Unacast: staying ahead of explosive growth with Google Cloud Platform

Location data drawn from mobile devices is often too imprecise for use by marketing and business. Proximity sensors offer a much greater degree of accuracy, but produce a wide range of different data types. By sorting and aggregating proximity sensor data and other location data into its Real World Graph™, Unacast has created a reliable resource for businesses who want to personalise messages to consumers. Now the world's biggest platform of proximity data, Unacast analyses large amounts of real world data from multiple partners, runs it through a proprietary scoring and verification engine and presents it in a coherent and trusted format. Because the proximity sensor industry is growing at high speed, Unacast was built on Google Cloud Platform for optimal scalability, minimal DevOps, and maximum flexibility.

“In only eight weeks, we built a data ingestion pipeline that fits all of our needs and can scale almost infinitely with Google Cloud Platform. I’ve thrown data at every part of it and it’s scaled perfectly from day one. GCP has helped us take giant leaps in working with more and more different “real world" data types." - Andreas Heim, VP of Engineering, Unacast

NoOps and optimal scaling

The number of proximity sensors worldwide is growing by over 30% every quarter, and the amount of proximity data exponentially, meaning proximity data platforms have to scale fast to keep pace with industry growth. Unacast stays ahead of its competitors by nurturing internal innovation and building new partnerships, instead of losing time to operations and maintenance. So when Unacast looked for a cloud provider, it knew it needed a scalable, managed solution that could give its team space to work on development.

Unacast chose Google Cloud Platform as the basis for its entire platform. They built a data processing pipeline based on Google App Engine, Google Cloud Pub/Sub, Google Cloud Dataflow and Google BigQuery. Because all of the GCP services are managed, Unacast devotes no personnel to operations, reassured about stability and security by Google’s internal use of its own tools and infrastructure. Unacast built its architecture on Google Kubernetes Engine with Kubernetes from the outset, and the platform now scales painlessly according to need. Dataflow delivers powerful responses to advanced jobs such as streaming, batch processing, large mapping analysis and major import jobs, and BigQuery has established itself at the core of the company’s work life.

“We didn't realize how powerful BigQuery was until we started using it extensively. The sheer speed, flexibility and user-defined functions are exceptional. We can run extremely complicated calculations in BQ in a couple of minutes instead of building a tool ourselves. We use it for prototyping data processing pipelines, analytics, learning and as our main storage. If we considered moving to another cloud platform, the thought of losing BigQuery would stop us. It's as friendly to developers as it is to BI staff, and because BQ’s really, really fast, we’ve never needed another type of database for our main storage.” - Andreas Heim, VP of Engineering, Unacast

Gearing up for the future with Machine Learning

By 2021, an estimated 500 million proximity sensors are set to be deployed worldwide, almost 40 times as many as exist today. Unacast is preparing for the future by focusing on Cloud Functions, Cloud Dataprep and Cloud Machine Learning Engine.

“We are always improving, onboarding different data sources and creating value by looking for how we can take advantage of new types of services that didn't exist when we started out. Google Cloud Platform makes it easy to build new products on top of the data ingestion and analysis platforms that we already have. The products fit together like Lego.” - Andreas Heim, VP of Engineering, Unacast