# Reels Insights readout

Fill this in after a Reel has been live long enough to read (24 hours
minimum for a trial, 72 hours minimum for a public Reel). The point is
to map each metric to the lever it points to, so the next Reel changes
in the right direction.

```
REEL TITLE / TOPIC:
PUBLISHED:                  (date / trial or public)
HOURS LIVE:
ACCOUNT BASELINE NOTES:     (your usual numbers for this kind of Reel)
```

## Numbers

```
Reach:
Total views:
Skip rate:                  (your baseline: __ )
Average watch time:
Retention curve cliff:      (timestamp where drop-off happens)
Share rate (sends/reach):   (your baseline: __ )
Save rate:                  (your baseline: __ )
Likes per reach:            (your baseline: __ )
Comments:
Followers from this post:   (your baseline: __ )
Post-watch: profile visits:
Post-watch: auto-play into next Reel:
```

## Read the numbers (check the one that fits)

- [ ] **High skip rate, low total views.** Hook failed the three-second
      cliff. Fix the opening three seconds before doing anything else.
      (See `references/reels-hook.md`.)
- [ ] **Hook held, retention cliff in the middle.** Body failed. Add a
      re-hook beat or compress the build-up.
- [ ] **High reach and watch time, near-zero follows.** Topic legibility
      failed. The Reel was watchable but didn't make the topic clear.
      Fix the on-screen text and caption keyword.
      (See `references/caption-and-search.md`.)
- [ ] **Watch time fine, sends below baseline.** CTA failed. Recut the
      end with a named-persona send-prompt.
      (See `references/sends-playbook.md`.)
- [ ] **Multiple recent Reels with sudden low reach.** Possible
      originality cap. Audit the last 30 days.
      (See `references/originality-and-audio.md`.)
- [ ] **Audio missing on test view.** Audio was stripped at publish.
      Recut with safer audio.
- [ ] **All metrics healthy, the Reel just hit.** Note the format. Plan
      a follow-up in the same shape.
- [ ] **All metrics flat with no clear failure.** Likely variance or
      topic-fit mismatch. Don't change the format on one outlier.

## Action for the next Reel

```
EARLIEST FAILURE FOUND:
ONE CONCRETE FIX:           (recut opening / new caption / re-trial /
                             fresh Reel built around named-persona payoff)
WHAT TO KEEP:               (what worked, don't change it)
RE-TRIAL OR FRESH?:
```

## Don't read these as verdicts

- View count alone. Without reach and share context, it's vanity.
- Likes alone. Cheapest action, weakest distribution signal.
- Day-one numbers on a public Reel. Reels can keep distributing for
  days if share rate stays healthy.

# Reels hook checklist

Run any candidate Reels hook through this. A strong hook clears most of
these. If it fails any of the first four, rewrite it before going further.
The three-second cliff is unforgiving on Reels.

## Pass / fail criteria

- [ ] **Lands in about 2 seconds when said out loud.** No "hey guys," no
      logo intro, no slow zoom.
- [ ] **First spoken word is the most interesting word in the line.** Cut
      anything before it.
- [ ] **Works on mute.** Three to five bold words on the first frame
      carry the hook for sound-off viewers.
- [ ] **Names a concrete viewer, number, or stake.** No vague benefits
      ("level up your skincare"), no abstract promises.
- [ ] **Opens a loop the body actually closes**, or names a stake the
      payoff actually addresses. No orphan clickbait.
- [ ] **Signals a shareable payoff.** The viewer can already think
      "I know someone who needs to see this" by the third second.
- [ ] **First frame composition holds the hook.** Face, reveal, motion,
      or sharp contrast. Not a flat static product shot.
- [ ] **No third-party app UI in the frame.** No TikTok timer, no CapCut
      overlay, no screen-recording chrome.
- [ ] **On-screen text sits clear of Instagram's UI.** Not in the bottom
      fifth where captions sit, not in the top fifth where account info
      sits.
- [ ] **Topic is legible from the hook alone.** A stranger can name what
      this Reel is about in three seconds, without sound.

## Fast upgrades when a hook is weak

- Add a number or a specific time / dollar figure.
- Name the exact viewer ("if you [specific situation]...").
- Cut everything before the first interesting word.
- Swap a vague benefit for a concrete result.
- Turn a statement into a withheld reveal ("I won't say the brand, but...").
- Pair the spoken hook with a contrast on screen (text says one thing,
  visual shows another).
- Rewrite the on-screen text as a three-to-five word headline, not a
  sentence.

## Reels-specific failures to look for

- Hook works with sound, fails on mute. (Sound-off viewers scroll.)
- Hook promises something the body can't deliver. (Tanks sends, the
  metric that actually buys cold reach.)
- Hook is muddy across topics. (Topic legibility is what converts views
  into follows on Reels.)

# Reels script template

Fill this in for a single Reel. Aim for 15 to 30 seconds unless the brief
says otherwise. Total seconds watched matters, so a longer Reel with strong
retention can beat a shorter one with weak retention, but the cold-reach
sweet spot stays short.

