A household name for kiwi shoppers, The Warehouse Group is the largest retail group in New Zealand. The Warehouse has over 90 retail locations across New Zealand, and as part of an ongoing dedication to customer excellence, they are leading the charge to provide customers with an online shopping experience on par with their in-store experience
This dedication has encouraged The Warehouse Group’s digital team to explore the elements of customer experience that are more difficult to replicate online - things like browsing the aisles until the perfect Christmas gift that a customer has been searching for presents itself.
The team at The Warehouse Group quickly identified an idea that could help create a version of this experience on their website. A recommendation engine - a tool that could be used to present gift ideas for different personas (for example, “gifts under $100 for a husband who loves the outdoors”) could be used to sort through the thousands of products listed on the Warehouse website so that shoppers could quickly hone in on a suitable gift. The team hoped that making this process as effortless as possible would yield greater conversion rates, and they were right!
The recommendation engine itself turned out to be the more straightforward portion of the equation. With thousands of products listed in The Warehouse Group’s various systems - many in different formats or with different feature listings - the idea of manually categorising and standardising listings was a herculean task, especially with the looming Christmas deadline approaching. The main hurdle that the ElementX team faced in implementing the gift finder was this categorisation.
The ElementX team identified that a Natural Language Processing solution would be the most efficient way to read and categorise product descriptions, and after some testing and iteration, the team determined that a GPT-3 model would be best suited to the wide variations in language and structure present in the product descriptions. The team used this to classify products with a high degree of accuracy, focusing on specific product types that would be useful to include in the gift finder. GPT-3 is pre-trained, which means our team didn’t have to spend a lot of time in the setup phase with training data for the AI to understand how to categorise products. This, in turn, led to a faster result for The Warehouse Group’s team and meant that the Gift Finder was ready in time for Christmas!
The ElementX Team worked closely with The Warehouse Group’s digital team to identify the success criteria for this project, and after running a successful proof of concept of the gift finder for Mother’s day, the team set their sights on Christmas gifts. After using GPT-3 to categorise the products to be included, the team built the customer-facing gift-finder page on a robust, scalable foundation, using Google Cloud so that the result could accommodate large spikes in user traffic as was expected of the Christmas holidays. The Spark 64 team also built an analytics dashboard to analyse and report on customer’s activity with the finder, allowing The Warehouse Group’s digital team to quantify success and track customer behaviours within the gift finder.
Over the three months leading up to Christmas, The Warehouse Gift Finder saw over ten thousand new users to the Warehouse website, and visitors to Gift Finder pages stayed on the Warehouse website nearly three times longer than visitors to the rest of the site. The Gift Finder was the highest online source of revenue generation for The Warehouse Group during the Black Friday and Cyber Monday periods, with an average value of Gift Finder orders 28% higher than the sitewide average. It contributed to the joy of hundreds of Christmas mornings nationwide - and we couldn’t be more proud to have been a part of this project!
Ming Cheuk, CTO: Ming designed the architecture of the solution, working closely with The Warehouse Group’s digital team to ensure that every eventuality was considered and that our build would be successful.
Dmitrii Goriunov, Senior Full Stack Engineer: Dmitrii built the backend of the gift finder, ensuring its stability and ability to handle large volumes of visitors. He was also responsible for connecting and training GPT-3 to categorise The Warehouse Group’s products for the gift finder so that customers could search the full range of products.
Aorangi Smith-Iri, Full Stack Engineer: Aorangi built the frontend interface of the Gift Finder, ensuring a smooth, effortless user experience for the Warehouse online shoppers.
Daniel Jimenez, Technical Product Manager: Daniel provided project management to keep the Gift Finder build process on-budget and on-schedule. He also provided quality control for the duration of the project.
Find the Gift Finder here
