> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getswan.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Scheduling and cadence

> How Swan paces sends across the day, the sending window, and roughly how many messages go out per sender.

Swan doesn't fire off every queued message at once. A separate scheduler runs for each step type, and each one moves contacts forward a little at a time, so sending accounts build up a natural, human-like pattern of activity instead of a burst.

## How pacing works

Each step type has its own schedule:

| Step type                   | Roughly how often                                    |
| --------------------------- | ---------------------------------------------------- |
| Email                       | Every 3 minutes                                      |
| LinkedIn message            | Every 15 minutes                                     |
| LinkedIn connection request | Three fixed times an hour — roughly every 20 minutes |
| Wait                        | Every 3 minutes                                      |

On every pass, Swan advances at most one contact per sender for that step type — so if a sender has ten contacts waiting on an email step, one moves forward per pass, not all ten at once. Wait steps are the exception: every contact whose wait has elapsed moves forward on the same pass, since letting a wait run its course doesn't put any load on a sending account.

The net effect, before any per-sender limits kick in, works out to roughly:

* Up to about 20 emails per sender per hour of sending window — though the default per-sender limit of 75 emails/day caps this in practice, so 75 is the effective daily ceiling
* Up to about 4 LinkedIn messages per sender per hour — capped by the default 40/day limit
* Up to about 3 LinkedIn connection requests per sender per hour — capped by the default 25/day (120/week) limit

Actual sending limits (see [Sending limits](sending-limits.md)) are checked independently and can slow this down further — pacing controls *how often* Swan tries to advance a contact, limits control *how many* sends are allowed per day.

## Sending window

Each org has a sending window that controls when regular messages (emails and LinkedIn messages) go out — by default, weekdays (Monday–Friday) between 9:00 and 19:00, in the sequence's timezone. Outside that window, Swan holds those steps until the window reopens; it doesn't queue up sends to fire in a burst when the window starts, since pacing still applies once it does.

Your org can configure a different set of days and hours for this window at [Tools](https://agent.getswan.com/tools) → Outreach → Configuration.

LinkedIn connection requests and wait steps are exempt from the sending window and can be processed at any time of day — a connection request doesn't put the same load on a sender as a message thread would, and a wait step isn't sending anything at all.

Sequence timezones use standard IANA time zone identifiers (for example `America/New_York`, `Europe/London`, `UTC`).

## Wait steps

A wait step pauses a contact's progress for a set amount of time (minutes, hours, days, or weeks, depending on how the step is configured) before moving to the next step. Wait steps aren't affected by the sending window — the clock runs continuously, and the next real send after the wait still respects the window and pacing rules above.

## Common questions

**Why didn't my sequence send anything overnight or over the weekend?**
Emails and LinkedIn messages only go out inside the sending window (default Mon–Fri, 9:00–19:00 in the sequence's timezone). Connection requests and wait steps aren't restricted this way.

**Can I set a different sending window per sequence, or per recipient's timezone?**
The sending window is configured at the org level and applied using the sequence's own timezone setting, not the recipient's local timezone.

**Why does LinkedIn move slower than email?**
Email is checked every 3 minutes, LinkedIn messages every 15 minutes, and connection requests three times an hour — so LinkedIn steps naturally advance less often than email steps, on top of LinkedIn's own lower daily caps.

**How fast will a new sequence start sending after I approve it?**
The first send happens on the next scheduling pass that falls inside the sending window, subject to normal pacing and limits — not necessarily immediately.

**Do wait steps count against the sending window?**
No — the wait itself runs on its own clock and is checked every 3 minutes regardless of the window. Only the actual email or LinkedIn message that follows is subject to the window.