Keep the audio column tight. People talk faster than they read, and the
on-screen text should reinforce the spoken line, not duplicate it.

```
WORKING TITLE:
TOPIC / KEYWORD PHRASE:    (the exact words a stranger would search)
NAMED VIEWER:              (the one specific person this Reel is for)
GOAL:                      (cold reach / saves / follows / sends)
TARGET LENGTH:             (sec)
ACCOUNT TYPE:              (Creator / Business)
TRIAL OR PUBLISH:          (see assets/trial-reels-decision.md)

──────────────────────────────────────────────────────────────────────
TIME  | SPOKEN / AUDIO                  | ON-SCREEN TEXT      | VISUAL / SHOT
──────────────────────────────────────────────────────────────────────
0-3s  | HOOK (concrete, named viewer,   | hook headline,      | face / reveal /
      |  payoff promise)                | 3-5 bold words      | in media res
      |                                 |                     |
3-Xs  | ESCALATION beat 1               | key phrase          | reset shot/angle
      |  (raise tension, add specifics) |                     |
      | ESCALATION beat 2               |                     | reset shot/angle
      |   use "but / therefore"         |                     |
      |                                 |                     |
X-Ys  | PAYOFF (deliver the promise,    | the takeaway,       | show the result
      |  named container if possible)   | nameable phrase     |
      |                                 |                     |
end   | SEND-PROMPT CTA                 | CTA text, own       | hold / loop
      | "send this to the friend who…"  | beat                | to start
──────────────────────────────────────────────────────────────────────

CAPTION:
  Line 1: [keyword phrase as headline]
  Body:   [1-2 lines expanding the hook, plain language]
  CTA:    [named-persona send-prompt, restate the one from the Reel]
  Tags:   [3-5 highly relevant hashtags]

AUDIO:
  Source:                     (original / Sound Collection / trending)
  Trending window check:      (upward arrow, < ~30K uses)  [ ]
  Audio attached on test view: [ ]

PROMISE -> PAYOFF CHECK:
  Does the body deliver what the hook promised?  [ ]
  Is the payoff useful to one specific person?   [ ]
  Could that person be named in the CTA?         [ ]

ORIGINALITY PASS (see references/originality-and-audio.md):
  No TikTok watermark visible:    [ ]
  No CapCut watermark visible:    [ ]
  No third-party app UI on screen:[ ]
  Substantial edit (your cuts,    [ ]
   your voice, your text):
```

Notes:
- If the hook needs a sentence of setup, the setup is probably the real
  hook. Start there.
- Cut the first and last lines last. They tend to be the most over-written.
- On Reels, the grid crop is 3:4. If the Reel is going to the grid, frame
  the hook subject in the upper two-thirds so the crop keeps it.

# Trial Reels decision sheet

One page. Fill it in per Reel. Default to a trial if cold reach is the
goal and the account is eligible.

```
REEL WORKING TITLE:
PRIMARY GOAL:        (cold reach / follower nurture / time-sensitive moment)
ACCOUNT FOLLOWERS:   (under 1K = no trial available)
ACCOUNT IS PUBLIC:   [ ] yes  [ ] no
ACCOUNT TYPE:        (Creator / Business)
```

## Eligibility gate

- [ ] Account has at least about 1,000 followers.
- [ ] Account is public.

If either is unchecked, you can't trial. Skip to publish, focus on the
0 to 1K slog. (See `references/trial-reels.md` for the mechanic.)

## Goal gate

- [ ] The goal of this Reel is cold reach (non-followers).
- [ ] You are uncertain whether the hook lands.
- [ ] You're not certain the topic fits cold audiences.
- [ ] You want a clean read before committing the Reel to the grid.

If all four are checked, run a trial. The flop is free, the win is
confirmed.

If you checked one or fewer, the Reel may be a publish, not a trial.

## Publish-direct (skip the trial)

Publish directly without trialling when one of these is true:

- [ ] Primarily for existing followers (BTS, community moment, thank-you).
- [ ] Time-sensitive in a way that needs the grid right now (product
      drop, calendar moment, live event).
- [ ] Part of a series your followers expect to see.

Followers don't see trials, so a trial wastes the audience that this
Reel was actually for.

## After the trial (24 to 72 hour read)

Compare against your own prior trials, not against follower-feed posts.

- [ ] **Numbers above your trial baseline.** Share to everyone now.
      Don't wait for auto-share. Early momentum matters.
- [ ] **Numbers okay but not obvious.** Let Instagram auto-share at
      72 hours. Passive yes, lets the platform's own judgement decide.
- [ ] **Numbers below baseline, hook held.** Topic likely didn't fit
      cold audiences. Don't publish. Save the footage, recut the topic
      framing, re-trial.
- [ ] **Numbers below baseline, hook failed (high skip rate).** Don't
      publish. Recut the opening three seconds and re-trial.

## Quick decisions

- Cold reach goal, eligible, uncertain about hook -> trial.
- Time-sensitive grid moment -> publish.
- Follower-only nurture -> publish (or skip Reels entirely).
- Under 1K followers -> publish, you can't trial.
- Trial converted strongly -> share to everyone immediately, don't wait.
- Trial flopped -> don't publish, recut the opening or the topic frame,
  re-trial.

# Caption and search

Hashtags don't move reach on Reels. Mosseri said so directly. Captions and
on-screen text now do the work hashtags used to: they tell Instagram what
the Reel is about so it can route the Reel to the right Reels-tab viewers,
and they get the Reel indexed in in-app search and on Google.

