Today’s State of Feature Management and Experimentation

Authors:

Table of Contents

In software development, success starts with the ability to consistently launch innovative products without interruption or incidents (or even worse–damage to revenue and reputation caused by outages). That takes the right tools and the right processes. To get a better understanding of where the industry stands today in its use of feature management and experimentation, we partnered with LeadDev to survey 500 engineering leaders. The findings reveal a strong tie between achieving success with Feature Management and Experimentation and leveraging feature-level monitoring. Also underscoring the report is the necessity for robust solutions like Split by Harness.

What Engineering Leaders Are Saying

Key Findings from the Report

82%

of those successful with feature management and experimentation have the ability to measure system performance and user behavior at the feature level.

79%

see risk mitigation and instant triage as a benefit of feature-level monitoring and experimentation when releasing new technologies and features.

78%

see feature iteration and optimization as a benefit that feature-level monitoring and experimentation delivers. (2nd place to risk mitigation and instant triage)

83%

expect integration between their feature management tool and existing CI/CD pipelines.

16%

succeed with feature management and experimentation without the ability to monitor system performance and user behavior at the feature level.

Key Takeaways from the Report

“Safety” and “Impact” Are the New Top Priorities

It used to be that speed was the top priority for software teams relying on feature management tools. Today, “safety” and “impact” have eclipsed that notion. In fact, 78% of survey respondents say risk mitigation and optimization play a central role in their feature management practice. And 54% are already using feature management tools for progressive delivery as a safety measure.

These findings are not a surprise. The challenges of mitigating risk at enterprises with over 500 engineers is not an easy feat, especially when teams are pressured to innovate quickly. Limiting the blast radius of a potential release issue with feature management techniques like progressive delivery is an effective way to roll out features safely.

Successful Feature Management Requires Release Monitoring

It’s also not a surprise that 82% of successful feature management respondents say they monitor system performance and user behavior at the feature level during their releases. Only 16% succeed without this capability.

Knowing the impact of every feature you release in a timely manner is an essential component to feature management. If your platform doesn’t have a way to do this, all you can do is cross your fingers and hope that nothing breaks.

Doing things right requires a feature management solution with release monitoring for every rollout or experiment and every kind of flag, an always on the job system that catches progressive delivery issues that typically fly under the radar. As a bonus, this approach automates the triage process, letting the right team focus on resolving the issue so the rest can get ship done.

Experimentation Is Becoming an Engineering Responsibility

Years ago, experimentation was seen as a product or even marketing team function. Today, that’s not the case. According to survey participants, experimentation is now a shared responsibility led by engineering 42% of the time and product teams 39% of the time. This shift underscores the collaborative effort required to harness the full potential of experimentation that drives product innovation. It also proves the importance of testing the impact of new features in production as a way to mitigate release risk shouldered by engineering teams.

Technical Debt Is Worse With Homegrown Solutions

Nearly half of the respondents acknowledge that homegrown feature flagging increases technical debt, highlighting the need for out-of-the-box solutions with automated flag removal. This advanced technology can minimize technical debt caused by unused feature flags.

Feature Flags Integrated With CI/CD Is the Next Logical Step

The feature management and experimentation platforms of the future will be deeply integrated with CI/CD pipeline solutions, as proven by 83% of leaders who expect such connectivity. An integration like this provides context and traceability, allowing developers to quickly understand the full picture of each feature flag and the associated code changes.

This is exciting news for Harness and Split customers moving forward. As Split becomes a native tool within the Harness application, the integration will offer a complete way to manage feature flags and CI/CD pipelines within the same enterprise platform.

The Industry Has Spoken:

It’s Time to Embrace a Proven Feature Management & Experimentation Solution

By this point, feature flags are table stakes. In fact, if you’re not using them, you’re already falling behind. One survey participant who didn’t use feature flags said it was “because the company is old and slow.” This statement is not a shocker.
However, despite the maturity of the category, around 70% of teams still rely on homegrown feature flag solutions, most of which have no ability to measure impact at the feature level. These are teams who still struggle to detect issues in a timely manner, even with the gradual releases called for by progressive delivery. These are the teams who fear that experimentation might slow them down. For these teams, here’s our advice:

It's time to transition to a platform that eliminates toil and unlocks the full potential of safe, impact-driven development.

If you're not yet doing feature management and experimentation with a solution that tracks the impact of every feature on your most important metrics automatically, you're missing out on the ability to innovate safely and efficiently.

Don’t wait. Download the full LeadDev report to explore these insights further and start your journey toward safer, more effective feature management and experimentation.