute cost per query after moving off Snowflake.
Lenovo
1/4
of the Presto cluster's resources
Lenovo Simplifies Data Processing and BI Analysis with StarRocks
StarRocks replaced a dozen nodes in a Presto cluster with 1/4 of the resources at Lenovo's Liansheng Zhida, and Tableau charts that used to load slowly now appear in seconds.
Herdwatch
0.7–1.5s
dashboard query latency
Herdwatch Enables Customer-Facing Analytics Using Apache Iceberg and PhoenixAI Cloud
Dashboard queries went from 2–5 minutes on Athena to 700 ms–1.5 s on StarRocks, with sub-second loads for multi-query dashboards and one Iceberg lake for all regions.
SplitMetrics
30.29→3.35s
P99 query latency
How SplitMetrics Replaced PostgreSQL with StarRocks for Customer-Facing Analytics
P99 dashboard latency down from 30.29s to 3.35s and 50% lower infrastructure cost after moving customer-facing dashboards from PostgreSQL to StarRocks.
TRM Labs
50%
better P95 response times
How TRM Labs Built a Petabyte-Scale Lakehouse on Apache Iceberg and StarRocks
TRM Labs moved customer-facing analytics from BigQuery to an Apache Iceberg and StarRocks lakehouse: 50% better P95 response times and 54% fewer query timeout errors.
Eightfold.ai
2×
lower latency, lower cost
Customer-Facing Analytics Meets Agents: Eightfold's Move from Redshift to StarRocks
Sub-second, multi-tenant analytics for humans today and AI agents tomorrow. Eightfold doubled performance and halved cost by migrating from Redshift to StarRocks.
Trip.com
200ms
average query response
Trip.com Uses StarRocks to Power High-Concurrency, Low-Cost Queries
70% of Trip.com's real-time lodging BI scenarios now run on StarRocks, with about 200 ms average response time and only 1% of queries taking more than 500 ms.
vivo
65%
lower P50 ad-hoc query latency
How vivo Accelerates Lakehouse Queries for Ad-Hoc Analysis and BI with StarRocks
vivo moved ad-hoc lakehouse queries from Presto to StarRocks: P50 latency down 65.06%, Agile BI P90 from 16 s to under 5 s, and R&D data available in under 3 minutes instead of hours.
Zepp Health
100ms
average query latency
Zepp Health Reinvigorates Smart Wearable Data with StarRocks
Zepp Health moved event tracking analytics from Apache HBase to StarRocks: about 100 ms average and 250 ms P99 latency, large queries from minutes to seconds, 20% lower storage cost.
SF Technology
1/3
of the resources for the financial system
SF Technology Chooses StarRocks for Real-Time Analytics
After moving to StarRocks, SF Technology's financial system runs with only 1/3 of the resources, and StarRocks serves the 2,000 sets of BI reports for delivery services.
Fresha
200ms
home-page analytics queries
How Fresha moved customer-facing analytics off Postgres with StarRocks
Home-page analytics for Fresha's largest partners went from 15–20 s loads or timeouts on Postgres to about 200 ms on StarRocks, with minute-level freshness.
PlaySimple Games
30×
faster funnel load
How PlaySimple Games Built User-Level Funnel Analytics with StarRocks
PlaySimple Games cut end-to-end funnel load time from ~1 hr to ~2 min with StarRocks on Delta Lake; heavy data queries run ~2× faster, and 6 of 8 test funnel queries beat Trino.
Haezoom
1.74x
higher throughput
How Haezoom and CloudShift Moved Solar Energy Analytics from Apache Druid to StarRocks
Haezoom replaced Apache Druid with StarRocks: 1.74x more requests served, 44.3% lower average response time, up to 4x faster complex queries and 30% lower infrastructure cost.