The caption is also the most underused CTA surface on the platform. It's the
one place where you can write a real send-prompt without burning seconds of
the Reel.

## Captions as SEO

Instagram indexes captions, alt text, and on-screen text for in-app search.
Public posts are also indexed by Google and Bing, which means your Reels can
show up for keyword searches outside Instagram entirely. That changes how a
caption should be written.

Practical rules:

- **First line is the keyword phrase.** Lead with the exact words a person
  would type if they were searching for this Reel. "Niacinamide for oily
  skin" beats "the one product changing my mornings."
- **Write in plain search-friendly language.** Avoid metaphor and inside
  jokes in the first line. Save the voice for later in the caption.
- **One topic per caption.** A caption that hedges across three topics
  trains the algorithm that the account is about all three weakly. Pick the
  one and own it.
- **Use the keyword phrase verbatim once or twice, not five times.** Keyword
  stuffing reads as low quality and is easy to spot.
- **150 to 300 characters is the working sweet spot.** Long enough to load
  the topic, short enough that viewers actually read it. Reels captions that
  are walls of text get ignored.

## On-screen text as a second search signal

Instagram reads on-screen text and adds it to the topic understanding of the
Reel. A Reel with the keyword phrase on the first frame is twice tagged: once
for retention (sound-off viewers see what it's about) and once for indexing
(the platform knows what the Reel is about).

For dual-duty on-screen text:

- The first-frame text should be the keyword phrase in a headline form, not
  a complete sentence. "Breakouts after every new serum?" beats "Today I'm
  going to talk about why you break out after every new serum."
- Use the same vocabulary on screen as in the caption. Consistency
  reinforces the topic signal.
- Don't try to caption the whole voiceover on screen. Punctuation, not
  transcription. (See `reels-hook.md` for on-screen text timing.)

## Hashtags: three to five, no more

The old advice was 30 hashtags stuffed in the first comment. That advice is
dead. Mosseri stated directly that hashtags don't improve reach. They still
act as labels for in-app search and topic categorisation. So use them, but
don't expect them to do anything they don't do.

Working rules:

- **Three to five hashtags, maximum.** More than five reads as engagement
  bait and signals low-quality content.
- **Highly relevant only.** A hashtag that doesn't accurately describe the
  Reel is noise. It also confuses the topic signal.
- **Mix one broad and two to four niche.** One broad (#skincare) for general
  surface area, two to four niche (#niacinamideroutine, #oilyskincare) for
  specific match.
- **In the caption, not in a first comment.** First-comment hashtags are a
  relic of the old algorithm. Put them at the end of the caption.
- **Don't reuse the exact same hashtag set on every post.** Instagram reads
  identical hashtag stacks as automated behaviour.

## Caption structure for Reels

A working caption shape for cold-reach Reels:

```
[Keyword phrase as headline, sentence-case]

[One to two lines that expand the hook and reinforce the topic]

[Send-prompt CTA naming the recipient]

[3 to 5 relevant hashtags]
```

Example skeleton (not a literal template):

```
Niacinamide for oily skin: what the routine actually looks like.

The trick most people miss is layering order. Get it wrong and the
breakouts come back faster than the serum can fix them.

Send this to the friend who buys every new serum.

#skincare #niacinamide #oilyskincare #skincareroutine
```

The caption restates the topic in search-friendly language, expands the
hook into context, drives a send to a named persona, and labels the Reel
with relevant hashtags. It's a complete unit that works with the Reel, not
a throwaway under it.

## Your Algorithm and topic legibility

Instagram now exposes the topic labels it has inferred about a user, and
lets them edit those labels. That cuts both ways for creators.

On the inbound side, your Reel is competing for placement against other
Reels for the same topic labels. If Instagram can't confidently assign a
topic to your Reel, it doesn't know who to show it to and reach suffers.

On the outbound side, the topics you build through your captions, on-screen
text, and engaged audience determine which topic labels your account gets
associated with. Inconsistent topic positioning across Reels makes the
account fuzzy and weakens recommendation.

Practical implication: be ruthlessly consistent on topic across multiple
Reels. Don't post a skincare Reel, a finance Reel, and a travel Reel in the
same week and expect any of them to compound. The clearer your account's
topic signature, the cleaner the recommendation surface for every Reel.

## Profile and grid placement

When you publish a Reel to the grid (an opt-in checkbox), it appears in the
profile grid at a 3:4 aspect ratio. A composition framed for 9:16 video
often crops awkwardly on the grid. If the Reel matters for your profile
brand, frame the hook subject in the upper two-thirds so the grid crop
keeps it.

You can pin Reels to the profile grid (maximum three pins across all
posts), but you cannot reorder or pin within the Reels tab. The Reels tab
itself is strictly reverse-chronological.

## Anti-patterns

- **Hashtag-stuffed captions** ("#fyp #foryou #explore" and 25 more). Reads
  as low-effort, possibly bait, and dilutes the topic signal.
- **Throwaway captions** ("link in bio," a single emoji, "thoughts?"). The
  caption is the SEO surface and the CTA surface. Use it.
- **Captions that contradict the Reel.** A caption that promises one thing
  and a Reel that delivers another confuses topic and tanks trust.
- **Different keyword phrases on screen and in caption.** Pick one and use
  it in both places.
- **Captions written for followers when the Reel is for cold reach.** Most
  Reels viewers don't know you. Write the caption for them.

