Enkel | Case - Smiles Sagemaker

Smiles - Sagemaker

Smiles embarked on a project as a strategic part of product recommendation within a larger initiative to redesign the Smiles digital platform, involving changes to login flows, CMS, backend, and API and service architecture. A recommendation engine was developed using various Cloud products with the goal of displaying storefronts, cards, and product recommendation events to pre-selected or non-pre-selected customers. DataRobot was chosen as the platform for developing, training, and scoring the data models used in decision-making regarding the recommendations to be displayed to the customer, when applicable. AWS Sagemaker was used to host the model and provide a persistent endpoint for prediction retrieval.

Client

Smiles is one of the largest loyalty programs in Brazil, with 16 million participants as of March 2019. Created 25 years ago as Varig's relationship program, it was later acquired by GOL and became an independent company in 2013, raising R$1.1 billion through an initial public offering (IPO).

The loyalty program is known for innovation and pioneering launches since 2013, such as Clube Smiles, Viaje Fácil, and Conta Família. The company also has the longest mileage validity in the market, ranging from 3 to 10 years, depending on the customer's category.

Challenge

Smiles requested an extremely scalable, flexible, and cost-effective architecture. With these premises, Enkel decided to use the following AWS solutions:

Amazon SageMaker;
Amazon EMR;
Amazon API Gateway;
Amazon CloudFront;
Amazon S3;
AWS Lambda;
Amazon SQS;
Amazon ECS;
Amazon Kinesis;
Amazon DynamoDB;
DataRobot Model;
Auth0.

Architecture

Recommendation Engine

Parameter Manager

Application

The platform displays the following recommendations:

Co-branded Cards

For pre-approved customer bases provided by the CRM department, containing information about the pre-approved bank and its variant (International, Gold, Platinum, Infinite).

Ancillary Revenue
Products related to travel and based on the purchased ticket (Baggage; Seat; Travel Insurance; Hotel; Car Rental; Uber Credit; Theme Parks and Attractions; Excursions and Sightseeing Tours; Cruise Travel).

Clube Smiles

Offer for subscription to the most suitable Clube Smiles package for the customer's profile, for eligible customers according to business rules.

Parameter Manager

To make the recommendation engine customizable based on the company's business strategies, a SPA application was developed for parameterization of various operational conditions, including: enablement, lifespan, and weight of the recommendation; product variants; multiplication factor; and others.

Authentication in the application is done via SSO with the corporate Active Directory, and parameters are stored in a DynamoDB table.

Automation

Infrastructure provisioning was developed with AWS CloudFormation and packaged with AWS SAM through the Smiles DevOps pipeline.

Results

With the project's GO LIVE, Smiles was able to boost its sales and enhance the user experience by suggesting products and services that align with the customer's profile and needs.