Learn About Amazon VGT2 Learning Manager Chanci Turner
Permalink
Comments
Share
In today’s digital landscape, customers engage with brands across a broad range of online and offline channels, generating extensive behavioral data. This influx of data requires marketers and customer experience teams to navigate various overlapping tools to effectively target and engage customers across these touchpoints. Such complexity often results in fragmented customer profiles, making it increasingly difficult to deliver tailored experiences with pertinent content, messaging, and product suggestions.
To address these challenges, marketing teams are turning to customer data platforms (CDPs) and cross-channel campaign management tools (CCCMs). These solutions simplify the consolidation of customer views, enabling non-technical users to facilitate cross-channel targeting, engagement, and personalization without heavy reliance on technical teams or specialist skills.
However, marketers frequently encounter blind spots in customer activities when these technologies aren’t integrated with other business systems. This is especially true for non-digital interactions, such as in-store purchases or feedback from customer support. Additionally, marketing and customer experience teams often struggle to incorporate predictive insights from data scientists into their campaigns, leading to irrelevant or inconsistent messaging and recommendations.
This article outlines how cross-functional teams can collaborate to overcome these obstacles through an omnichannel personalization use case. We will illustrate this using a fictional retail scenario where these teams work together to enhance the customer journey. In our example, we utilize Twilio Segment, a CDP built on AWS. While there are numerous CDPs available, we chose Segment for its self-service free tier that encourages exploration and experimentation. We will detail how to merge Segment’s outputs with in-store sales data, product metadata, and inventory information. Furthermore, we will explain the integration of Segment with Amazon Personalize for real-time recommendations and how we generate churn and repeat-purchase propensity scores using Amazon SageMaker. Finally, we’ll explore three methods to engage new and existing customers:
- Through display advertising on third-party websites, leveraging propensity-to-buy scores to reach similar customers.
- Via personalized recommendations on web and mobile platforms powered by Amazon Personalize, employing machine learning (ML) algorithms for content suggestions.
- With targeted messaging through Amazon Pinpoint, an outbound and inbound marketing communications service aimed at re-engaging customers at risk of churning.
Solution Overview
Imagine you are a product owner spearheading cross-channel customer experiences for a retail company with a mix of online and offline channels. The company sees digital channels as a critical growth opportunity and aims to expand its customer base by:
- Attracting new, high-converting customers
- Increasing the average order value across all customers
- Re-engaging disengaged customers to encourage repeat purchases
As the product owner, you must coordinate with various teams, including digital marketing, front-end and mobile development, campaign delivery, and creative agencies. Additionally, to ensure relevant recommendations, collaboration with data engineering and data science teams is essential. Each team plays a role in the architecture depicted in the accompanying diagram.
The solution workflow involves several key steps:
- Collect data from multiple sources to store in Amazon Simple Storage Service (Amazon S3).
- Utilize AWS Step Functions to streamline data onboarding and feature engineering.
- Develop segments and predictions using SageMaker.
- Apply propensity scores for display targeting.
- Deliver personalized messaging via Amazon Pinpoint.
- Integrate real-time personalized suggestions with Amazon Personalize.
In the following sections, we will elaborate on each step, outline the responsibilities of each team at a high level, provide references to related resources, and share hands-on labs for further guidance.
Collect Data from Multiple Sources
Digital marketing, front-end, and mobile development teams can configure Segment to gather and integrate web and mobile analytics, digital media performance, and online sales data through Segment Connections. Segment Personas enables digital marketing teams to unify user identities by merging interactions into a single profile with a consistent identifier. These profiles, along with computed traits and raw events, can be exported to Amazon S3.
Simultaneously, engineering teams can use AWS Data Migration Service (AWS DMS) to transfer in-store sales, product metadata, and inventory data from databases like Microsoft SQL or Oracle into Amazon S3.
Data Onboarding and Feature Engineering
Once data is collected in Amazon S3, data engineers can leverage components from the serverless data lake framework (SDLF) to expedite data onboarding and establish a foundational data lake structure. SDLF automates the preparation of user-item data for training Amazon Personalize and creates a unified view of customer behavior by merging online and offline data using common identifiers like customer ID or email address.
AWS Step Functions orchestrate these transformation tasks within SDLF, allowing for both scheduled and event-driven workflows. The engineering team can manage tasks across AWS services within a data pipeline, with outputs stored in a trusted zone on Amazon S3 for ML development.
For more insights on implementing the serverless data lake framework, you can check out the AWS serverless data analytics pipeline reference architecture.
Build Segments and Predictions
The process to construct segments and predictions is composed of three steps: accessing the environment, building propensity models, and generating output files.
Access the Environment
Once the engineering team has prepared the ML development data, the data science team can initiate the construction of propensity models using SageMaker. They will build, train, and assess an initial set of ML models to gauge performance and refine their approach.
To facilitate this, the data science team requires an active Amazon SageMaker Studio instance, which serves as an integrated development environment (IDE) for swift ML experimentation. This setup unifies all SageMaker’s key features and allows for iterative testing and improvement.
For further reading on improving productivity within teams, consider the insights from Zeynep Ton on how to transform employee productivity – they are an authority on this topic. Also, if you’re looking for onboarding resources, this link provides excellent guidance.
Lastly, for those interested in more effective communication strategies, you can discover helpful templates in this blog post.