## Quick checklist before publishing

- First caption line is the keyword phrase a stranger would search.
- On-screen text on the first frame matches the caption's keyword phrase.
- Caption is 150 to 300 characters with a clear send-prompt CTA.
- Three to five highly relevant hashtags, in the caption, at the end.
- Topic of this Reel is consistent with what the account usually posts.
- If the Reel is grid-bound, the hook subject is composed for the 3:4 crop.

# Diagnose a flop

When a Reel underperforms, run the diagnostic in order. Most flops fail at
exactly one of these levers and the earliest failure is the only one worth
fixing first. Working out of order wastes the next Reel.

The order is: originality, hook, body, topic legibility, send signal,
audio, caption. Stop at the first lever that explains the result.

## Before you start

Ask for or pull these numbers from Reels Insights:

- Total reach and total views.
- Skip rate.
- Share rate (or sends per reach).
- Average watch time and retention curve.
- Followers from this post.
- Save rate.
- Any non-standard signal (sudden distribution stop, no notifications,
  etc.).

Compare against the account's own baseline for prior Reels, not against
absolute numbers. A 5% share rate is great for one account and weak for
another.

## Step 1: Originality flag

This is the silent killer. Check first because no other fix matters if
the account is in a recovery window.

Signs the account may be capped:

- Multiple recent Reels with sudden drops in non-follower reach, with no
  hook problem to explain it.
- A run of recently posted reposted content, or Reels with visible TikTok
  or CapCut watermarks.
- Significantly lower reach than usual across the last several posts, not
  just this one.

If the signs are there, the fix is not for this Reel. The fix is to stop
posting unoriginal content and consistently publish original Reels for at
least 30 days. See `originality-and-audio.md` for the rolling 30-day
window mechanic. If this is the cause, no in-Reel edit will save the
current post.

If only this one Reel underperformed and the account history looks
healthy, originality probably isn't the cause. Move to step 2.

## Step 2: Hook (the three-second cliff)

Pull the skip rate and the first three seconds of the retention curve.

- Skip rate notably above the account baseline, retention cliff in the
  first three seconds: hook failed. This is the most common cause.
- Skip rate normal, retention held past three seconds: hook worked. Move
  on.

If the hook failed, recut the opening three seconds before doing anything
else. See `reels-hook.md` for what the opening needs to do, and
`assets/reels-hook-checklist.md` for the pass/fail check.

Common hook failures to look for:

- "Hey guys" or any intro before the first interesting word.
- Slow visual: black frame, logo, slow zoom.
- Vague benefit: no concrete noun or specific viewer named.
- Sound-only hook with no on-screen text. Sound-off viewers scrolled.
- Cover thumbnail (in Reels tab) doesn't match the hook content.

## Step 3: Body (mid-Reel bleed)

Hook held but average watch time is low and the retention curve flatlines
in the middle. Pull the retention curve and find the drop.

- Cliff at 5 to 10 seconds: the body didn't deliver on the hook's promise
  fast enough. Front-load more of the payoff.
- Gradual flatline through the middle: no escalation. Add a re-hook beat
  ("but here's the part nobody talks about") right where attention sags.
- Cliff just before the payoff: build-up reads as filler. Compress or cut.

See the retention discussion in the `viral-short-form` skill's
`references/retention.md` for the failure-mode framework. Reels-specific
note: the body sag matters more on Reels than on TikTok because the
auto-play continuation into your next Reel depends on a clean finish.

## Step 4: Topic legibility (views without follows)

Healthy reach and watch time, near-zero follows. The Reel was watchable
but didn't make the topic clear enough for a stranger to want to follow
the account.

Check:

- On-screen text on the first frame: does it name the topic in plain
  words a stranger would search for?
- Caption first line: is it the keyword phrase, or a clever line that
  obscures the topic?
- Profile bio: does it match the topic of this Reel? A viewer who lands
  on the bio after the Reel needs to see "more of this."
- Account topic consistency: does this Reel match what the account
  usually posts? An off-topic Reel from a topically-focused account
  attracts viewers who won't stay.

The fix is in the caption and on-screen text for this Reel, and in topic
discipline going forward. See `caption-and-search.md`.

## Step 5: Send signal (no shares)

Watch time and saves are fine, sends are below baseline. The payoff is
good, but the Reel didn't earn the favour of a DM share.

Check the CTA:

- Was it a generic "follow for more" or "like if you agree"? These
  suppress distribution by reading as engagement bait, and they spend
  the closing seconds on a low-weight action.
- Was the CTA a named-persona send-prompt ("send this to the friend
  who..."), or vague ("share with a friend")? Specificity is the lever.

Check the payoff:

- Is it useful to a recognisable specific person, or to a vague "anyone
  interested in this"?
- Is there a memorable container (a named list, a clear takeaway)
  someone could repeat in a DM thread?

The fix is a recut CTA at the end and possibly a name-anchored payoff
restructure. See `sends-playbook.md`.

## Step 6: Audio strip

A Reel can play muted if licensing failed silently, especially on
Business accounts using trending songs outside the commercial pool.
Check the Reel on a test non-creator account. If the audio is stripped,
that explains the underperformance regardless of everything else.

