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

# Rehearsal

> Play the customer yourself and run the whole arc — messages, waits, nudges, even a live call — before the guide ever reaches a real person.

Every guide can be rehearsed against you before it touches a customer. Press **Try it live** on the guide page and a drawer opens where **you play the customer**: the same instructions, the same wake loop, the same log — with sends shadowed, the pager shadowed, and a clock you can fast-forward. A three-week arc of nudges, decided silence, and the backstop paging plays out in minutes.

Rehearsal works on drafts. You don't publish first — trying it before reality is the point. The only requirements are the same authoring basics publish asks for: a way to reach people (a texting number, or an email sender if email is on) and written instructions, because the rehearsal runs them verbatim.

## What's real and what's shadowed

A rehearsal is deliberately honest about which is which:

* **Sends are shadowed.** Nothing reaches a phone or an inbox — the guide's texts and emails render in the drawer's log exactly as they would on the watch page.
* **Paging is shadowed.** When the guide would escalate, the log shows *would have paged you* instead of firing a real page.
* **Reads are real.** Attached tools run genuine reads by default — if you bind a real record from your CRM, the guide reads it live. Reads are safe; that's the point of rehearsing them.
* **STOP is real physics.** Reply STOP as the customer and the suppression lands exactly as it would in production — rehearsed for real.
* **Rehearsals never touch the roster.** Sandbox conversations never appear on the Outcomes tab, never count in the funnel, and can never text, email, or call a real customer. Real conversations, in turn, have no time machine — the virtual clock exists only in rehearsal, by construction.

## Starting a rehearsal

The start form mirrors enrollment, because that's what it stands in for:

* **Who do you play?** — name the customer you're acting as, and optionally their company.
* **Their team** (optional) — the account owner and AE the guide can loop in. Next to the AE email there's a one-click **use my email**, which stands you in as the AE — so when you rehearse the loop-in, the escalation goes to the person running the rehearsal.
* **Their record** (optional) — when HubSpot is attached and connected, pick a real record for the guide to read against, seeded with the one the tool was tested with. Leave it unbound and the rehearsal honestly closes by your tap.

Press start, and the guide opens the conversation — wake #1, exactly as a real enrollment fires it. Reply in the composer as the customer; when both texting and email are armed, a toggle picks which channel you're answering on.

## The time machine

Guides live on timers — "give them until Friday, then nudge" — and the rehearsal compresses the waiting. Each rehearsal carries its own virtual clock, shown as a chip in the drawer's header with the days skipped (`+3d`). Skipping moves everything together: every rendered timestamp and the guide's own sense of "now," so a Friday wake never fires into a context that still says Monday.

* **Skip to next wake** — jump just past the guide's next scheduled wake and fire it. When the guide is waiting on you and has no timer armed, the button becomes **Wake it now**.
* **+1 day / +7 days** — move time forward; the drawer reports how many timer wakes fired.
* **Start over** — abandon this rehearsal and set up a fresh one.

## Pretend evidence

The close rule doesn't bend in rehearsal — a guide still can't close without evidence. So each attached read tool gets one **Pretend it proves it** button: on its next read, that tool reports that the source confirms the outcome *as your instructions define it*, so you can see what "true" means for this guide and watch the close fire. **Back to real reads** clears it.

Pretend reads are branded **SIMULATED** in the log, and a close citing one renders **rehearsed** — never verified. Green is earned by a real read, even in rehearsal.

## Rehearsing a call, in your browser

If the guide has a voice agent attached, rehearse the whole call flow without a phone:

<Steps>
  <Step title="The guide asks for a call">
    When the conversation warrants a call, the guide asks — showing the full number it would dial. In rehearsal nothing can ring; live, this same ask would page you if the guide has **Ask me before dialing** on, and otherwise would go straight to the dial.
  </Step>

  <Step title="The call is armed">
    With approvals on, **Approve the call (rehearsal)** is the real approval door — live, that one tap dials; here it arms the call for you to take. With approvals off, it arms as soon as the guide asks.
  </Step>

  <Step title="Take the call">
    **Take the call — as the customer** opens a live web call in your browser: a real voice session, briefed on the conversation so far, with you on the customer's end of the line.
  </Step>

  <Step title="The guide picks the thread up">
    After you hang up, the call's transcript lands in the log within moments and wakes the guide — the same door a production call returns through — and the conversation continues from what was said.
  </Step>
</Steps>

## Reading the rehearsal

The drawer's feed is the real [watch page](/guides/watch-page) log: the same wake labels, message bubbles, read receipts, and close citations you'll supervise in production. If the guide proposes done, the **Confirm done** tap is right there — the fallback arm, rehearsable end to end.

<Tip>
  Rehearse the failure paths, not just the happy one. Go silent and skip a week to watch the nudges and the backstop page. Reply STOP. Claim you're done before you are, and watch the guide check the record instead of taking your word for it.
</Tip>

## Common questions

<AccordionGroup>
  <Accordion title="Why is Try it live disabled?">
    The button's tooltip names the missing piece: turn on a channel, pick a texting number or email sender, or write the instructions — the rehearsal runs them verbatim, so an empty box has nothing to rehearse.
  </Accordion>

  <Accordion title="I skipped time and nothing fired.">
    The drawer says so honestly: *time moved — nothing fell due.* Usually the guide has paged you and is waiting with its own timer off — text it as the customer, or use **Wake it now**. Skipping time is also never the way past an armed call: the one thing waiting there is a microphone.
  </Accordion>

  <Accordion title="Can a rehearsal accidentally reach a real customer?">
    No. Sends and pages are shadowed at the tool level, rehearsal conversations live on sandbox instances that never enter the Outcomes roster, and the drawer's only live audio is the web call you take yourself, in your own browser.
  </Accordion>

  <Accordion title="Does closing the drawer lose the rehearsal?">
    No — reopening **Try it live** picks up the newest rehearsal for that guide where it left off. **Start over** is the explicit way to abandon one and begin fresh.
  </Accordion>

  <Accordion title="The call ended but the log still shows it armed.">
    The transcript is delivered a few seconds after hangup, and the drawer waits for it briefly. If it never arrives, the drawer says so and the call stays armed — take it again, or wake the guide to continue without it.
  </Accordion>
</AccordionGroup>

When the rehearsal reads the way you want — the opener, the nudge cadence, the honesty when you claim done early — [publish the guide](/guides/building-a-guide) and [enroll your first customer](/guides/enrolling-customers).