Crafting marketing campaigns that resonate with customers

Overcoming legacy data and infrastructure challenges

HP’s vision is to engineer experiences that amaze, serving individuals and businesses through innovation in computing, printing and digital solutions. Operating globally, the marketing data sciences team supports the continued measurement and optimization of HP’s marketing investments — ensuring the right data and infrastructure are in place to meet critical measurement outcomes, drive investment effectiveness and predict future actions to enhance spending efficiency. HP has managed a robust customer relationship management (CRM) of owned data across numerous systems and sources for years, but with over 50% of this data going unactivated combined with approaching third-party cookie deprecation, they knew their visibility into customer behaviors and acquisition effectiveness was at risk. Separately, the sheer volume of data — coupled with a limited view of the customer — created heavy reliance on the marketing data sciences team to pull insights from HP’s internal assets via costly, time-consuming processes. This resulted in data latency for campaign activation and hindered their ability to target customers in a meaningful, timely way.

Managing data across multiple platforms and systems also created friction in maintaining data quality and consistency. Different business units operated in silos via their legacy systems — including cloud environments and some homegrown systems. Since these tools weren’t designed to handle the scale and complexity of contemporary data environments, they lacked the modern capabilities required for effective data movement and integration. Luis Alonso, Head of Customer Data Strategy and Engineering at HP, elaborated: “It was our job to ensure that a solid data infrastructure was in place to leverage the CDP. We needed to strategize and build a next-gen data ecosystem for our measurement framework, and audience activation was key, and our data needed to effectively support our marketing teams.” Plus, keeping up with evolving privacy regulations and consent concerns became a challenge. The company needed to implement a federated approach to data activation, enabling a unified view of their customers without requiring data movement and copying.

According to Kumar Ram, Global Head of Marketing Data Sciences at HP, the company isn’t alone in this uphill battle. “I think this is the universal theme for most organizations in our space — the volume of data, the siloed data and the lack of a consolidated view of the customer. Not just the physical centralization, but having that logical layer on top of our customer 360, allowing us to leverage the federated approach and piecing together our different data environments across the organization.”

“With our size and complexity, it’s one thing to consolidate a few hundred data sources into 10–15 different sources, but trying to consolidate everything into one repository is a challenge. There was a lot of potential revenue that we were leaving on the table,” Ram explained.

In 2022, HP’s IT organization invested in unifying all customer data in the Databricks Platform — unlocking the ability to access insights more seamlessly across the enterprise and more strategically, laying a data foundation that could support a composable architecture. It leverages Delta Lake, an open source compute layer that brings atomicity, consistency, isolation, durability (ACID) transactions to big data workloads, ensuring that the data in Delta Lake is reliable and consistent.

By combining the power of Delta Lake, Unity Catalog, the medallion architecture and other relevant components, HP is building a framework that supports efficient data pipelines, better data governance and improved collaboration across their data environments. The architecture ensures that HP’s data is organized, accessible and secure across the organization, leveraging bronze (engineering), silver (data scientists) and gold (data analysts) for various data consumers. Leveraging several ingestion components, raw data is ingested, cleansed, enriched and refined incrementally. By organizing data into these layers, HP can streamline data workflows, reduce processing times and facilitate seamless integration with analytics and reporting tools. With this widened access to data, Unity Catalog is a key component in the mix, enhancing access governance for HP’s data assets and defining and enforcing granular permissions.

Embracing first-party data to get ahead of cookie deprecation

As third-party cookies began to phase out, HP needed to pivot toward a first-party data strategy and democratize the untapped insights in their CRM system across the enterprise. The brand aimed to turn traditional strategies on their ear, especially with the resurgence of contextual targeting, to unlock powerful multi-channel use cases with first-party insights. HP was committed to future-proofing their stack against cookie deprecation quickly, doubling down on Databricks and investing in the right customer data platform to enable more-precise audience segmentation at speed and scale.

Before cookie deprecation, the technology company relied on cookie-based data via a data management platform (DMP) for all media activation. Moving forward, the brand aimed to migrate off their DMP and invest in centralizing and activating their first-party data with a solution that had the flexibility and agility to evolve. Ram explained, “Our objective was not only to execute many of the DMP use cases without the use of cookies — creating segments, offering media platform and database connectivity and being able to assess performance, for example — but also to establish a platform that we could evolve into an enterprise customer data platform.”

