For the 90 Days of DevOps community, I presented the concept of the digital factory. After years of doing DevOps transformations across industries at Zühlke, I have developed a holistic approach to scaling DevOps that goes beyond just tools and pipelines. In this talk, I explain why we still struggle with walls of confusion, how platform engineering enables teams to do DevOps at scale, and how digital factories bring everything together.
The Broken Value Stream#
The pattern I see in company after company is the same. The business has great ideas. They put them into Word documents and Jira tickets and throw them over the wall of confusion to development. Development builds something and throws it to testing. Testing verifies something (that rarely matches the original specification), and throws it to operations. Operations says “this will never work in production” but somehow manages. The customer gets the result and says: “That is not what we wanted.”
The value stream is broken by these walls of confusion. They come from silo organizations with different goals, missing alignment, slow and ineffective processes, and cultural resistance. Security and quality become afterthoughts instead of built-in practices.
From Projects to Products#
These problems originate from how we organize work. In the past, we did waterfall projects with fixed scope, budget, and time. Then agile brought smaller increments, but we are still doing projects. Our clients want products.
A project focuses on output: maximize features, user stories, tasks, and code. A product focuses on outcome: understand the customer’s need, solve their problem, change their behavior. DevOps supports this shift because it is a mindset, a culture, and a set of technical practices that organizes people across the value stream to continuously deliver value.
“DevOps is about bringing all the people, process, and technology together to continuously deliver value.”
The 24 Key Capabilities from Accelerate#
The book Accelerate scientifically identified 24 key capabilities that drive software delivery performance. They fall into five categories:
- Continuous Delivery: Version control, deployment automation, continuous integration, trunk-based development, test automation, test data management, shift-left security
- Architecture: Loosely coupled architecture and empowered teams (teams of maximum five people with clear inputs and outputs)
- Product and Process: Customer feedback, value stream mapping, working in small batches, team experimentation
- Lean Management and Monitoring: Change approval processes (which science shows add no value and often slow things down), monitoring, WIP limits, visualizing work
- Culture: Westrum organizational culture (pathological, bureaucratic, or generative), supportive learning, job satisfaction, transformational leadership
The research also maps the relationships between these capabilities. When you want to achieve the outcomes on the right side, you invest in the capabilities on the left side. This is one of the most important pictures in DevOps.
The Tesla Example#
To illustrate what modern continuous delivery looks like, consider Tesla. On October 7, 2021, Elon Musk tweeted that the self-driving module FSD 10.2 would roll out to 1,000 owners with a perfect safety score. This tells us several things: the software is modularized, over-the-air updates work, they monitor driving behavior continuously, and they can target specific user groups. That is a canary release.
On October 15, version 10.3 rolled out to a larger group. On October 24, they rolled back to 10.2 because of a problem. In a regulated industry, with cars on the road, they performed a rollback. Not even 24 hours later, they deployed 10.3.1 as a fix forward. Many companies cannot do rollbacks with their software. Tesla does it with hardware on the road.
The Cognitive Load Problem#
Modern software development requires an enormous set of technical practices: infrastructure, runtime, CI/CD, monitoring, security, tooling, cost management, maintenance, and access management. And somewhere in there, you also want to build an application.
When you scale this across multiple teams, each team builds its own stack. This leads to inconsistencies, redundancies, lack of operational experience, no synergies, and difficulty moving people between teams. The cognitive load becomes crushing.
Platform Engineering: The Foundation#
Platform engineering solves this. A platform team builds a product, the platform, that provides standardized capabilities to product teams:
- Application runtime (environments, Kubernetes clusters)
- DevSecOps (vulnerability scans, license scanning, container scanning)
- Access and identity (centralized authentication across all tools)
- Monitoring and observability (pre-configured dashboards and alerting)
- CI/CD pipelines (standardized, ready to use)
The product teams build, run, and maintain their products on top of this platform. They still own their CI/CD pipelines, still monitor their applications, still respond to incidents. The platform team just enables them to do DevOps without reinventing the wheel.
This is not a new silo. The platform is a product that teams want to use. The platform team generates value for the teams. The product teams generate value for the customers.
The Digital Factory: A Holistic Approach#
A digital factory brings everything together. Picture a company that builds drones (or replace “drone” with your software product):
Board level: The management has a vision (increase market share) and a portfolio of initiatives. They prioritize: build a drone that can carry heavy weights.
Product level: Product managers take this epic and design features: bigger battery, updated software, new engine.
Team level: Existing teams work on software and battery changes. For the new engine, a new team needs to start. The platform team provides a standardized CI/CD pipeline so the new team can begin with zero delay.
Delivery: All parts are assembled and delivered. The customer is happy.
Feedback loop: Telemetry data from the drones flows back to the teams for continuous improvement, and also back to the board for strategic decisions.
This is what I call a digital factory. Lean portfolio management at the top connects strategy with execution. The platform team at the foundation enables DevOps. Product teams in the middle deliver value.
Creating Great Products Requires a Holistic View#
Building a digital factory is not just about DevOps and platform engineering. You also need:
- Scalable architecture with a modular, API-driven approach
- Data management to make sense of telemetry and business data
- Customer experience with an end-to-end focus on the customer journey
- Agile program delivery to manage backlogs, dependencies, and team alignment
- Product management to connect strategy with execution
Key Takeaways#
- Move from projects to products. Focus on outcomes, not outputs. Put the customer at the center.
- DevOps is a holistic approach that brings people, process, and technology together across the value stream.
- Platform engineering reduces cognitive load and enables product teams to do DevOps at scale through standardization and self-service.
- Digital factories are the holistic model for industrialized software development: lean portfolio management at the top, a platform at the foundation, and empowered product teams in the middle.
- The feedback loop never stops. From telemetry back to teams, and from business metrics back to the board, data-driven decision making keeps the factory running.
- We are entering the age of industrialized software development. Platform engineering is the foundation of the digital factory, and the digital factory is how we continuously deliver value.