DEV Community

zsuzsi-kun for ConfigCat

Posted on • Originally published at

Switches for future - feature flags in product development

How using feature flags makes your product development successful and safe

By the given example you can follow the process from idea to satisfied customers

Software (or even hardware) development could be a real challenge. Not just because it needs all your attention and a lot of time to do but you cannot be sure (even if you made research on the market) that your customers would like the new feature or product.

That is the reason why product development is not just a part of a job, but it is an independent work. It needs specialists, managers, tactics, and tools too. Having a good idea is far too little to give off the perfect software product - you are going to need the best people and the best tools for that.

But what if you have a really good idea and all the needed people and tools to start? On the stage of development, you have only a short period of time and you do not have too many chances: we can say that you have only one. Thus, you are going to need not just the previously mentioned people, tools and methods but some sector-specific solutions are indispensable too.

Feature flags for software development are like prototypes and tests for mechanical engineering.

This special toolset given by feature flags could guarantee a light and less risky process. To show you how it works and how can you implement it in your development process we give you an example. So have a cup of tea, let's enjoy the ride and get familiarized with the world of feature flags by using a smart interface developed by ConfigCat's team!

What is a feature flag? How can it be used for targeting?

Feature flags are like small switches in your code. They can be used to start or switch off given functions - in this way they can be used to control which users are able to see or use a specific option, feature or program part. Using of feature flags can make much easier the testing process since you can choose your target audience by a switch - you do not need any complicated choosing methods or rewrite the code by each testing step.

From a problem to an idea and just one more step to the solution.

A feature for university student which makes the planning and applying on a course easier.

Neptun unified education system is a web-based administrative interface that can be used by not just the administrative body of a university, but it is open for tutors, teachers, and students too. The system is made for all the administrative tasks since the registration at the beginning of a semester to logging into any classes.

A few years ago, a new feature appeared in the system: the schedule (or timetable) planner. This function enables the users to see the preferred classes in a table - as they have registered to these courses, but it is just a plan for the semester. This feature makes easier the process of applying for the selected courses too. Now we are going to see how we can manage the introduction of a new feature like this.

In this article, we are presenting this process by using ConfigCat but this is not the only tool to use in a case like this, you can manage it with another feature flag tool too. For example, you can use the product of Unleash.

1. step: Idea and development

Since this is a part of the software like any other function, the team must develop it. After writing the codes and finalize the first version, the testing process can be started.

Get familiarized with feature flags at this point. Using ConfigCat you can easily create switches for specified groups. These groups could be your testers or the small circle you do not want to show the latest release of your function. It is up to you, but without feature flags, the testing and implementation is a way more complicated process. Connect ConfigCat to your written code and let's start the testing!

2. step: Choosing the right audience for the first test

As you can feel, this is not a change you can easily build in the code and let rock. Minor changes must be tested - so one bigger like the scheduler must be too. And this is the point where feature flags step on the stage.

After you build the code in the program or a parallel test version of your program, you can implement some feature flags. First, let's chose the developer team - they can easily notice all the smaller issues at first sight.

Alt Text
Fig. 1: Feature flag for the first tester group: developers

ConfigCat's intuitive interface lets you develop feature flags based on the role of the users or any other identification attributes like beta testers, age group - the borderline is given by the collected data from the users.

Here I would like to mention that this is one of the good reasons why you should collect all the needed and important data from your registered users. There are some data you do not have to ask for (e.g. country, used platform, browser) since they are accessible by using cookies or simply log in to the app. But other parameters like age, sex, or program they are enrolled should be asked for.

Well, now we have chosen the group of the developer teams. This targeting strategy works well if we did give attention to tagging the people using our software. (No, it is not a waste of time, as you can see!)

After they accepted the changes and did not find any major issues, the testers' range could be widened.

Step 3: Giving access to more and more users

You can continue the testing process step-by-step. After giving access to your developers, you can find beta testers and your friendly users. (Friendly is a tag you can give to your user based on his or her activity and mindset.)

In ConfigCat, you can manage these flags as you can see on Fig. 2.

Alt Text
Fig. 2: Managing different feature flags in ConfigCat

You can even merge some lines by using the same attribute but different identification:

Alt Text
Fig. 3: Merging identifications

In case of the scheduler in Neptun we can say that the beta tester group can be selected from the volunteering testers or by random. The friendly ones are users who have given some feedback before. Since we need a lot of feedback now, it is not a negligible aspect and we can use their opinion later in the public release process for marketing purposes.

Step 4: Start the test process on a selected faculty

Like Facebook does with countries, we can implement new features by faculty to faculty. In this example, we have chosen GPK (Faculty of Mechanical Engineering) and have made the new function available for students on this faculty.

Alt Text
Fig. 4: Giving access to students of GPK in ConfigCat

As you can see on Fig. 4. students of other faculties are not able to see this new feature. We can manage different testing periods by changing the accessibility and see which faculty is the most open for the new function.

This method is like the technique used by bigger developer teams, like Facebook, etc. They usually test a new function in a selected country and after evaluating test results they make the decision if the new function could be implemented globally or should be eliminated from the codes.

Step 5: Start testing on the whole university

After trying out the new function on a faculty we can start the 'global' tests. The function could be reachable for all university students.

Alt Text
Fig. 5: Giving access to all the faculties at the university

In some cases, there could be a smaller group, the skeptics and you should be careful with them. They are the users are not open to new functions and features. Since they can be decision-makers you should evaluate the risks before you make the given feature available for them. They are sensitive to the changes and they notice all the minor problems too. Maybe you should make another global test before giving them the right to see the new development. You can lock them out by creating a category for them as you can see on Fig. 6.

Alt Text
Fig. 6: Making unavailable a function for a specific group of users

In case of Neptun we would lock out the 'older generation' as they are less open to the new features (in this case they are Ph.D. students and teachers).

At the beginning of the testing process, we can choose only a few students from the university. In ConfigCat this can be done by making a target in %, as you can see on Fig. 7. The percentage of the users can be adjusted based on the feedback.

Alt Text
Fig. 7: Selecting only 5% of the whole university and locking out the sensitive users

Step 6: Start the new feature as a built-in option and make it accessible to all users

After made all the steps above you have finally arrived at the most exciting part of a development process: implementation. Now you can turn all your switches on - giving enough light to your work and its results.
Now the sensitive decision-makers and the skeptics can try the new feature out. In case of Neptun we should finally make enabled the new features for the 'older generation' too.

After giving green light to the new functions we would like to think that our work is done. But I can tell you, it is not true at all. In this century you must be eager to new development opportunities not just for getting on top of your sector but to remain competitive in it.

ConfigCat can help you make your development process safe - but it cannot work instead of you. It is just a tool like feature flags are in the testing process. We are gladly helping you with not just our product but with our experience too.

Try out ConfigCat on our website and please do not hesitate to contact us with any doubts of yours!

Top comments (0)