With more than 114 million items available on Magalu, more efficient searching paired with personalized recommendations are vital to driving a better customer experience, brand loyalty and sales for Brazil’s largest e-tail super app and website
Retailers ultimate goal is to deliver a great experience both in-store and online like they might enjoy from a personal shopper who understands their needs and preferences. In the immortal words of Apple founder Steve Jobs: “a lot of times, people don’t know what they want until you show it to them.”
That ability to help online shoppers “discover” products they’re more likely to click on and purchase by serving up relevant items on the right pages at the right time is what internet marketers call customer experience personalization. It is what has driven incredible growth and revenues for some of the world’s most valuable tech brands from Amazon to Netflix.
On the back end, what’s driving hyper-personalized discovery through search is a deeper understanding of shoppers’ backgrounds and preferences – all driven by accessing the right data and sifting through it at lightning speeds by applying data science, analytics, and various forms of A.I. such as machine learning, deep learning, and large language models.
Engineering opportunities for more on-target, “serendipitous” discovery that leads to more click-throughs and conversions is a high-stakes race for the world’s biggest digital companies – including Brazilian retail company Magazine Luiza (Magalu), the country’s #1 e-commerce player. Brazil is amongst the world’s fastest growing e-commerce regions. Magalu’s search R&D efforts to accomplish this are led by Caio Carneloz, Search Squad Leader at Luiza Labs.
Magalu partnered with CESAR to improve its search engine
Carneloz has led a squad of 12 researchers, designers, and data scientists for the last several years who are squarely focused on developing the best personalization innovations for Magalu – including experts from CESAR who specialize in A.I. and data science who add more academic knowledge and methodology to the team, he said.
Caio Carneloz, Search Squad Leader at Luiza Labs.
“CESAR’s expertise in advanced A.I. has enabled us to deliver more relevant, personalized results for the long tail of search queries. To give an example, when searching for a Galaxy S23device, people might type in something like, ‘a smartphone that can play games’ – but with only basic search, we cannot solve this problem,” said Carneloz. “To address this, we mined our CESAR colleagues’ expertise from the cutting edge of the academic field such as solving it using transformer neural networks , for example.”
Using transformers – a sophisticated technique in the field of natural language processing – helps “machines” (computers) tackle various languages and human communication with a sequence-to-sequence model that lets the A.I. understand relationships that can span multiple sentences or in various directions (a.k.a. “parallelism”).
From a practical viewpoint, the use of transformers let shoppers who mistype or misspell a product name that’s common in the U.S. such as “air fryer” into Magalu’s search function, but still have the top recommended air fryers at the best prices and shipping timeframe populate that shoppers’ screen, according to Carneloz.
“Even though they mess up the word air fryer, they get the results they’re looking for,” he said. “And because of this, we get more of the desired click-throughs and conversions. This same algorithm lets our shoppers find complicated tech products in long search strings with too many words, like: Galaxy Z Fold 64 gigabytes and 13 megapixels.”
Scrum methodology is key to faster insights and complex problem solving
Much like the philosophy guiding winning sports teams, the scrum methodology and agile project management framework deployed by Magalu’s Luiza Labs in partnership with CESAR lets the RD&I team self-manage, learn from experience, and rapidly adapt to change. It’s a proven way to solve complex problems by prioritizing efficient planning, collaborative execution, and a core focus on a team’s continuous improvement.
This methodology and framework has led to key insights about what Brazilians prefer and will react to positively in making their shopping decisions. For example, within a massive country such as Brazil which is nearly a continent unto itself, people prioritize delivery time of products based on the location of distribution centers. Magalu will display different results for shoppers from São Paulo versus the northeast region because the “best” products will vary based on delivery times and shipping costs.
“The top iPhones I search for will be delivered to me in São Paulo in just one or two days, whereas people in the Northeast might not be able to get those same iPhones until 10-15 days have passed,” said Carneloz. “To use the same product in another example, we found people will click on a low-priced iPhone first, but most won’t buy it once they find out it’s been used. And the majority of people will select the lower Pix price or installment payments in Brazil. All of these considerations are used to refine our search results.”
Deeper personalization leads to boosted brand loyalty and sales
Through the use of data science and advanced A.I., Magalu’s customers can more easily find what they need from the more than 114 million items across specialized categories such as books, digital entertainment, groceries, fashion, food delivery, sports and technology brands. This lets customers find what they’re looking for even if they type in the wrong words (if the phonetics are the same), search by voice, or even find items that fit them best by reviewing images of the shopper to make recommendations for them.
During the recent Google Cloud Next event in August in San Francisco, Magalu announced it is introducing “Lu’s Brain” – so its shoppers can interact with “Lu” – the most followed avatar on Earth who has long been the virtual face and voice of the e-commerce giant since 2003. In order to improve Lu’s prowess at answering shoppers’ questions and guiding them to the perfect products based on their personal preferences, diverse backgrounds and locations, Lu is being further developed and trained with generative AI in an e-commerce context for the modern day.
“Moving forward into 2024, Generative AI is one of the top priorities for our RD&I squad, along with leveraging new large language models to better understand what shoppers are looking for,” said Carneloz. “In the search field, these types of long-tail queries can be some of the most difficult to interpret and react to without LLMs and hyper-personalization.”
CESAR Innovation Center in Recife, Brazil is one of Latin America’s most renowned institutions for world-class research, design, and innovation on demand. For more than 25 years, CESAR has been a trusted partner for some of the best-known brands seeking to co-develop breakthrough innovations that up-level their competitiveness and productivity.