icône 1

2 billion

logged transactions

icône 2

600 million+

transactions in 12 months

icône 3

- 400 %

processing time

The background

For several years now, coupon marketing has become a key strategy. It is flexible, customisable and financially affordable, and has helped brands boost their power at an incredible rate.

A major customer loyalty programme, based on distributing coupons, has completely reworked its digital platform. One of its objectives was to offer users the ability to load discount coupons directly onto the distributor’s loyalty card.

To help this feature take shape, they called upon the skills of Publicis ETO. This data-driven agency specialises in customer relationship management solution architectures, and supplies tools for data integration, management and recovery. It also helps brands design and tailor content for their omni-channel customer experience. Today, Publicis ETO manages 50 platforms (single customer databases, data lakes, customer loyalty programs, data management platforms, e-CRM tools) for popular brands in France, and around the world.

At Publicis ETO, the professionals that handle data are organised into two complementary skill types: data science and data management. Data scientists explore and analyse all of the data, to provide models that can transform customer data into actionable data. Data managers develop effective platforms that make the customer data and defined models (segments, clusters, scores, etc.) actionable.

The challenge

Enriching the customer experience with a big data platform

Publicis ETO overcame several challenges to ensure that they met their customer’s needs.

Developing the right solution

They needed to find a software ecosystem that was innovative, open-source, interoperable and scalable. The new ecosystem would need to help them orchestrate storage, and explore, process and manage data. It would also need to support the development of an environment that could offer a collaborative, secure workspace.

Building the infrastructure

The first part of the infrastructure needed to be on-premises (internal), in order to collect and organise personal data flows. The second part needed to be cloud-based (external), in order to give both data scientists and data managers an exploration space adapted to their needs in terms of resources (including scalability). By adopting this approach, the agency wanted to benefit from the scalability offered by this technology, while managing its costs.

Guaranteed security

Since the agency handles personal data, it was vital to have rules in place regarding data isolation, pseudonymisation and security. They also needed to ensure that these rules would be followed throughout the entire process.

The solution

Solutions that meet both customer requirements and agency specifications

There was a long list of limitations to take into account — data security, certification, hosting regulations in France, and much more.

The software layer

Publicis ETO decided to use the Hadoop framework — a tried and tested solution with a vast, varied ecosystem. Due to the complexity involved in assembling these technologies, they decided to use a distribution, and selected one from Cloudera, an open-source big data platform publisher. As a market leader and a contributing member of the Hadoop community, Publicis ETO assessed this solution and decided it suited their needs perfectly: a centralised platform, high security, encryption, and more. With the Cloudera Enterprise solution, they had a managed, scalable and secure cloud solution.

Publicis ETO software layer

The hardware layer

Publicis ETO chose OVH as their cloud provider. In addition to guaranteeing that the data would be hosted in France, the agency could build a varied infrastructure composed of Private Cloud solutions and HG Dedicated Servers.

With the OVH Private Cloud, they get a fully-dedicated, scalable, high-availability infrastructure, with ISO 27001 certification. It also offers VMware virtualisation and NSX software-defined networking (SDN) solutions. This way, Publicis ETO can up or downscale its infrastructure as required, and add or delete resources on-demand.

“We work for many customers, and we needed a solution that was adapted to suit our customers’ highest needs. This is why we choose OVH. There were a number of reasons we chose them, but what stood out for us was the fact they can guarantee hosting and maintenance in France. On top of this, the platforms chosen by Publicis ETO (IaaS, rather than PaaS or SaaS) would not have been profitable with a public cloud model.”

Emmanuel Guiffroy, CRM Solutions Architect

By opting for Big-HG Dedicated Servers, the agency could customise its servers in order to handle intensive workloads, and boost high-performance processing with next-generation processors.

Thanks to its partnership with Claranet, Publicis ETO also benefits from OVHcloud Connect. They can use this solution to create an isolated, secure network connection between their own network and the OVH private network.

Configuring and managing the platform

Publicis ETO has entrusted Claranet, its established partner, with this responsibility, as Claranet’s expert teams specialise in data projects. This way, the agency could focus on its core business, and delegate the technical aspects of it without any hassle.

Publicis ETO hardware layer

The result

Since October 2017, the platform has been functional and stable. It has helped significantly improve two aspects of the business:

1/ Customer relationships, by authorising the use of new algorithms.

2/ The productivity of data scientists, since it has reduced processing time by 400%.

With this big data platform setup, Publicis ETO can now industrialise the model, in order to deploy it for other projects. The agency will also be able to offer machine learning on a much larger scale, and for an increasingly wide range of applications.

Through a supervised machine learning approach, they will be able to generate algorithms capable of forecasting an event (e.g. a product purchase) and a quantity (e.g. a sales prediction).

Unsupervised machine learning will organise customers with similar behaviour into segments.

And finally, unstructured data — namely text mining — will provide automated classification models for documents and insight detection, targeted towards the luxury goods industry.