Disclaimer ❗️

I realized this test technical assignment for back-end development using Nestjs and TypeScript.

Source: Backend developer (any)

Test assignment: backend

We create mobile applications, and sometimes we need to run A/B tests to validate hypotheses. For this, we need a system that implements the simplest REST API, consisting of a single endpoint.

API and distribution

Upon launch, the mobile application generates a unique client ID (which persists between sessions) and requests a list of experiments by adding a Device-Token HTTP header. In response, the server provides a list of experiments. For each experiment, the client receives:

• Key: the name of the experiment. The client has code that will alter certain behavior depending on the value of this key
• Value: a string, one of the possible options (see below)

It is crucial that the device is assigned to one group and always remains in it.

Experiments

1. Button color

We have a hypothesis that the color of the "buy" button affects the conversion to purchase

• Key: button_color
• Options:
• #FF0000 → 33.3%
• #00FF00 → 33.3%
• #0000FF → 33.3%

Thus, after 600 requests to the API with different DeviceTokens, each color should be assigned to 200 devices

2. Purchase price

We have a hypothesis that changing the purchase price in the app may affect our marginal profit. However, to avoid losing money in case of an unsuccessful experiment, 75% of users will receive the old price, and we will test the change on only a small portion of the audience:

• Key: price
• Options:
• 10 → 75%
• 20 → 10%
• 50 → 5%
• 5 → 10%

Requirements and constraints

1. Once a device has received a value, it will always receive only that value

2. The experiment is conducted only for new devices: if the experiment was created after the first request from the device, then the device should not know anything about this experiment

Task

1. Design, describe, and implement the API
2. Add experiments (1) and (2) to the application
3. Create a page for statistics: a simple table with a list of experiments, the total number of devices participating in the experiment, and their distribution among options

You can use any technologies and libraries

The following will be considered a plus:

• Presence of tests
• A deployed version of the application
• Server response speed <100ms