Fix: recut with a Sound Collection track if you're on Business, or
switch to a Creator account for trending audio access. See
`originality-and-audio.md`.

## Step 7: Caption SEO

If reach was low across the board (and the hook was fine), the topic
signal may have been too weak for Instagram to know who to show this
Reel to.

Check:

- Caption first line is the keyword phrase, not a clever line.
- Caption is 150 to 300 characters, not a single emoji or a wall of
  text.
- Three to five highly relevant hashtags, not a stack of 20.
- Topic vocabulary matches between caption and on-screen text.

Fix in the caption for next time. You can't usually fix a caption
retroactively in a way that recovers distribution.

## Reading the result

After walking the steps, name the earliest failure first. Be specific:
"the hook failed the three-second cliff, skip rate was 62% versus your
baseline of 38%, the rest of the Reel was actually fine." Vague verdicts
("the hook needs work") don't help.

If multiple steps failed, fix the earliest one first. Fixing the CTA on
a Reel with a broken hook is pointless because the hook is what's
capping reach.

If the account looks healthy at every step and the Reel still
underperformed, the most likely answers are topic fit (Instagram didn't
have a confident audience match) or simple variance. Single-Reel
variance is real. Don't change the format on the strength of one outlier
in either direction.

## Recommend the next Reel as well as the fix

End the diagnostic with one concrete next action: a recut of the
opening three seconds, a new caption draft, a re-trial of the same
footage with a different first frame, or a fresh Reel built around a
named-persona payoff. The diagnostic is only useful if it changes what
the user does next.

# Edits app

Edits is Meta's free standalone video editor for Reels. It's the in-house
alternative to CapCut and the safer default for anyone who's been getting
CapCut watermarks into their exports. It is mobile-only at present and
publishes directly into Reels with deeper Insights integration than a
generic third-party editor.

It is not a hook generator, a viral-script tool, or a substitute for the
craft work in the rest of this skill. It's an editor. The reason it gets
its own reference here is that two specific Edits features actually move
distribution: no watermark on export, and tighter Insights wiring.

## When Edits helps

- **You've been getting CapCut watermarks through.** This is the most
  common case. Edits exports cleanly with no third-party watermark, which
  removes one of the biggest silent reach caps on Reels (see
  `originality-and-audio.md` for the watermark penalty).
- **You want template workflows tied to trending audio.** Edits surfaces
  templates that pair pre-cut beats with currently trending sounds. If the
  goal is to ride a sound trend, the in-app template flow saves time and
  keeps the audio attachment clean through publish.
- **You want Insights tied tighter to the edit decisions.** Edits passes
  metadata through to Reels Insights more cleanly than third-party
  exports. The retention curve and post-watch behaviour read against the
  cuts you made.
- **You're optimising for one platform.** Edits is built for Reels first.
  If you're not also cross-posting to TikTok or YouTube Shorts from the
  same export, the in-app flow is faster.

## What Edits doesn't fix

- **A weak hook.** Edits gives you more control over cuts and captions, but
  it does not rewrite a hook that opens on "hi guys." Hook work happens
  before you open any editor. (See `reels-hook.md`.)
- **A muddy topic.** Edits can put on-screen text on the first frame, but
  it can't make you choose the right keyword phrase. Caption and on-screen
  text decisions belong in the draft, not in the editor. (See
  `caption-and-search.md`.)
- **A non-shareable payoff.** Edits is not a content strategy. A Reel that
  earns no sends because the payoff isn't useful to a specific person
  won't be saved by a smoother edit. (See `sends-playbook.md`.)
- **Originality category violations.** A Reel that is mostly someone
  else's footage is still unoriginal in Edits. The watermark goes away
  but the substantial-edit bar still applies (see
  `originality-and-audio.md`).
- **Cross-platform optimisation.** Edits exports a Reels-shaped file. If
  you're also publishing to TikTok or Shorts, you may want to edit
  somewhere that gives you cleaner control over multiple aspect ratios.

## Practical workflow notes

- Edits is mobile-only at present. There is no desktop version. If your
  craft workflow lives on a laptop, Edits is a downgrade for editing
  itself. Use it for the export step or for the trending-audio template
  flow.
- The template library leans toward currently trending audio. Templates
  expire as audio trends saturate (see `originality-and-audio.md` on the
  trending audio window).
- Edits supports direct publish to Reels with caption, on-screen text, and
  the trial toggle in the same flow. You don't have to re-export and
  re-upload.
- Captions written in Edits feed straight into the Reel. Make sure the
  first line of the Edits caption is the keyword phrase, same as you
  would in the Instagram caption flow.

## When CapCut or another editor is still the right call

- You're already cross-posting and your CapCut workflow exports to
  TikTok-shape and Reels-shape from the same project, and you've fixed
  the watermark export setting. Re-tooling around Edits just for Reels
  may not be worth the friction.
- You're editing on desktop. CapCut and other desktop editors stay better
  than going mobile.
- The Edits template you want doesn't exist. Templates are a small slice
  of all possible Reel formats.

