Slalom’s Experience with Amazon Personalize
Consumers are becoming increasingly used to dynamic, deep experiences for products they are using. Succeeding at personalization and recommendation is a way for our clients to form stronger relationships with customers and develop deeper consumer insights. Despite these benefits, adopting machine learning can be a difficult and time-consuming process with mixed results. In this post we will consider a real-world use case and evaluate how Amazon Personalize can accelerate development of personalization models and enable stronger customer relationships.
Amazon Personalize is a fully-managed, machine learning (ML) based service that enables organizations to efficiently add real-time personalization to their application. This AWS product is trained using an organization’s own data to provide the most relevant customer recommendations to drive material improvements in key business metrics.
In our experience, at Slalom, there are six key tenants that make it into production and produce measurable improvements for machine learning initiatives. These key tenants include a clearly defined business imperative, a strong data foundation, workflow and automation, expertise and tools, an agile mindset and a focus on the user and their adoption.
For organizations that are starting on a machine learning journey, services such as Amazon Personalize allow for teams to quickly create value by putting their unique business data, such as application activity stream data, inventory data, or customer demographic data, into action without extensive machine learning experience. This is done with two major machine learning techniques, deep learning and recurrent neural networks. This provides customized recommendations reflecting how end-customers engage in an organization’s specific environment. To get started, Amazon Personalize essentially turns sophisticated code development into a simple point and click configuration effort at little to no cost.
An electronics client of Slalom was beginning with machine learning and needed to deliver measurable results quickly. We decided to start with a ‘customer also bought’ recommender engine for the client’s website. This decision was made because there were already images on the checkout page. This meant measuring the increase in average sales would be easily understood.
Amazon Personalize helped to accelerate the development effort. This innovative product automated many of the individual tasks, tweaks and optimizations that usually take a lot of a time. We started by registering ecommerce interaction data sets with Amazon Personalize and then pushing ecommerce interaction data to the system.
Then we determined the “recipe” (instructions for what type of personalization model we wanted) that we would be using with Amazon Personalize. Setting up our data feeds and determining the appropriate recipe was key to getting an impactful model from Amazon Personalize. Establishing these data feeds allowed us to take advantage of real-time recommendation updates based on what products were currently in the customer’s shopping cart.
Amazon Personalize provided two important additional benefits, speed of model build and continuous model improvement. This allowed our client to act quickly on customer insights and gain an advantage over competitors. They were able to retrain and deploy the model nightly verse the typical weekly or monthly processing schedule.
The speed of automated model development and continuous model improvements allowed our client to spend more time thinking about creating differentiated customer experiences and new strategies to optimize checkouts. As an example, our client was able to introduce new packaged offerings at a lower price than if those products were purchased separately.
Amazon Personalize allowed our client to quickly see value from machine learning by bringing cutting-edge algorithms to a key business problem. At Slalom we know organizations are in different stages of maturity in adopting machine learning and Amazon Personalize is a great way to get started.
David Frigeri leads the Data and Analytics practice for Slalom Philadelphia.