Corrily can take user features other than location into account when it calculates experimental prices. A feature is any attribute of the user that you think may affect the optimal price for that user.

Supported values for user features

Any single-leveled object of key-value string pairs is valid user features data. Example:

  "platform": "ios"

The values can be strings, integers, or floats.

You can provide whatever keys and values you want. The keys and values have no intrinsic meaning to Corrily. To Corrily these are simply labels for training its AI models. For example, suppose that you provide a platform key every time you calculate a price. Sometimes the value is ios, sometimes it’s android, and sometimes it’s web. If the users who have their platform field set to web consistently convert at a higher price more than the ios and android users, Corrily takes this data into account when calculating future prices for any user with platform set to any of these 3 fields. The web users would probably be presented higher prices, and the ios and android users would probably be presented lower prices.

The important thing is to consistently provide the user features data as much as possible when calculating prices. In other words, you need a consistent set of keys, and each key needs a consistent set of values. Corrily’s AI models can’t learn if the keys and values are constantly changing.

How to set user features

  • Use the features parameter of the calculate price API endpoint when you want to take into account temporary or frequently changing factors that may affect the user’s optimal price.
  • Use the set user features API endpoint.