# Step 3 · Treatment Phases

Here you define the trial's **arms**: the conditions participants can be in, and set
your **recruitment plan**.

## Defining the arms

Each arm is a row in the treatments table. For each one you set:

Name
: A label for the arm.

Type
: One of **Active**, **Sham**, **Washout (nothing)**, or **Washout (unmodulated)** (see
  {doc}`../concepts`). A sham is paired with the active arm it mirrors so blinding lines
  up correctly.

Sound type
: For active and sham arms, the category of sound this arm plays (configured in
  {doc}`6-audio`). Washout (nothing) arms have no sound.

Duration (days)
: How long a participant spends in this arm.

## The enrolment plan

Below the arms is the **Enrolment plan**. This is where you tell the portal how many
participants you need and let it work out the recruitment ceiling for you.

Participants to complete (per arm)
: For each active or sham arm, the number of participants you expect to **complete**
  that arm. This is a soft goal, arms may overshoot it, and it does two jobs: it adds
  up to the trial's total, and it sets how likely a new participant is to be assigned to
  that arm.

Estimated completion rate
: The fraction of enrolled participants you expect to finish (a number from 0 to 1,
  default 0.5).

From these, the portal shows a live read-out that updates as you type:

- **Total to complete**: the sum of the per-arm targets, in participants.
- **Total enrolment limit**: total to complete ÷ rate, rounded up. This is how many
  participants may be enrolled at once.
- **Randomisation %**: each arm's share of assignments, equal to its target ÷ the
  total.

:::{admonition} Worked example
:class: note
Three arms with targets **50 / 25 / 25** and a completion rate of **0.5** give
randomisation probabilities **50% / 25% / 25%**, a total completion target of **100**
participants, and an enrolment limit of **200** (100 ÷ 0.5).
:::

:::{important}
Targets are **soft goals**. Randomisation uses these fixed shares and does **not**
rebalance as arms fill, so final arm counts will scatter around the targets rather than
hit them exactly. Recruitment ends when the total completion target is reached and
pauses at the enrolment limit, never per arm. The full behaviour is in
{doc}`../enrolment`.
:::

Every active and sham arm needs a completion target before the trial can launch, since
the total and the randomisation shares are undefined if one is blank.

Next: **{doc}`4-schema`**.