Killing a Feature

Killing a feature is a hard task for a product manager, because you can be sure that someone will complain, and that you break someone’s workflow.

But in order to build a great, focused product, you need to sunset a feature when it does not fit anymore.

Here is a guide on how to kill features properly:

  1. Confirm the condition is terminal
  2. Stop offering it to new sign-ups
  3. Divide users into two camps
  4. Announce end of life to remaining folks

Focus on actual tasks and jobs to be done

When planning a new feature or improvement, one should break it down to the very tasks and jobs that are to be performed by the feature. In other words, one should ask:

  • Why will I need this feature?
  • What would I like to achieve with this feature?

An example of such an analysis is made on self-driving cars: Exactly what jobs will self-driving cars satisfy?

POEM Framework

The POEM Framework is an opportunity evaluation framework for new business and product ideas. In proper software product management, whenever you plan a new feature or product, you need to make sure that it will be successful. Therefore, numerous factors must be evaluated. The POEM Framework provides a good overview of all factors to be considered.

The content itself is not new and should be common sense, but the POEM structure ensures that nothing is forgotten. Plus, it’s open source and free.

This post explains the framework and is a good starting point.


Features for very small user groups

Have a look at this feature: Yelp for iOS — When in New York City, the two tightest distance filters are measured in blocks, not miles. Isn’t this neat? I guess every user who uses Yelp in New York City appreciates this feature. It shows that Yelp is a user-friendly application that cares about the users’ needs.

I wonder, however, about the feasibility of developing such a feature. Yelp is a world-wide service, and only a minimal fraction of users actually live in New York or visit the city. The feature, despite being small, needed to be thought of, described, assessed, developed, tested, rolled out… this is a lot of effort and cost. Is it really worth implementing a neat feature for a very small user group? I think not. What is your opinion?