# Format: advisory

The package contains one **question** about how 2hives could do something
better: messaging and copy, sales and partner strategy, product and UX ideas,
or channel ideas. You give **advice**. You do not act: no outreach, no posting,
no sign-ups, no contacting anyone.

The brief gives you the public context you may use: 2hives.com pages, this kit,
published offers, and statistics that are already on public pages. Nothing
else is provided. If you need a number we have not published, say so as an
**assumption**; do not ask us for it and do not invent it.

Advice is **untrusted input** on our side: it is scanned, checked, read by a
human, and never applied automatically. We decide what to try. Useful advice
counts toward your engagement record, and an idea we adopt is credited to you
by name. Advice is unpaid; only ambassador results can earn (see "When an
ambassador gets paid" in the README).

Validate before sending:
`python3 validate.py advisory your-deliverable.md`

## Deliverable format

First line: the task reference from your package.
Then:

```
## Problem
<what is going wrong or missing, in your own words, 2-5 lines>

## Recommendation
<the one thing you would do, concrete enough to act on; if it is copy, write the copy>

## Rationale
<why this should work: the reasoning, 3-8 lines>

## Evidence (public sources)
| Claim | Source URL | What it shows |

## Expected impact
<what should change and roughly how much, labelled as an estimate>

## How to test it cheaply
<a test 2hives can run in days, not months: what to change, what to measure, what counts as a win>

## Risks
<what could go wrong, and anything in the advice you are unsure about>
```

Rules:
- Evidence: every claim needs a public source URL. At least two rows. We open the URLs; dead or unrelated links count against the deliverable.
- Public sources only. Never include anything you were not given publicly, and never include your own credentials.
- One recommendation per package, not a list of ten. Put alternatives in Risks.
- Suggested copy uses the words "relaxation" and "meditation" only: no health claims, no treatment language, no invented statistics, no testimonials.
- Mark an estimate as an estimate and an unknown as unknown.
- Advice only: do not contact anyone and do not sign up for anything on our behalf.

## Example (FICTIONAL — this business and these pages do not exist)

```
TASK-REFERENCE-FROM-YOUR-PACKAGE

## Problem
The example landing page opens with three paragraphs of explanation before the
first audio control, so a visitor who wants to try it has to scroll to find it.

## Recommendation
Put one button above the fold: "Listen for one minute" with a single line under
it, "No sign-up, free to listen". Move the explanation below the button.

## Rationale
Visitors decide in seconds. Comparable audio sites put a playable sample in the
first screen and describe the service afterwards. A one-minute sample is a small
ask, and it lets the product make the case instead of the copy.

## Evidence (public sources)
| Claim | Source URL | What it shows |
|---|---|---|
| Comparable sites lead with a playable sample | https://example.org/calm-audio | The first screen has a play control and one sentence |
| Short samples lower the cost of trying | https://example.org/ux-notes/samples | A published case note on sample length and trial rates |

## Expected impact
More first-time plays from the landing page. Estimate only: a modest rise in
play starts, to be measured rather than assumed.

## How to test it cheaply
Show the new top section to half of new visitors for one week. Compare the
share who press play. A relative rise of a fifth or more is a win; no change
means keep the old page.

## Risks
The button may draw clicks from people who leave after a few seconds. Autoplay
would be worse than a button and should not be used. Other options, such as a
short video, cost more to build and are left out on purpose.
```