Natura has just launched its Intelligent Search, a search service based on AWS AI that assists over 1 million consultants, who currently perform about 400 searches per minute daily.

Overview

Natura is one of the leaders in the beauty and cosmetics sector in Latin America, with a significant and impactful presence in the region. With renowned brands like Natura and Avon, the group covers a wide range of products and services, ranging from skin and hair care to fragrances and makeup. Today, Natura has 14.5 thousand employees distributed across 18 countries, in addition to five factories and 19 distribution centers.

This structure serves over 900 proprietary and franchise stores and approximately 3.5 million beauty consultants and brand representatives, with more than 1.6 million of them in Brazil. In addition to the physical structure, the consultants also rely on a product search system, which the company has just modernized using generative artificial intelligence (AI) based on AWS.1186690677

Natura's senior cloud engineer, Eli Melo, explains that the previous search system only allowed for exact term searches, which could often discourage a purchase. "A person would type a term, and if it didn't have an exact match among the products, there was no return. The user was left without an answer and wouldn't buy," he explains, citing the example of the Kaiak perfume, which the system wouldn't find if the consultant searched for "caiaque".

Furthermore, the search engine had a character limit per product name, which also hindered the search, since many registrations were made with abbreviations. To meet the consultants' expectations, Natura decided it was time to modernize the system using more sophisticated and efficient technologies that would make the search smarter.

"We started thinking about using algorithms and AI to perform the search. During the study period, we contacted AWS, which provided us with an immersion in the use of AI-based services," he says. According to Melo, one of the products that materialized as a solution was precisely the Amazon OpenSearch Service, which already has built-in algorithms that allow for context and phonetic searches, in addition to text.

Eli Melo - Senior cloud engineer at Natura

"During the study period, we contacted AWS, which provided us with an immersion in the use of AI-based services"

Solution | Building an intelligent solution

The executive explains that for the search to perform as it should, there was a need to apply AI algorithms to the product database. Thus, the solution was found in existing algorithms, such as phonetic search, vector search, and tag generation through Large Language Models (LLMs), enabling a search that could be considered intelligent.

Hence the decision to develop an internal search solution based on AWS services such as Amazon OpenSearch, which offers flexibility in storing product registration data and enables the application of a variety of Machine Learning (ML) algorithms, allowing for phonetic, vector, and other searches; and Amazon Bedrock, which allows for pay-as-you-go pricing, eliminating concerns about instantiating and maintaining infrastructure. "With Bedrock, in addition to accessing a variety of available LLMs, we can also take advantage of native integrations with other AWS services, which makes the solution simpler to operate," explains Melo, noting that the LLM models also allowed for the addition of terms (in the form of tags), relating registered products with contextually relevant terms.

The solution also utilized AWS Lambda, which, with its serverless approach, allows code execution without the need to manage servers, featuring automatic scalability and payment only for execution time. Melo emphasizes that this combination forms a more flexible infrastructure, designed as follows:

This architecture provides a flow that begins with the client's input to the search API, performing a semantic query on the Amazon OpenSearch cluster, while also feeding a pipeline with data on user behavior (most searched products, searches with no returns, etc.). Meanwhile, upon the insertion of new products, generative AI makes up the pipeline to enrich the product metadata, assisting at the time of the search:

"We worked on these ideas through a path of studies, trials, and errors. A cool point is that on AWS we have the ease of spinning things up, testing, validating, etc. A simple test doesn't take months. This speed helped us show value quickly," he states. The executive explains that, in this way, a first non-production environment was developed, showing that the issues related to the previous system were resolved. Thus, the search system was improved until it was deployed in a production environment, initially with 5% of the sales team, focusing on testing and collecting metrics. From ideation to this first version, all development was completed in one month.

Natura's senior cloud platform engineer & DevOps for Latin America, Marcelo Fabricanti, recalls that this first version was delivered fully functional, allowing for enhancements such as the inclusion of new metrics. "Today we map what people search for and we are moving to a third phase: whether the search result went to the cart and on what page the search result was. We are working on relevance. It is a long-term process," he says.

Next steps | Improving results and sales

Fabricanti highlights that since the implementation of Intelligent Search, an increase has been noted not only in the results obtained but also in their accuracy. Today, the solution is available to 100% of the over one million Natura consultants in Brazil, who are responsible for performing about 400 searches per minute daily.

Compared to the previous system, Natura recorded gains such as 205 thousand searches with products added by the new system; 54 thousand searches returning results that would not be met by the previous system; and 99% of searches with related products added to the results. "We can conclude that the new search engine not only includes products but also generates returns that, in many cases, would yield no products with the old system. This results in a significant increase in the probability of the user adding products to the cart, which consequently increases the possibility of conversion," he states.

Fabricanti also points out that, when analyzing the performance of Intelligent Search, his team found that, on average, queries are completed in less than 50 milliseconds, which indicates not only functional improvements but also a performance gain compared to the previous system. "The idea for the future is to use the new engine across all our channels: apps, end consumers, etc. Today we are in one of the applications, but there are other channels that can still receive this revamped search. The idea is to expand to all channels," he concludes.

About Natura

Natura is one of the leaders in the beauty and cosmetics sector in Latin America. With brands like Natura and Avon, the group has about 16 thousand employees distributed across 18 countries, in addition to five factories and 19 distribution centers.

AWS services used

Contact