## Anti-patterns

- **Treating Edits as a virality tool.** It's an editor. The patterns in
  the rest of this skill still do the work.
- **Using Edits templates blindly.** A template tied to a current trending
  audio works for the first wave of users. By the time a template is
  obvious, the audio is saturated and the lift is gone. Treat templates
  as starting points, not as guaranteed hits.
- **Skipping the originality pass because Edits exported it.** A clean
  Edits export with no watermark can still be unoriginal under the
  guidelines if the underlying footage is someone else's clip. Edits
  doesn't change the substantial-edit bar.

## Quick decision

- Watermarked export from CapCut hitting Reels? Switch to Edits for the
  export step.
- Want to ride a trending audio with a clean template? Try Edits.
- Already have a working desktop edit workflow with no watermark? Stay
  put.
- Hook is weak? Don't open any editor yet. Fix the hook first.

# Reels Insights

The Insights dashboard for a single Reel is the only honest feedback loop on
the platform. View count alone tells you almost nothing. The post-watch
behaviour metrics tell you what actually happened: who shared, who skipped,
who watched and forgot, who watched and followed.

The job of this reference is to teach you which metrics map to which levers,
so a flop or a hit can be read for what it actually was.

## The metrics that matter

The Reels Insights dashboard is organised by tabs (Overview, Engagement,
Audience). The metrics below are the ones worth acting on. Other numbers
in the dashboard (impressions, profile visits, etc.) are downstream signals
that mostly track the ones below.

### Skip rate

The percentage of viewers who scrolled away before a meaningful watch
threshold (typically the first three seconds). This is the cleanest read on
your hook.

- **High skip rate, low total views**: the hook failed the three-second
  cliff. The seed test never expanded. Rewrite the opening three seconds.
- **High skip rate, high total views**: rare. Usually a Reel that benefited
  from a strong audio trend or a topic spike. The hook still needs work for
  the next Reel.
- **Low skip rate**: the hook held. Whatever else underperformed, the
  opening did its job. Look downstream for the failure.

Your baseline depends on your account and your niche, so compare against
your own prior Reels, not against an absolute number.

### Share rate (and sends per reach)

The percentage of viewers who sent the Reel in a DM. The single most
important number for cold growth.

- **Above your baseline**: the payoff is shareable and the named-persona
  CTA worked. This is the metric that buys non-follower reach. Reels with
  above-baseline share rate tend to keep compounding for days.
- **Below your baseline**: the Reel didn't earn the favour. Either the
  payoff isn't useful to a specific person, or the CTA was generic, or
  both. See `sends-playbook.md` for the rewrite.
- **High views but low share rate**: classic "viral but flat" Reel. The
  audio or topic spiked the reach, but the Reel didn't earn the share that
  would have made the reach sustained.

### Retention curve and average watch time

A graph of where viewers dropped off, plus the average percentage watched.
Reading the curve is more useful than reading any single number.

- **Cliff in the first three seconds**: hook problem. (Match this against
  skip rate; they tell the same story two ways.)
- **Cliff at 5 to 10 seconds**: hook held but the body didn't deliver on
  the promise fast enough. Front-load more of the payoff.
- **Gradual flatline in the middle**: no escalation. The Reel opened well
  and then bled out. Add a re-hook in the middle or compress the section.
- **Spike at the end (rewatches)**: the loop landed. Lean into this format.
- **Cliff right before the payoff**: viewers gave up before getting the
  value. Either the body is too long, or the build-up reads as filler.

A Reel with strong retention can outperform a shorter Reel with weak
retention because total seconds watched matters. Don't shorten reflexively
if the retention curve is healthy.

### Followers from this post

The number of new followers Instagram attributes to viewing this specific
Reel. The most actionable single addition to Insights.

- **High views, near-zero follows**: the Reel got reach but didn't make the
  topic legible. Viewers watched and scrolled. The fix is in the topic
  signal: on-screen text, caption keyword, profile bio (does it match the
  Reel's topic?).
- **Moderate views, healthy follows**: the Reel did the job a Reel is meant
  to do. The topic was clear, the value was clear, the profile sold the
  follow.
- **Followers from this post versus your average**: the relative number
  matters more than the absolute. A Reel that drove three times your usual
  follow rate is a format to repeat.

A "viral" Reel that brings no followers is not actually growth. Use this
metric to separate views from real growth.

### Save rate

The percentage of viewers who saved the Reel. Weighted below sends, but
still a real signal of useful content (especially for tutorials, lists,
recipes, and reference material).

- **High save rate**: viewers want to come back to it. Common on educational
  or product-list Reels. Pairs well with named-list payoffs (see
  `sends-playbook.md`).
- **Low save rate on educational Reels**: the content wasn't actually
  useful enough to keep, or the payoff was unmemorable. Compress the
  takeaway into something nameable.

### Likes per reach

The like rate from non-followers. Mostly a connected-reach signal (it
helps you get re-shown to existing followers in main feed), not a cold-
reach signal. Don't optimise heavily for it. Sends and watch time matter
more for new audience.

### Comments

Comments matter mostly for connected reach and for community signal. A Reel
with a healthy comment thread keeps surfacing back to your followers in
main feed. The thread itself can also be the reason a viewer sends or
follows. But a Reel can be a hit on cold reach with very few comments if
the topic doesn't naturally invite discussion.

### Post-watch behaviour

A grouped read of what viewers did after the Reel: visited the profile,
followed, scrolled past, or kept watching another of your Reels (auto-play
into your next Reel). The auto-play continuation is particularly useful:
a high "watched your next Reel" rate means your account's topical
consistency is paying off.

