We can’t end our discussion of recommendation systems without surveying the different ways they can improve your product. I’ll skip the “similar products” interface you are familiar with in favor of some applications that might not be obvious:
- Interface Customization. Any modern product will have so many features that it can overwhelm new users. You want to hide that complexity from users at first, but what features do you expose them to? You can use a recommendation system, based on what they tell you during onboarding, to determine which features they are likely to enjoy and show them first.
- Selecting A/B Testing Groups. When rolling out a new product feature you can randomly select a group of users for an A/B test, but there is no guarantee that random sample will be interested in using the feature. A recommendation system can give you a strong indication of who is likely to use that feature and you can focus your A/B testing on that community to get a higher fidelity test result.
- Learning Customer Preferences. If your product has a wide variety of preferences available, it might be too much to expect a customer to manage them all. A recommendation system focused on the features of your product can produce a profile of customer interests and allow you to automatically configure the product for them as they use it.
In Review: Recommendation systems are not just for movies! By using recommendation system methods to predict customer satisfaction you can improve both your product and decision making.
Next Week: Even more important than what the data can tell you is how you communicate those insights to others. Data Storytelling is an art that can be used for both good and evil, so we’ll review how to communicate data insights to ensure you are on the side of good.
Quote of the Day: “His fear, he says, is that whatever he selects, the other option would have been better.” NY Times writer Alina Tugend