Research & Development at DISH Wireless: Data Products

By Darshit Gandhi, Hamza Khokhar, Vinayak Sharma

Data researchers and professionals are buzzing about data products, but what exactly are they? In this article we will cover the basics of data products and discuss data product use cases.

What are Data Products?

A data product is like any other product: a ready-to-use object. Think about raw data as the ingredients for a recipe (e.g., flour, eggs, butter). Just as you can take ingredients and bake it into a dessert, you can take raw data and turn it into insights, predictions or strategic decisions using code and infrastructure.

At DISH Wireless, we’re researching and developing data products using data that’s been generated within the DISH SMART 5G™ network. We started this process by defining data products as:

“All data, linked to a data platform or a data mesh, that can provide business value when they are easily intelligible and have comprehensive documentation explaining methodologies required to access data and elaborated metadata”

Let’s break this definition down further. It starts with understanding the concept of data mesh, an idea coined by Zhamak Dehghani in her book, Data Mesh: Delivering Data-Driven Value at Scale. She defined data mesh as:

“Data mesh is a decentralized sociotechnical approach to share, access, and manage analytical data in complex and large scale environments within or across organizations. Data mesh is a new approach in sourcing, managing, and accessing data for analytics use cases at scale.”

Meaning, data mesh is an approach rather than an entity.

Another important component of the DISH Wireless definition of data products is the data platform: a centralized area that catalogs datasets from distributed sources.

Our goal in Research & Development is to publish data products as applications that are available on the DISH Wireless Developer Hub, a portal DISH built for the developer community. Our vision is that DISH Wireless data products will enable all types of users to build the next generation of disruptive technologies that harness cloud-native 5G network capabilities.

How to Identify a Proper Data Product

Data products become useful when they have certain features. Examples include:

1. Linkage to data mesh and a data platform so that its output can be used by other data products.
2. Metadata Comprehension: Any user who accesses the data and its metadata can easily make sense of what the data is showing, where the data is collected from, and potential use cases of the data
3. Access: Documentation that shows the methods to access the data (e.g., what are the API parameters?, SQL query, etc.).
4. Well-maintained metadata: The data product owner and any of its vendors share the responsibility of metadata maintenance. Maintenance processes should be automated for maximized efficiency.

Can You Change Data into a Data Product?

Imagine you have executed a query on AWS Athena, a serverless query service, and the results are stored in S3, a data storage service. Now this data would not be considered a data product until it satisfies the following requirements:

→ Reliable foundation/source: It’s built above all the raw data coming from batches, streaming data and files.

→ Metadata bank: This provides information about the rows and columns in a data set. Some characteristics that are valuable in a metadata bank include:

• Purpose of the data: use cases applicable with this dataset
• Category of the data: organization of the data. An example in the telecom world, categories of data group data as performance data, data logs or error logs.
• Owner: who generated the data and when was it generated
• Statistical values: mean, median, mode, counts, min./max., etc.
• Error handling: rules that detect and handle errors
• Rating the data: same as rating any product on e-commerce website

→ Centrally managed data: All data and metadata are found in one true source.

→ Versioning: Data products are versioned appropriately. All changes to the dataset are tracked and well documented.

→ Secure: Access to the data is only given only to authorized users.

Why Should Industries Be Excited About Data Products?

Data products have become much more significant with the advent of data mesh architecture. A data mesh architecture enables users access to large chunks of data without having to do any pre-processing tasks or move data to a centralized source, like in a traditional data warehouse approach.

Data products by DISH Wireless serve businesses by providing a large variety of ready-to-use data for which different users can consume for their business needs. Imagine a company that builds autonomous vehicles, for instance. The engineers at this company want to build better autonomous vehicles on a grid network utilizing the DISH SMART 5G™ network or even a different MVNO. Engineers can browse data products on the DISH Wireless Developer Hub, which can help the engineers to optimize their network slice, find insights about their network slice and utilize models from other autonomous vehicle vendors on the DISH network. Having easy access to insights and data products allows the engineers to think more analytically about problems, rather than spend time researching.

Categories of DISH Wireless data products include:

1. Application

• A complete product which serves a data problem. One example is the DISH-built Automated Contract Governance product. This product extracts key contract segments and provides AI-generated recommendations on whether to accept, decline or renegotiate terms of the contract.

2. Framework

• A product which etches a set of processed data enabling the customer to configure data according to their needs.

3. Dashboards

• A visual interpretation via business intelligence tools or custom made python dashboard. For instance, EKS architecture tree map.

Understanding data products are key to transforming our network and should be part of every developer’s process. The network is a data generating machine, which yields an avalanche of data that our team and the developer community can turn into incredible data products for enterprises.

About the Authors

Darshit Gandhi is Lead Data Analyst who is working towards creating Data Products and its process. He has over 7 years of experience working with data in the technology industry.

Hamza Khokhar is a Data Scientist working towards creating machine-learning (ML) powered Data Products. He has over 3 years of experience as a software and computer vision engineer.

Vinayak Sharma is a Data Scientist focused on machine learning research to create a self-healing network at DISH Wireless. He is passionate about creating value and solving hard problems, and he hopes to change the world with his work on Data Products. His background is in parameterizing earth models to assist with global climate change at CU Boulder.