Customer challenge: This e-retailer wanted to modernize its e-commerce platform, implement recommenders, and improve search functionality.
Provided solution: Combining data from legacy systems into a consolidated data platform. Using this data to offer near real-time personalization to customers.
Outcome: Increased click-through and conversion from recommendations and shoppers that used search function.
This e-retailer is a top-3 player in the Netherlands which serves 1.7 mil customers per year with a combined total of 145 mil site visits. On top of that, the e-retailer experiences customer interactions such as the opening of and clicking in e-mailings, calls to the contact center, and the reviews that are written on comparison sites. This large e-retailer worked together with GoDataDriven to implement smart personalization algorithms in its new e-commerce platform. These algorithms allow the organization to respond in near real-time to individual customer interactions during the customer journey.
Consolidating data from legacy systems
The technology to store and analyze customer data in near real-time on one integrated platform has become available at low-cost in almost the blink of an eye. Although every organization with large datasets looks into the opportunities of extracting value from data, not every organization is capable of actually creating solutions on top of their data.
It proves to be rather difficult to consolidate data from various legacy systems and develop solutions that add business value. This large e-retailer took the step, knowing that the technological challenges are only minor compared to everything else. “You need technology, people, and methods”, says their manager of marketing technology.
”Technology might be the easy part. In practice it proves far more difficult to change ingrained habits and methods. And then we are not even speaking about attracting data scientists capable of extracting value out of data.”
The e-retailer ventured out for partners that were capable of helping the organization with this transition, who were able to implement the right technology and new methods, as well as a new culture, and contacted GoDataDriven.
Completely new e-commerce platform
As mentioned, the turnaround to a DataDriven organization is all about technology, people, and methods. Starting with the technology: the e-commerce platform had been built from the ground up. The organization used to work with many in-house developed applications. For the new platform it was decided they would look for applications that are able to execute one specific aspect of the broad marketing spectrum, with a preference for open-source solutions. Applications that were not available open-source were developed by a software development team. “A patchwork of applications has been developed. All applications are kept together by an in-house developed platform that makes sure that all data from underlying systems becomes available in Hadoop.”
Missing in this configuration still was a solution to capture clickstreams from the website. Ultimately, the organization opted to further develop Divolte, an open source clickstream collector. The smart algorithms themselves are not in the open source tool Divolte, but are developed based on the data it collects.
Looking for a new type of developer
Developing smart algorithms has become a specialty of the e-retailer. Which brings us to the subject of people: a new type of intelligence specialist is required to do this. Instead of business intelligence people, this means data scientists. The e-retailer trains its own people for this, and is continuously looking for new talent. Both data scientists as well as software developers are granted a lot of freedom to work on cool new features and smart self-learning algorithms in self-managed full-stack development teams. This has led to a new culture, with the corresponding new methods. The organization now applies agile, Scrum, and DevOps to its software development process. Teams act as product owners and are responsible for their contribution to the progress. This culture shift started at the IT-department and has spread to other departments as well.
Smart algorithms for personalization
As said the proof is not so much in the software but in the smart algorithms that perform personalization. Not that this organization is new to personalization; they used to be frontrunners in this area. At that time it was not possible technically to analyze data from customer behavior in real-time, so the e-retailer had to personalize the website based on more or less static profiles. That proved to be not effective, as the products that someone bought in the past only partly reveal something about one’s current interests. As of recently, the organization uses brand preferences, a technique which works far better. The sky is the limit when you develop the capability to use the data that you collect in real-time. The better your ability to optimize the algorithms, the more relevant an individual customer journey will be.
The foundation of the new strategy is formed by the development and optimization of algorithms. The impact that developers have on the core of the organization has left them inspired.
Understanding individual customer needs
As soon as the foundation of the new e-commerce platform was laid out, a data team consisting of internal specialists and data scientists from GoDataDriven developed personalization algorithms.
These algorithms allow the e-retailer to drastically improve its personalization efforts, not only based on order history, but also based on real-time site interactions. Analyzing the behavior has lead to a better understanding and insight into individual customer needs. Different customer groups will be served by different search algorithms, providing a faster and better search experience. This might even lead to offering products a customer had not thought of before. Another aspect is that the organization will be able respond to current events, like the weather, by executing marketing campaigns more quickly, improving the customer experience even further. The objective is mainly to improve the customer relation, so that customer retention improves. That by itself will lead to more revenues, which is a great bonus, but not the primary target.
New phase
The new e-commerce platform and the data driven marketing approach herald a new phase for the e-retailer. The current state will be continuously evaluated. What if technology is no longer the limiting factor? How quickly is your organization able to adapt its strategy to individual customer needs?
Industry
RetailProject type
Customer predictive modellingDivolte clickstream collector
Technologies used
HadoopR
Python
Divolte
The objective of developing machine learning algorithms is mainly to improve the customer relation, so that customer retention improves. That by itself will lead to more revenues, which is a great bonus, but not the primary target.