7 takeaways to "Accelerate" your DevOps

In the age of digital disruption, success increasingly is tied to speed. Financial services firms compete by trading faster and more securely. Retailers win customers by offering superior real-time experiences. Even government agencies are transforming themselves to serve their constituents more efficiently and effectively.

In 2003, Nicholas Carr made waves with his argument that IT doesn’t matter. But the authors of the new book Accelerate: The Science of Lean Software and DevOps found that, not only does IT matter, but it’s the key to creating competitive advantage as well.

Authors Dr. Nicole Forsgren, Jez Humble, and Gene Kim launched the largest DevOps research project to date to investigate how the most innovative organizations are leading the way in using DevOps principles and practices. The key to moving faster, they argue, is building and scaling a high-performing technology organization.

“It’s not universally accepted yet, but high-performing, innovative organizations have embraced [DevOps] and are leveraging it to their advantage,” says Forsgren. “The faster you move, the more likely you are to be able to delight your customers, to beat your competitors to market, to pivot when you need to pivot. Even in situations where you don't have competitors, as is the case with the federal government, the ability to move faster puts you at a distinct advantage.”

Here are 7 takeaways from Accelerate.

How to Build a DevOps Toolchain That Scales

Inside 'Accelerate': Key findings

The book, co-authored by Forsgren and her DevOps Research and Assessment (DORA) co-founders Humble and Kim, goes beyond anecdotes and case studies of DevOps adoption to quantify what constitutes a high-performing technology function and the business results such technology functions can deliver. Forsgren, Humble, and Kim spent four years “sciencing the crap out of DevOps,” as Forsgren puts it, analyzing data from the State of DevOps reports (presented by DORA and Puppet Inc.) to create a capabilities model for software-delivery performance using rigorous statistical methods.

[ Special Coverage: DevOps Enterprise Summit 2018 London ]

As the DevOps movement takes hold more broadly—from the early adopters in technology, telecom, and financial services to more recent implementations in healthcare, government, manufacturing, and nonprofits—the research provides a playbook for continuous improvement. “We do this work so that we can understand how to help these organizations continue to be better,” says Forsgren.

Accelerate offers new insights into what enables both software-delivery performance and organizational performance, as represented by profitability, productivity, and market share. It provides readers with information on how to measure the performance of their technology teams and what capabilities they should invest in to drive higher performance going forward. Here are some of the book’s key takeaways.

Case studies are good, but not good enough

Stories of DevOps adoption have clear value, but more rigorous research into what works and what doesn’t is the key to helping more organizations adapt their approaches. “The two go together and inform each other,” says Forsgren, who says ethnographic research and in-depth case studies help shape and inform the research in Accelerate.

Forsgren and her co-authors analyzed 23,000 survey responses from around the world from thousands of organizations—both startups and global enterprises, for-profit and not-for-profit, those with legacy systems and those born digital. As a result, the quantitative analysis is more actionable for any company than is a single case study. “These broader results give you applicability,” Forsgren says. “We can say, 'We’ve seen this work across dozens of industries and thousands of teams around the world, and it is highly likely to work for you, too.'”

Maturity models won’t help you move faster

Maturity models are a popular way to benchmark technology performance, but Accelerate’s authors argue that they are not the appropriate tool to use—or mindset to have—when it comes to organizational or technology transformations. “Maturity models focus on a destination,” says Forsgren.

“The concern is that the destination will change as the industry changes or the landscape changes. Also, once you arrive at that destination, your resources—not just money, but also time and attention—may disappear because your leadership has decided your transformation effort is complete.”

A capabilities paradigm is more effective when transforming complex organizations such as the technology function. It allows for the fact that progress is not always linear and the destination continues to change with market and customer demands.

DevOps performance isn’t determined by the factors we thought it was

The prevailing wisdom has been that things such as the age of applications, the question of whether operations or development handles deployment, and how detailed the planning processes are will have the greatest impact on the technology function’s performance. Not so, according to this research, which uncovered 24 key capabilities in such areas as automation, architecture, lean management, and culture that drive improvements in software-delivery performance, and in turn, organizational performance.

The impact of DevOps on business performance is even greater than anyone thought

The most exciting discovery came during the first year of research, says Forsgren, when the team found early evidence that IT performance does matter. This showed that companies with high-performing technology organizations were twice as likely to exceed their profitability, productivity, and market-share goals.

“At first the team was, like, yeah, of course it does, but ... this was actually a really big deal,” says Forsgren. “For decades, academic research had been unable to find a good link between IT and overall organizational measures like profitability, productivity, and market share.” Companies could buy a server and throw it in a closet or purchase software and implement it, but so could their rivals; those technology investments were not creating sustainable competitive advantage.

DevOps is different. It enables companies to “combine technology, process, and culture to create something unique, and suddenly that’s showing up on the bottom line and providing a point of competitive distinction,” says Forsgren.

The gap between high- and low-performing technology organizations continues to grow

In 2017, the DORA research team found that high performers executed 46 times more code deployments, and had a lead time, from commit to deploy, that was 440 times faster; a mean time to recover from downtime that was 170 times faster; and a change failure rate that was 5 times lower. And while the gap widened for reliability, it narrowed in measures of throughput.

“What we think we’re seeing there is that high performers have been maxing out on their ability to develop software with speed and stability every year. They’re optimizing both,” says Forsgren. “Low performers, we believe, have been under lots of pressure in market. They have been pushing to catch up in terms of throughput, but they’ve been doing it in reckless ways. They’re trying to keep up in terms of speed, but it’s coming at the cost of stability, because they aren’t developing their capabilities.”

Low-performing teams may not be investing enough in building quality into the process, resulting in larger failures that take more time to recover from, while high performers understand that they don’ t have to trade speed for stability because, by building quality in, they get both.

Architecture matters; technology doesn’t

One of Forsgren’s favorite takeaways is that a company’s actual technology stack has nothing to do with its overall performance. “The subtext,” she explains, “is that there’s no excuse.” It’s not legacy systems that are holding IT organizations back; nor are born-digital companies guaranteed to outperform.

Working in a mainframe shop was not statistically correlated with performance, nor was having a greenfield environment. “Where we did see meaningful differences was in how systems were architected,” Forsgren says. Having a loosely coupled architecture was the biggest determinant of performance.

Improvement is possible for everyone

The researchers found teams from every industry, in every size of company, in both highly regulated companies and unregulated companies, and in startups and global behemoths in each of the high-, low-, and medium-performing groups.

The bottom line, they say, is that performance improvement is attainable for any organization as long as leadership provides consistent support, including time and resources, and team members commit themselves to the work. Even for the highest-performing organizations, there’s always more work to be done.

“My co-authors and I have been in technology for years and have seen organizations fail. To quote W. Edwards Deming: 'Survival isn’t mandatory,'” says Forsgren. “If you aren’t continually pushing and moving forward in ways that drive value to the organization and value to customers, at some point you will be passed by.”

For more on Forsgren's lessons learned from "sciencing the crap out of DevOps," don't miss her presentation at DevOps Enterprise Summit London on June 25-26. And for more DevOps must-reads, see our list of top picks.