AAA: how to structure a test

One of the most widely used patterns for structuring a unit test is AAA: Arrange, Act, Assert.

Arrange → Act → Assert

Arrange sets up the state the test needs in order to run. Act runs the behavior we want to verify — it’s generally better to have a single main action, so the test’s purpose is obvious. Assert checks the expected result.

describe('calculateCancellationFee', () => {
  it('charges 50% of the deposit when cancelled with less than 24h notice', () => {
    // Arrange
    const deposit = Money.of(2000);
    const hoursBeforeReservation = 3;

    // Act
    const fee = calculateCancellationFee(deposit, hoursBeforeReservation);

    // Assert
    expect(fee).toEqual(Money.of(1000));
  });
});

The value of AAA isn’t in writing a comment on each block. Its real benefit is that it makes the test’s intent obvious — given this state, when this behavior happens, then this should be true — which is why it lines up naturally with the Given/When/Then format from What is testing.

One action per test. If the Act block has three calls, the test is usually verifying three behaviors and its name can’t describe all of them. When it fails, the message doesn’t say which one broke.

A single concept, not a single line. It doesn’t mean a single expect line: it means a single concept verified. Checking the four fields of a newly created Reservation is one concept; also checking that the email was sent is another, and it goes in another test.

When ReservationService depends on other components, you have to decide what to replace and with what: that’s what test doubles are about.