Harnessing consumer data for personalization
Very few purchases in life play as significant a role as a jewelry gift for those we love most. Whether it’s an engagement ring or anniversary present, every piece celebrates a one-of-a-kind personal connection. Pandora’s purpose is to help people give a voice to their love. The company fulfills this purpose by creating affordable, hand-finished jewelry for the many rather than the few, with pieces that can be personalized to reflect our many facets and express who we are, creating unforgettable moments.
This has driven Pandora to offer a more personalized experience throughout the customer journey. Core to this strategy is the ability to tap into data captured through customer interactions. Pandora ingests three different types of data: customer profiles with product-level information, online and offline orders, and web page activity. Their legacy data solution struggled to efficiently manage both the sheer volume of data and the complexity of it. Cross-team collaboration was not supported, as each team used different programming languages and had different levels of access to data, preventing them from approaching problems through the same lens. And their legacy system, which required resource-intensive manual processes to build and train complex machine learning (ML) models, also slowed data science teams. Numan Ali, Solution Architect, Data and Analytics Center of Excellence at Pandora, explains, “It was too complex for our data teams to explore, collaborate and analyze data at the scale we had. And our struggles to deliver the data to our teams blocked our ability to innovate with analytics.”
The company also faced challenges in building reliable and performant data pipelines on their previous infrastructure to support downstream analytics and ML use cases. Not only did this compromise Pandora’s time to market and, therefore, their ability to compete — it also made the goal of personalized customer journeys impossible. “It took more than a day for the data to go from the source to the very endpoint,” Numan says. “In order to deliver an accurate personalized shopping experience, we needed a simpler and streamlined approach that would allow us to take advantage of our valuable data.”