Hawk is the first independent European platform to offer a transparent and technological advertising experience across all screens: Desktop, Mobile, CTV, DOOH & Digital Audio.
Their clients, media advertisers and agents, are on a constant lookout for more efficient digital strategies: they offer them access to the best locations on mobile and desktop screens, on digital advertising boards, audio platforms or connected TVs.
Their solution meets the needs of traders wishing to set up effective and impactful advertising campaigns:
- At first, traders need to understand the available traffic, the market trends, the current opportunities in order to set up the most efficient and performant strategies.
- Once the campaign is live, traders are very attentive to real-time results, and possible optimisations. They want to monitor their specific KPIs, and understand how they're progressing by playing on the different business levers at their disposal.
- Campaign ending goes with reporting. Traders need a holistic view of their campaign, check their KPIs, compare the performance with previous runs and draw conclusions for future campaigns.
At every stage, Hawk account managers need precise tools to assist their users, regardless of their level of experience, by giving them relevant insights in a fast and intuitive way. But with more than 50 billion messages processed every day, providing a real-time analysis tool is a true and highly rewarding challenge.
Druid as a Foundation stone
To meet this challenge, they started in 2016 with the open-source Druid distribution, hosted by themselves on AWS, as the foundation stone of their data/analytics platform.
But keeping a stable and performant service was a daily challenge. Druid is a complex solution difficult to grasp, constantly evolving, that needs to be heard by qualified data-engineers.
QoS was decent in the early years, but with extensive use by more and more users, they quickly spotted the limits of their approach: despite the mobilization of increasing resources, the sub-second commitment was an old memory.
Led by its CTO Julien Lafont, the company undertook in 2018 a major modernisation of its big-data platform. The goal is to enable each team to produce value from its data. It has to be simple, fast and adapted to all kinds of analysis, while reducing infrastructure costs.
This modernisation is all the more important because it must go hand in hand with the significant growth of the company, especially by opening its analytics tools to external users.
The new architecture will consist of two key pillars:
- A real-time analytics solution that will ingest the data from the streaming pipeline, and offer a rich data-mining interface.
- A cloud data-warehouse capable of storing the raw, anonymised data for heavy workloads.
After comparing Druid to other technologies on the market, they decided to reiterate their confidence in this solution, but by relying on external expertise to drastically reduce its operational cost, and thus allow their data-engineering team to invest its time on new projects.
Imply has been considered as a partner for the new managed Druid cluster along with other commercial solutions. We were selected for our expertise on the Druid solution, for our investment through the open-source contributions to Druid core, as well as for the tools we have developed and made available to our clients.
"The feature that we immediately loved is the Cluster Manager. With it, we can start a new Druid cluster in our AWS VPC or launch a major update in one click. Before, it was a time-consuming and dreaded operation for our data team."
The Pivot UI has also been greatly appreciated by their users, including the ability to create dashboards, to schedule reports or create automatic alerts.
The migration of their Druid cluster to the Imply Cloud solution helped them to match several business requirements
- Offering to their customers an all-in-one solution to follow their campaign KPIs throughout their lifecycle
- Allowing their customers to build business dashboards, and visualize the evolution of their KPIs
- Giving to each media-trader the capacity to setup tailored reports and send them to the right people in their organization with the automatic reporting module
- Accelerating the Time-To-Market thanks to a simplified data pipeline, with Druid directly connected to their Kafka cluster and RDS database.
In the future, they plan to replace their Aerospike NoSQL cluster with a high-performance Imply cluster that will be directly accessed through the Druid APIs. This migration will allow them to reduce the complexity of their Data infrastructure while providing new opportunities.