A broadcast sends one message to a chosen audience. It is composed, checked, and
then run — the three steps are deliberately separate, because a send cannot be
recalled.
Composing
Name is for you, not the recipients. It is how the broadcast appears in the
list afterwards, so name it by what it is rather than when you sent it.
Message has a tab per language the tenant has enabled. Each recipient gets
the message in their own language, falling back to the assistant’s default
when the contact’s language is unknown or has no translation. Basic HTML is
supported.
Filling only one language is normal and safe — everyone falls back to it. What
is not safe is filling several and leaving one half-written: a contact whose
language matches the unfinished tab gets the unfinished text, not the complete
one from another tab.
Choosing the audience
Three ways, in increasing precision:
Reach
As the audience is chosen, Reach shows how many contacts would actually
receive the message.
This is the number to read before sending, and it is not the same as the size of
the audience. A contact is counted only if the assistant can still reach them —
a segment of five thousand can have a reach far below that if most of those
people have blocked the bot or never opened a channel.
A reach of zero means the send would do nothing. That is worth resolving before
running it rather than after: the usual causes are an audience that matches
nobody, or contacts who exist in the database but have no reachable channel.
Running it
A broadcast is editable while it is still a draft. Once a run has started or
finished, it is a record rather than a document — the message that went out
cannot be revised after the fact.
Large sends are split into batches and paced, so they cannot delay replies to
people in live conversations. How aggressively is a tenant-wide setting; see
Settings.