> ## 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.
# Outcomes and verification
> How a guide's outcome closes — only on a fresh read from a connected source or your explicit confirmation — and the derived five-stage funnel on the Outcomes tab.
A guide exists to drive one outcome per customer, and Thread is strict about what "done" means. **The guide never verifies its own work.** An outcome closes exactly one of two ways:
* **The read.** The guide cites a fresh read from a connected source — your CRM, your project tracker, a named teammate's sign-off in a connected channel — taken in the same wake it closes in. The receipt (source, reference, value, read-at) is recorded on the close, and the outcome shows as **Done — verified** against your system of record.
* **Your confirmation.** When no read tool is attached, the guide cannot close anything on its own. It proposes done, pages you with its evidence, and waits for your one-tap **Confirm done** on the [watch page](/guides/watch-page). The outcome shows as **Done — confirmed by you**.
Neither the guide's word nor the customer's ever closes an outcome. A customer saying "all set!" is a claim — the guide's habit is to verify it at the source before celebrating, and the log records both the claim and the read.
## The close mode is never a surprise
Which arm a guide closes on is decided by one fact: whether at least one read tool is attached and connected. The guide page discloses it everywhere it matters:
* After publish, the guide's status line states it outright — "closes on fresh evidence — reads: HubSpot" or "closes by your tap".
* Publishing reads back your instructions' definition of done alongside the attached read tools, so the definition and the evidence bind together in front of you.
* Detaching the last connected read tool asks for an explicit confirmation that names the consequence — the close mode never flips silently.
## The five-stage funnel
The guide's **Outcomes** tab opens with a left-to-right pipeline of five stage cards. Every stage is a code-observed fact computed from the log — **derived, never authored**. Nobody edits a customer's stage, and the guide cannot narrate itself forward.
| Stage | The fact that lights it |
| ------------- | -------------------------------------------------- |
| **Enrolled** | Consent attested — the customer was enrolled |
| **Reached** | First text or email delivered |
| **Engaged** | First reply in from the customer |
| **Verifying** | First record read from a connected source |
| **Done** | Closed — verified by the read, or confirmed by you |
Each card shows a cumulative count (customers who reached at least that stage), with the stage-to-stage conversion percentage in the seams between cards. The terminal card keeps the two arms honest: a mixed cohort shows the split — for example, `16 verified · 2 confirmed by you` — because green is earned by the read, and taps are counted apart.
The funnel adapts to what the guide can actually observe:
* A guide with **no read tools attached shows four cards** — never a ghost Verifying rung — with a footnote: closes are confirmed by you.
* A rehearsed close from [rehearsal](/guides/rehearsal) renders as "Done — rehearsed", never as verified. Green is earned by a real read, even in practice.
A **Band** toggle renders the same numbers as a single tapered funnel silhouette — a different view of the same derivation, nothing stored.
## The roster
Below the funnel sits the roster: one row per customer being driven to the outcome, with the customer's name, current stage, and last activity. Rows that need your attention wear a **Needs you** chip — the only warm accent on the page, and effectively your call list. Paused instances are marked too.
The funnel and the roster are one surface:
The roster filters to exactly the customers that card counted — click **Verifying** to see everyone the guide is actively checking, or **Done** to see the closes.
Every row opens that customer's [watch page](/guides/watch-page), where the full log and the per-customer ladder live.
Clear the filter to return to the whole roster.
## Where the evidence lives
Every read the guide takes lands in the watch log as a monospace receipt — `source · reference · value · read-at` — and the close event cites the specific receipts that satisfied it. When you want to audit a close, the citation is one click away on the close's own log entry. The stages, the funnel, and the Done labels all derive from those recorded events, so the numbers you report from this tab are backed by receipts, not summaries.
Which evidence counts is yours to define. Your guide's instructions carry the "done sentence" — what done means and which source shows it — and the attached read tool is configured with the exact record and field it checks. See [building a guide](/guides/building-a-guide).
## Related pages
The certified per-customer log the funnel derives from.
The connected sources guides can read for verification.
The floor that makes self-verification impossible, not just discouraged.
The same stages, rolled up across bulk enrollments.