AWS for Industries

Succeeding with Industrial Data Platforms

Data is the key enabler to digital transformation. Retailers combine data to get a single view of customers to build better experiences and reduce acquisition costs, financial institutions use data to manage risks and personalize offerings, and manufacturers connect to data in production systems to lower unit costs or reduce quality issues. Across all those industries, linking data from disparate data sources into a single industrial data platform is the first step to unlock this value.

The technical aspects of building industrial data platforms are well understood. Yet even a well-architected industrial data platform can fail. This blog outlines three business-related mental models that can drive the success of an industrial data platform.

Working Backwards from Users

Industrial data platforms and their use cases are diverse. Not only will an industrial data platform vary depending on industry, but each business within an industry will have different needs for one. All too often, we see businesses embark on building an industrial data platform without a granular enough understanding of their users. What data do they want? How will they use the data? What business value will they gain? Just making data available and democratizing access alone rarely yields the anticipated business outcomes.

Before building an industrial data platform, it is essential to develop a deep understanding of the business’s needs for an industrial data platform. This means interviewing users across the business, documenting their requirements, articulating the value to the end users, and developing end-to-end user journeys. At Amazon, we call this process Working Backwards, and it results in a future-dated press release that articulates the value from the customer perspective.

Putting this work in upfront takes effort and time, but it pays off in measurable benefits. First, the experience will reflect expectations that users have, such as ease of use, speed, or flexibility. Second, a better user experience will lead to faster and higher adoption, which in turn results in more feedback that can be used to iterate and improve the industrial data platform. Third, business outcomes materialize faster and more reliably, because the industrial data platform was built with the user’s needs in mind.

Thinking Big, Starting Small

Industrial data platforms are typically long-term investments that can take months – or sometimes years – to complete. In light of this journey, businesses are keen to deliver value along the way. Balancing long-term scalability with short-term business value is challenging. We see two types of mistakes: 1. organizations start with architecture, governance, and processes and don’t define a path to value;  and 2. they build point solutions to use cases in the business that are not scalable across the business into a cohesive platform.

For an industrial data platform to succeed, i.e. reaching scale while generating continuous buy-in, businesses need to think big but start small. This means building an industrial data platform incrementally, use case by use case, while leveraging components that have been built to extend and scale the platform. In practice, this takes four steps:

  1. Define the industrial data platform vision and backlog, as well as its architecture, data standards, and data models.
  2. Make portfolio-level decisions to prioritize use cases so that each use case contributes to extending the capabilities of the platform along the vision.
  3. Build each use case in a way that it yields re-usable components and microservices that are small enough to be composable for other use cases.
  4. Ensure that components are incorporated into a cohesive industrial data platform that makes it easy to discover and implement them.

Building an industrial data platform in this way yields several benefits and can lead to a flywheel effect. The industrial data platform demonstrates immediate business value from use cases, while also extending platform capabilities in a directed manner over time. As the industrial data platform grows, the pace of use case development will accelerate. Additionally, the industrial data platform will receive continuous feedback from the use case adoption, which ensures development can be course-corrected or pivoted prior to investing in a large-scale rollout. A virtuous cycle entails.

Building through a product operating model

An industrial data platform has many stakeholders, spanning both business and IT. Leaders in business seek to influence the direction of the industrial data platform towards their needs, while IT leaders are often siloed between product managers, technology strategists, and developers, as well as functions such as the cloud business office, architecture governance, and security. We see many businesses struggle to define an operating and governance model for industrial data platforms that is inclusive of the key stakeholders while linking accountability and resources to ensure effective execution.

The core mental model for building an industrial data platform is to think of its components as loosely coupled products, each of which is formed around a value stream and has a single-threaded leader (STL). For example, each microservice yielded from use cases should have an STL who maintains and operates it, while a separate STL at the platform level ensures that the microservice, along with others, is easily discoverable and composable. Where this requires a team, it must be separable, autonomous, and aligned with the portfolio of value streams based on decomposing the architectural design into autonomous modules interconnected via APIs.

Building an industrial data platform as a collection of loosely coupled products that are owned and delivered by separable teams simplifies and accelerates the development process. It also reduces the overlap of responsibilities and with it the need for committees and extended alignment processes. As a result, industrial data platforms will be built faster and with greater accountability.

Conclusion

Building industrial data platforms is hard, both from a technical and a business perspective. But the business value they deliver makes the effort worthwhile. For businesses to release this value, they need to work backwards from their users, think big but start small, and assign clear ownership to  each of the industrial data platform components.

Further Reading

Rishi Kumar

Rishi Kumar

Rishi Kumar is an Innovation Delivery Specialist in Innovation and Transformation Program at Amazon Web Services (AWS). In his role, Rishi leverages the Amazon’s working backwards mechanism to support customers across industries through their innovation and transformation journey. Rishi is passionate about helping customers with their data platform strategy and works with industrial customers to shape their data platform vision and define initiatives and roadmap to achieve that vision.

Peter Gratzke

Peter Gratzke

Peter Gratzke is part of the Innovation and Transformation Program team at Amazon Web Services (AWS). He supports enterprise customers to build new products and businesses, and transform to become more innovative.