## Reading a Reel as a whole

Most Reels fail for one of four reasons. The Insights dashboard tells you
which one.

1. **Hook failed.** Skip rate is high. Retention cliff in the first three
   seconds. Few views, regardless of everything else. The fix is in the
   opening. (See `reels-hook.md`.)
2. **Body failed.** Hook held but retention bleeds out in the middle. The
   fix is in escalation, pacing, and the re-hook.
3. **Topic failed.** Healthy reach and watch time but zero follows and low
   sends. The Reel was watchable but didn't make the topic legible enough
   to act on. The fix is in on-screen text and caption keyword.
4. **CTA failed.** Watch time and saves are fine, sends are low. The
   payoff is good, but the Reel didn't earn the favour of a share. The fix
   is a named-persona send-prompt. (See `sends-playbook.md`.)

A Reel can fail for more than one reason. Read the metrics in order and
fix the earliest failure first. Fixing the CTA on a Reel with a broken
hook does nothing.

## What not to read

- **Total view count.** Without a denominator (reach) and the share /
  follow context, view count is vanity. A Reel with 100K views and zero
  follows is a worse outcome than a Reel with 10K views and 200 follows
  for an account that's trying to grow.
- **Likes alone.** Likes are the cheapest action and the weakest distribution
  signal on Reels.
- **Day-one numbers as a verdict.** Reels can keep distributing for days
  or weeks if the share signal stays healthy. Don't write off a Reel on
  the 24-hour read alone unless the hook clearly failed.

## Working with the rest of the skill

Use `assets/insights-readout.md` to capture the numbers in a structured
form, then use the failure-mode list above to decide what to fix. For
diagnosing a specific underperforming Reel, walk `diagnose-flop.md`.

# Originality and audio

Meta's Original Content Guidelines are the structural rule that decides
whether your Reel is eligible for cold distribution at all. Accounts that
primarily post unoriginal content are excluded from recommendations to
non-followers. Followers can still see the posts in feed, but the cold-traffic
tap is closed. Originality is also the only lever in this skill that can flip
from "fine" to "shadow-capped" with no warning, so it deserves its own pass
on every Reel.

Audio is the second half of the same pass. The trending-audio lever still
moves cold reach, but only on accounts that can use it (Creator, not
Business) and only inside a short trend window.

## What counts as original

Meta defines original content as material you shot, designed, or
substantially edited. The bar is transformation, not novelty. A Reel using
licensed audio and your own footage is original. A Reel reposting someone
else's video with a watermark on top is not.

The following do not count as transformation on their own:

- Adding borders, frames, or letterboxing.
- Stamping a watermark on top.
- Speed changes (sped up or slowed down) with no other edit.
- Attribution overlays ("credit to @creator").
- Basic crops or zooms.

The following do count as substantial transformation:

- Adding your own voiceover or commentary throughout.
- Cutting and re-sequencing clips with new pacing.
- Adding original on-screen text that changes the meaning.
- Reaction format where your face and reactions take significant screen time.
- Translating, subtitling with original analysis, or restructuring into a
  new format.

If you're not sure whether a Reel qualifies, it probably doesn't. Treat
borderline cases as risky.

## The watermark penalty

Mosseri has explicitly said Reels with TikTok or CapCut watermarks get
reduced reach in Explore and recommendation surfaces. Your own brand
watermark is fine. Instagram clarified that own-logo watermarks don't trigger
the penalty.

Practical rules:

- Never publish a Reel with a TikTok watermark visible. Re-export the footage
  through any editor that doesn't stamp one (Instagram's Edits app does not).
- Never publish a CapCut watermark. The free CapCut export adds one by
  default. Remove it before posting or edit somewhere else.
- Your own brand logo, handle, or website on screen is safe.
- App UI in the corner of a screen-recording counts as a third-party
  watermark in practice. Crop it out or recreate the screen capture cleanly.

The penalty is silent. You won't see a warning. The Reel just doesn't get
distribution to non-followers. By the time you notice the pattern across
multiple Reels, you may already be in a recovery window.

## The rolling 30-day recovery window

Originality is judged over a rolling 30-day window of your account's posts.
Once most of your last 30 days qualify as original, you become eligible for
recommendations to non-followers again.

What this means in practice:

- One unoriginal Reel won't cap your account. A run of them will.
- Recovery is automatic but not instant. It takes posting original content
  consistently over the window for the account to come back into
  recommendations.
- The window resets day by day, not after a single corrective post. There's
  no "I posted one original Reel, I'm fixed" lever.
- New accounts don't get a grace period. The same rules apply from day one.

If an account suspects it's in a recovery window (sudden drop in non-follower
reach across multiple Reels, no policy notification, no obvious hook
problem), the move is to publish original Reels consistently for at least
30 days and stop reposting in the meantime. There's no appeals form.

## Creator vs Business audio access

Audio licensing is split by account type, and the difference matters a lot.

- **Creator accounts** get the full Instagram music library, including the
  popular trending songs that drive most audio-based distribution lifts.
- **Business accounts** are restricted to the commercial-use Sound Collection,
  which is a much smaller pool and largely excludes the songs driving current
  trends.

