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

# When a switch between organisations does not take

> Changing which organisation you are acting in can fail. This is what you see when it does, what we already know about it without you telling us, and what that record contains.

If you belong to more than one organisation, the control at the top left of
every product screen is how you move between them. It names the organisation you
are acting in right now, and opening it lists the others.

A switch that works reloads the page and says so: **Now acting in Acme Payments
EU. You were in Acme Payments.** That sentence is the confirmation — the pages
either side of a switch look alike, and a reload on its own is not a statement.

This page is about the other outcome.

## What you see when it fails

You land on a page that says the switch did not take, and it names the
organisation you are **still** acting in. That is the part that matters: the
danger in a failed switch is not the failure, it is a screen that leaves you
unsure which organisation's data you are reading.

On that page you will find:

* the organisation you are still in, named;
* the organisation you were trying to reach, named, when it is one you belong to;
* what went wrong, in one sentence;
* every organisation you belong to, each one a button that tries the switch
  again.

Nothing you were doing is lost, because nothing changed. A failed switch leaves
you exactly where you were.

## We already know

The failure is recorded on our side before that page is drawn. You do not have
to report it, and we would rather hear about it from our own records than from
you.

That is a change of position rather than a courtesy. Until this shipped, a
switch that failed was known only to the person it failed on: they were told,
on their own screen, and that was the whole of it. Nobody could say whether it
happened once a month or forty times a day.

If the page tells you we could not record it — which happens when the network
that failed the switch also fails the report — then that is the one case where
telling us adds something. The time it happened, to the minute, is enough.

## What the record contains

Identifiers and outcomes. Specifically:

| Recorded                                   | Why                                        |
| ------------------------------------------ | ------------------------------------------ |
| The organisation you were acting in        | Which account this happened on             |
| The organisation you were switching into   | Which pair of accounts it happened between |
| Your user id                               | Whether it is one person or many           |
| When it happened                           | Whether it clusters around anything        |
| Which of a fixed list of things went wrong | Whether it is one fault or several         |

And that is the whole of it.

**It contains nothing you typed.** Not the names you gave your organisations,
not what was on the page you were on, not the address you were reading, not a
message from the identity provider that might quote any of them. What went wrong
is recorded as one of a short list of outcomes this product knows about, and a
short machine-readable code from the identity provider — never as free text.

This is a deliberate limit and it costs us something: there are failures we will
be able to count and locate but not fully explain from the record alone. We
would rather investigate a narrow record than hold a wide one.

## What happens next, and what does not

Somebody reads these. A failure carries the organisation it happened on, so it
can be looked into rather than counted and forgotten.

**Nothing about this wakes anybody up.** No threshold in this product pages a
person on its own — not this one, and not any other. If failed switches climb to
a rate that deserves somebody's night, the system says so and a person decides.
That is the standing rule here, and it is why you will never be told that an
automated alert has been raised on your behalf.

## If it keeps happening

Try once more from the page you land on. If it fails again:

1. Sign out and sign back in. A switch depends on your session, and a fresh one
   fixes most of what can go wrong with it.
2. If it still fails, write to us with the time. We will have the record; your
   message tells us which one of them is yours.