WPP cuts AI campaign deployment times with Google Cloud

WPP has built a unified data and platform engineering system with Google Cloud for its AI marketing operations, cutting campaign deployment times from months to days.

The work underpins WPP Open, the company's marketing system, which uses AI models to analyse market shifts and support campaign planning. Before it could apply those models at scale, WPP had to address fragmented marketing data spread across hundreds of agencies.

To do that, it created a centralised data backbone and standardised how applications and data workloads are deployed. In its production environment, WPP moved to a service-based project structure, placing Google Cloud Storage and BigQuery in shared data projects and separating processing workloads into dedicated projects.

This setup gives internal users a single source of data while maintaining tight controls across teams and business lines. Access restrictions are applied at the level of individual storage buckets and BigQuery datasets, so teams see only authorised information.

Data structure

Raw information from external partners is stored in separate buckets before processing. Managed Service for Apache Spark runs Scala and Spark jobs to clean, normalise and standardise the data into what WPP calls standardised cohort definitions.

Those cohort definitions are based on five keys: age, gender, geo, product and interest. According to WPP, this model lets teams join and aggregate data globally without exposing sensitive underlying details or relying on shared identifiers.

The core processing engine was built in Scala with a focus on traceability. WPP said every data point in curated datasets can be traced back to its source, an important feature for model oversight and audit requirements.

Jonas Dahlbaek, Senior Data Engineering Lead at WPP, described the approach taken with Google Cloud.

"What made our collaboration with Google Cloud successful was the balance they struck between uncompromising professionalism when it comes to best practices and timely delivery of incredibly pragmatic, real-world solutions," said Dahlbaek.

Software lifecycle

Beyond data processing, WPP's platform engineering team developed a set of reusable GitLab continuous integration and continuous deployment templates. The aim was to reduce the burden on individual development teams and apply common security standards across deployments.

These templates cover Cloud Run deployments for web applications, batch processing and scheduled pipelines, along with deployment-only workflows and event-driven services. WPP said the shared templates have helped it manage a wide range of workloads with less variation between teams.

The group also adopted a build-once, deploy-many method to avoid rebuilding container images in production. Developers build and test images in development, then promote the same immutable images into production once validated.

WPP said this zero-rebuild promotion model reduces the risk of configuration drift between environments and improves release consistency. The deployment process also allows traffic to shift gradually to new revisions before a full rollout.

A senior platform executive linked that consistency to trust in AI systems and release management.

"Immutable deployments. Traceable data. Unshakable trust. When you know exactly what goes into your AI, you can ship at the speed of light," said Poldas.

Security controls

Security scanning is built into the development pipeline, with Wiz checks added before code is pushed for merge. WPP also uses Google Cloud Identity-Aware Proxy to control access to internal applications under a zero-trust model.

It also introduced templates with automated virtual private cloud logic to identify and resolve networking conflicts between older connectors and newer direct access methods. That is intended to reduce failed deployments and simplify network configuration.

Suleman Khan, Product Manager for OI & Google Partnerships at WPP, said the infrastructure work has helped both engineers and marketing teams.

"Working with Google Cloud has been instrumental in accelerating and standardizing our engineering efforts. In a world where massive volumes of fragmented data present a daily challenge, having the right infrastructure is paramount to thriving in the AI age and helps our developers and AI marketers alike," said Khan.

Measured impact

WPP said its teams now monitor operational health through latency across p50, p95 and p99 measures, along with 4xx and 5xx error rates, startup times, and CPU and memory use. The goal is to track the resilience of both data pipelines and serverless infrastructure rather than judge performance mainly by deployment frequency.

The commercial effects cited by WPP were substantial. It said the combined data and platform overhaul reduced creative and strategy time from four weeks to three hours, increased production efficiency by 70%, lifted content volume 33-fold and raised campaign return on investment 2.8-fold.

Yang Yue, Program Manager at Google Cloud, outlined how the work was managed across both organisations.

"Navigating a transformation of this scale across multiple complex workstreams-spanning data engineering, platform infrastructure, and AI integration-required more than just alignment; it demanded deep, mutual trust. Working as true partners, Google Cloud and WPP moved in lockstep to deliver production-ready platform capabilities on time," said Yue.