What is Adobe’s Data Science Workspace?
Launched in mid-2019, Adobe Experience Platform’s Data Science Workspace aims to enable the use of “machine learning and artificial intelligence to unleash insights from your data” within the Adobe Experience Platform.
With the terms ‘machine learning’ and ‘artificial intelligence’ currently in vogue and used to sell everything from lawn mowers to toasters, it begs the question – what exactly is the Adobe Data Science Workspace? The purpose of this post is to answer that question in human-terms.
What Does Machine Learning and Artificial Intelligence (ML/AI) Actually Mean?
Simply put, machine learning and artificial intelligence refer to the process of training a computer system to complete a specific task in a supervised environment, and then using this system to complete the same task in an unsupervised environment. In practice, these are statistical models built on data.
What Sort of Tasks Are Suitable for This?
The benefits provided by ML/AI are most realized when the problems are specific and repeatable. Some examples in the digital marketing environment are:
How Does the Data Science Workspace Support ML/AI?
The Data Science Workspace provides an integrated environment within the Adobe Experience Platform that supports four key stages of a data scientist’s workflow.
What Data is Available Within the Data Science Workspace?
The Data Science Workspace is fully integrated with all data contained within the Adobe Experience Platform via Adobe’s Query Service. This includes the Data Lake, Real-time Customer Profile, and Unified Edge.
Can I Use Third Party Data?
Yes! Anything from external CRM applications such as Salesforce to offline purchase data can be imported into the Data Science Workspace. Beyond traditional marketing data, this also allows for the incorporation of novel data sources such as weather or county-level economic trends. In short – if the data exists, it can be used.
How Does One Develop A Model?
Models are built in the Data Science Workspace using different ‘recipes.’ Users have the option of either leveraging a pre-built recipe or creating a custom one.
What Is A Recipe?
A recipe is Adobe’s term for the complete set of required input data and processing steps to build and train a model that solves a specific business task.
What Pre-built Recipes Are Available?
At the time of this post, there are three pre-built recipes available, with more planned for development. They are:
1 Prebuilt Recipe for Forecasting Retail Sales in Python
How Does One Build A Custom Recipe?
Integrated into the Data Science Workspace is JupyterLab, which is a programming environment popular with data scientists. This provides the analyst with a familiar tool to build new recipes and train models, with the caveat that basic coding is required.
What Custom Models Can Be Used?
Data Science Workspace allows the incorporation of essentially any modeling or analytical approach available in Python, R, PySpark or Spark (Scala).
2 Available Notebooks for Recipe Building
How is A Trained Model Operationalized?
Without writing any code, the model can be published as an intelligent service through Adobe I/O where it can be automatically updated with new data. The insights from the model are then seamlessly available through the Real-time Customer Profile for actioning.
So What?
By integrating access to the data, the Data Science Workspace allows analysts of all skill levels to concentrate more of their time on generating actionable and valuable business insights. These insights can then be consumed within other Adobe solutions such as Analytics or Target with only a few clicks.
Conclusion
While we are not quite at the stage of handing Skynet the keys to the marketing department, Adobe’s Data Science Workspace brings an enormous amount of power into the hands of an analyst. With its full integration with the Experience Data Model, its flexible modeling environment and ease of operationalization, Adobe’s Data Science Workspace promises to empower your analytics team to use state-of-the-art techniques to extract the full value of your data.
