The manufacturing world embraced lean principles decades ago, so why has the concept struggled to find a foothold in software development?

The latest State of DevOps Report from Puppet Labs, Inc. provides compelling evidence that traditional throw-it-over-the-wall approaches to programming make less and less since in the face of DevOps, a lean-like discipline that emphasizes rapid deployment, continuous testing, collaboration and platform integration.

Understand that Puppet Labs has a vested interest in promoting DevOps, since it provides automation tools for software development. However, its four annual surveys on this topic have demonstrated rigor that is rare in vendor research. The research merits the attention of any one who is trying to make a case for advancing DevOps in an organization.

The top line results of the just-released 2015 survey are compelling: High-performing IT organizations deploy code 30 times more frequently with 200 times shorter lead time and 60 times fewer failures. They also recover from failures 168 times faster. Interestingly, the speed and frequency of deployment in this survey didn’t increase from last year’s results, but the meantime to recovery and change success rates were significantly higher. This indicates that the productivity boost from DevOps eventually tops out, but the resulting code is more stable and easier to modify. That’s critical in the business world that demands constant adaptation. Not surprisingly, high development performance tended to correlate with business success.

A matter of culture

The survey authors stress that success with DevOps is more a function of culture than of process or tools. Performance-oriented organizations are characterized by high levels of cooperation, openness to constructive feedback, shared risk and willingness to experiment. They engage in what the authors call “blameless postmortems,” in which mistakes are treated as learning opportunities rather than as failures.

Perhaps this explains, in part, the plodding acceptance of DevOps in corporate IT organizations. Programmers tend to think of themselves as lone wolves whose code is the equivalent of artistic expression. Encouraging individualists to collaborate can be a challenge, but the authors point out that a few basic management tactics can help.

Sharing risk means that everyone is rewarded equally for the contributions of the team. One simple way to do this is by establishing performance metrics and displaying them for everyone to see, whether it be code defects, delivery performance or customer satisfaction. In the same way that many factories proudly display the number of days since the last on-the-job accident, IT organizations can use displays to keep everyone focused on the real goals.

Another is to avoid finger-pointing at all costs. This is a difficult task at risk-averse organizations that routinely punish failure. I suspect that is still the case at the majority of large organizations. Managers can play a critical role by limiting use of the words “you” and “I” in team meetings in favor of more inclusive terms. They can also put rewards for team achievements ahead of individual incentives.

Use of tools like configuration managers, version control, automated testing and shared code repositories also correlated strongly with success. The arrival of containers should turbocharge those organizations that have already embraced DevOps.

One dismaying finding was the appallingly low representation of women on development teams. Only 5 percent of the nearly 5,000 respondents identified themselves as female, and more than half of the respondents said they work on teams that are less than 10 percent women. Thirty years ago this kind of representation led to the creation of government affirmative action programs. The tech industry could use a similar kick in the pants. Not surprisingly, the Puppet Labs report found that teams with the most diverse membership were also the highest performers.