What are Experimental Features in Software?

Every open-source software application and the features in it are born out of ideas. Some ideas are phenomenal, and some are ineffective at the same time. So, it’s evident that not every feature is a good idea. But again, how to identify which idea is the best or is going to work for the software and which one would be terrible? Well, the answer lies in Experimentation! Only with experimentation, one can try out the new ideas, game plans, and activities in the software. Again, it is very different from testing. With experimentation, it is possible to explore trends, understand metrics, and make adjustments accordingly.

About Experimental Features

Experimental features are the features that are still under the development phase and not ready to be deployed into the actual software or product. Usually, these features are limited to just some modules in the software and are mostly incorporated to gain knowledge and derive comprehensive insights for future development in the software.

Why are Experimental Features Important?

Experimentation should be carried out regularly over the software applications and infrastructure for improvisation. They assist in validating developments and changes in the software and how it will have a positive impact on the user experience. As per the results obtained, you can either focus on the feature that had a better impact or brainstorm to incorporate a new feature. Additionally, instead of thinking about the discrete versions of the software, you can collectively work on discrete features that could be on and off for a specific user group.

How to Actually Experiment with These Features?

A/B Testing

Also, known as bucket testing or split-run testing, these are very common experiments. In this type, the experiment is aimed to compare the performance of two or more features and scrutinize them based on engagement, growth, adaptability, and other metrics. In the end, the better performing feature is deployed into the application.

Chaos Experiments

Chaos experiments aim at figuring out what doesn’t work rather than what works in favor of the software applications. The procedure involves intentionally incorporating an error or bug in the production environment of the software and tracking its effects on the performance of the software. For example, you can add wrong data in the database.

Multi-armed Bandits

Multi-armed bandits are just a modified version of A/B testing. Where you have to wait for a finite amount of time for experiment completion in A/B testing, this test uses algorithms to drive direct traffic to better performing features.

Enabling experimental features can bring exciting new features to the solution. So, play with them. However, test them out in non-production instance only, to see if they match your needs.