HP needed a platform that facilitated self-service segmentation, which made Databricks’ partner, ActionIQ, a natural fit. ActionIQ enabled HP to leverage their existing CRM data effectively, to create precise audience segments and seamless activation across various media channels. Additionally, ActionIQ’s ability to integrate consent data into the whole process helped HP replicate DMP use cases while remaining compliant, optimizing marketing spend and enhancing campaign effectiveness.

Connecting Databricks and ActionIQ to reduce data copy and move faster

Despite making significant strides in activating first-party data and overcoming data processing inefficiencies, there were still opportunities to further streamline campaign processes and improve governance by adopting a composable architecture. The team explored the idea of federated query pushdown via ActionIQ’s HybridCompute capabilities — connecting ActionIQ and the Databricks Data Intelligence Platform via a native integration.

“We tried to centralize all our data in one place, massively onboarding it to a data warehouse and layering reporting tools on top, but as we experienced challenges with adding data sources we ultimately realized this was not scalable. This is where open source standards and data federation — having our data sitting in one or more systems and the ability to bring it together without having to copy it — became really key for us. This also unlocks a major opportunity to improve the quality and accuracy of the data,” Alonso said.

Ram added, “We were using extremely expensive resources to do data integration and collection, which was not a good choice. The answer there was not having to move the data or make multiple copies of it. This is the key value prop that HybridCompute brought. Thankfully for us at HP, we had our IT organization already creating the building blocks by moving all our customer data into Databricks. All we had to do was layer in the ActionIQ CDP on top of it using its HybridCompute architecture.”

HP conducted a proof of concept to test the Databricks-ActionIQ connector. Unlike typical reverse ETL solutions, HybridCompute offers flexibility for organizations that don’t have their data centralized in one system to federate across multiple sources and to leverage the investments of their homegrown data models more effectively.

“Our old process required a long road of querying, validation and transfers with at least one full-time data scientist dedicated to this manual work. Once HybridCompute pushdown query configuration was in place, it was just a matter of getting on the ActionIQ platform and building the audience.”

Alonso reflected, “Being on both Databricks and ActionIQ and having HybridCompute capabilities will allow us to leverage our data models and governance components more effectively. Now, it’s about operational efficiency and adding value with additional features. We are considering other components, like MLflow to build models and journey orchestration that can impact our overall data strategy.”

Driving impact with federated query pushdown

With Databricks and ActionIQ, HP has generated measurable impact across stack optimization, time to market, campaign performance and cost efficiency. ActionIQ’s intuitive business-friendly user interface enabled HP to democratize their data across the enterprise, empowering marketers to create audience segments themselves, reducing reliance on data science teams and improving campaign velocity. The brand observed a drastic reduction in the time required for data processing and audience building, which now only takes 1 to 2 hours to complete, as opposed to 5+ hours. This has improved the overall accuracy and reliability of HP’s 360-degree customer views.

With ActionIQ HybridCompute alongside Databricks, HP has unlocked faster querying, allowing for the processing of 400 million records in seconds and eliminating the need for extensive change management or data pipelining. This setup allows HP to retrieve and analyze specific data to create targeted audience segments. According to Alonso, “Now, we can stand up a new campaign within a couple of hours compared to 2 weeks previously.”

In using best-of-breed solutions, HP has more clarity on their future roadmap. The multinational information technology company plans to build a comprehensive data quality framework that ensures accountability for upstream data sources through a scorecard system. The brand also plans to leverage Databricks’ MLflow for measuring and reporting model performance, transitioning from traditional dashboards to AI-driven capabilities that can generate graphs and charts on demand. Finally, with the eventual broadening of Delta Sharing use cases, HP will allow secure access to data across different clouds and improve performance measurement. Ram concluded, “When I think about the work that we do in marketing data sciences — it’s focused on measurements. ‘How have we performed lately?’ ‘How are our investments doing?’ ‘What are we going to prioritize next?’ With Databricks and ActionIQ, we have a much more predictive point of view and the maturity to the answers.”