Bigabid Case study

Bigabid uses Intel Granulate to reduce bid timeouts, latency and compute costs 

Download Case Study

The Challenge

Bigabid is an advertising technology company specializing in mobile user acquisition & re-engagement for gaming, dating and productivity apps. The company’s automated, real-time optimization tools, focusing on post-install events, harness the power of machine learning to give its top-tier clients an edge in a time-sensitive industry.

Bigabid’s AI-driven media buying engine is at the heart of the operation. To optimize the performance of its high frequency real-time programmatic bidder, Bigabid needed to reduce latency, dropped requests, and bid timeouts—without succumbing to a runaway increase in server costs.

Their environment is based on AWS spot instances, running a multi-tiered, C++ application with an RTB (Real-Time Bidder) module. The application handles over 80,000 requests per minute with each request needing to be processed within 120ms. Doubling the workloads they support and capturing revenue lost due to bid timeouts, while not adding costs, was going to be a challenge.

We achieved impressive improvements, with no time and effort. Within seconds from installation, Granulate was able to eliminate 90% of bid timeouts. We managed to reduce compute costs by 60%.
Amit Attias, CTO

Why Intel Granulate

Needing to meet their challenges with minimal impact on staff, Bigabid turned to Intel Granulate’s real-time, continuous optimization solution.

After Intel Granulate was installed and able to observe Bigabid’s application, it automatically optimized the allocation and release of memory. These changes reduced memory fragmentation and enabled greater concurrency, delivering better application performance.

Results

After a short POC on two machines, Bigabid decided to deploy Granulate throughout its network. Within 5 days, Bigabid vastly improved performance, reducing cloud cost by 60%, eliminating 90% of bid timeouts and reducing p95 latency by 70%. The following graph shows the steep drop-off in latency:

70% Reduced p95 Response Time
Figure 1: Bid response time before and after installing Intel Granulate

90% Reduced Bid Timeouts
Figure 2: Number of timeout bids before and after Intel Granulate

90% Reduced Dropped Requests
Figure 3: Number of dropped requests before and after Intel Granulate

90%
Fewer Bid Timeouts
60%
Reduced Cloud cost
70%
Reduced P95 Latency

AdTech
HQ: New York, USA
Employees: 100+

Bigabid, a leader in the Demand-Side Platform (DSP) market, specializes in mobile user acquisition and re-engagement campaigns for games, casino, dating and instant-services apps. They help their customers acquire, engage, and retain high value users with scientifically-based advertising.

Their leading technology analyzes millions of data fragments in real-time, providing Bigabid with the ability to automatically bid on the right advertising placement at the right time for their customers.