Join Celtra Director of Engineering and Analytics Grega Kešpre as he shares hard-won insights about harnessing the cloud to deliver a world-class data processing pipeline that handles over 2 billion analytics events daily.

Over 5,000 brands, more than 15,000 campaigns, over 188,000 ads representing 2 billion analytics events – that all adds up to a terabyte of data every day. Celtra, the fast-growing provider of creative technology for data-driven digital display advertising, has to handle a vast ocean of data every day, and they needed to harness the cloud with a high-performance data processing pipeline to manage it all.

The director of engineering and analytics at Celtra, Grega Kešpret, sat down with VentureBeat’s Evan Schuman and Jon Bock, VP of marketing and products at Snowflake — the data warehouse innovator backed by leading VC firms — to cover how big data technologies in the cloud can deliver a world-class data analytics solution for unstructured, messy, galaxy-sized data pools.

Choosing a cloud-based solution was a several-year project for his team, Kešpret says, and they went through several tries before landing on their final choice. The biggest challenge was the abundance of solutions out there. Expanding from a very narrow marketplace of just a couple of vendors, the market has exploded.

Finding the unique data warehouse solution that meets your needs sounds simple, but it’s too often overlooked by companies either overwhelmed by choice or who succumb to a salesperson’s siren song of bells and whistles.

But here’s the solutoin: Before you even consider shopping, you need to thoroughly understand your requirements, top to bottom. You need to know exactly what you need your data to do, and where you need it to go, and to whom. You need to know the state of your current system, and whether you can cobble your legacy architecture into a brand-new structure, or if it’s time to make the leap into the cloud.

(Hint: It’s almost certainly time to take that leap, because traditional data warehousing, which is expensive, complex, and inflexible, simply doesn’t meet modern needs.)

And Celtra needed flexibility above all, says Kešpret, which precomputed aggregates simply don’t provide in the face of unique counting, order statistics, outliers, and more.

Celtra wanted short development cycles, and faster experimentation. And they needed answers to questions that their existing data model did not support: everything from analyzing the effects of placement position on engagement rates to troubleshooting ad loading time performance.

Knowing what they required, and having a plan in place, allowed them to eliminate a number of platforms out of the gate, and narrow their focus on the cloud-based solutions that could not only support their current data load, but scale cost-effectively into the future as they continue to grow.

That meant data warehousing as a service. A SaaS data warehouse eliminates the overhead by shifting the burden of managing data partitioning, ensuring availability, and securing data and monitoring load onto the service provider in the cloud. It allows full, on-the-fly flexibility to support any size data pipeline and workloads.

Celtra turned to Snowflake, which is designed to parse data from any source, making it possible to combine structured and semi-structured data like JSON, Avro, or XML in a single query, and uses SQL to allow direct access to raw data.

And because in the end it’s all about the data, Celtra had found their unique data warehouse solution. The takeaway: every company has unique and sometimes extraordinary needs. How can you uncover yours — especially the hidden ones — and find the solution that meets your own complex data analytics?

Don’t miss out!

Access this VB Live event on demand right here.

You’ll learn more about how you can:

  • Build a simpler, faster solution to support your company’s data analytics load
  • Support diverse reporting and ad hoc analytics in one system
  • Take advantage of the cloud for flexibility, scaling, and simplicity


  • Evan Schuman, Moderator, VentureBeat
  • Grega Kešpret, Director of Engineering, Analytics, Celtra
  • Jon Bock, VP of Marketing and Products, Snowflake

    This VB Live event is sponsored by Snowflake.