For any creator with commercial intent who still wants access to trending
audio, the working answer is to use a Creator account, not a Business
account. Switching is free and reversible.

Legal note: switching to Creator does not actually license you to use a
copyrighted song commercially. It just removes Instagram's in-app block. If
you're posting a Reel that ads a product, the trending song you can now
attach may still be a copyright risk for your brand. Most small creators
operate inside Instagram's de facto tolerance; large brands and ad
placements have a separate licensing problem.

If the Reel will be promoted as a paid ad or repurposed off-platform, default
to commercially licensed audio or original audio. Trending songs don't carry
through to ad versions cleanly.

## Trending audio windows

Trending audio still moves cold reach. The window is short.

- A typical audio trend runs about one to three weeks from breakout to
  saturation.
- The strongest lift is in the first five to seven days, before the sound
  is overused.
- A useful filter when picking an audio: it shows an upward trend arrow in
  the audio page, and the total uses are still relatively low (rough rule
  of thumb: fewer than around 30,000 uses).
- Many audio trends originate on TikTok and migrate to Reels with a one to
  two week lag. Cross-platform creators can sometimes catch a trend earlier
  on Reels by watching TikTok first.
- Audio trends that are obviously past peak (everyone has used the sound,
  multiple parodies exist, brands have jumped on it) no longer carry a
  distribution lift. Use the sound only if it actually fits the Reel, not
  for the lift.

## Audio attachment failure modes

A Reel can play with no audio if licensing fails silently:

- A Business account picking a trending song outside the commercial pool
  will see the audio strip on publish. The Reel plays muted for most viewers.
- Music rights vary by region. A sound that's available in one country may
  be blocked in another. International audiences may see the Reel muted.
- An audio added at edit time may be removed at publish if Instagram
  detects a copyright issue.

After publishing, check the Reel on a test non-creator account to confirm
audio is attached. If it's stripped, recut with a safer audio choice.

## Originality and audio pass checklist

Before publishing or trialling, confirm:

- No TikTok watermark visible anywhere in the frame.
- No CapCut watermark in the export.
- No third-party app UI in the corner of any screen recording.
- The Reel qualifies as substantially edited (your voice, your cuts, your
  on-screen text), not a repost with a wrapper.
- Audio choice fits the account type (Creator for trending, Sound Collection
  for Business).
- Trending audio is still inside its window (upward arrow, not yet saturated).
- The Reel was test-played on a non-creator account to confirm audio
  attached cleanly.

If any item fails, fix before publishing. The cost of one slipped item is
days of capped distribution.

# Reels hook craft

Hook archetypes work the same across short-form. The Reels-specific work
happens in three places: the three-second cliff that decides whether the seed
test expands, the sound-off viewer who needs the hook in text on the first
frame, and the payoff promise that has to read as shareable from the start.

For archetypes themselves (curiosity gap, withhold, stakes, bold claim, named
viewer, mistake confession, listicle, in media res), see the `viral-hooks`
skill. This file is what changes when the surface is Reels.

## The three-second cliff

Reels has an explicit public hurdle at the three-second mark. Drop-off before
three seconds reads as a failed hook and caps the seed test before the
algorithm can expand the audience. Everything in the opening is in service of
keeping the viewer past that cliff.

Practical implications:

- The first frame has to do real work. No logo intro, no slow zoom-in, no
  black frame, no "hi guys."
- The first spoken word should be the most interesting word in the line. Cut
  any preamble before it.
- Front-load the visual hook. Movement, a face, a reveal, a contrast. A flat
  static frame in the first second is a flat-line in retention.
- Length doesn't excuse a slow open. A 30-second Reel and a 90-second Reel
  both die at the same three-second cliff if the open is weak.

## Sound-off legibility

A meaningful share of Reels viewers watch with the phone muted, especially
during the early seed test. If the hook is only spoken and not on screen, the
sound-off viewer scrolls. That's a failed seed signal you didn't need to
spend.

Rules for the on-screen hook:

- Three to five words, bold, on the first frame. Not a sentence. A headline.
- Restate the hook, don't transcribe the whole voiceover. The text should
  carry the meaning even with no audio.
- Place it where Instagram's UI won't cover it. Keep text out of the bottom
  fifth and top fifth where captions and account info sit.
- Use high-contrast styling. The default white sans-serif on busy footage
  fails on a phone in sunlight.

The same hook line you'd say works as the on-screen text most of the time.
Tighten it to a headline. Spoken: "If you keep breaking out after every new
serum, this is for you." On screen: "Breakouts from every new serum?"

## Payoff preview

The hook is also a promise. On Reels, the share metric punishes hooks that
make a promise the body never pays off, because viewers who finished feeling
cheated don't send the Reel to anyone. That's a direct hit to sends per reach,
which is the metric that buys cold reach.

A good Reels hook signals shareable curiosity, not generic curiosity. The
viewer should think "I know one specific person who needs to see this" before
the payoff even arrives. Pair the archetype with a named-persona angle so the
share intent forms in the first three seconds.

Spoken hook: "Every dermatologist I asked told me the same thing about niacinamide."
Implied share trigger: "I should send this to the friend who tried niacinamide
last month."

## First-frame composition

The first frame is the thumbnail in the Reels tab, the cover on the grid (if
shared), and the visual that decides whether a thumb keeps scrolling. Treat
it as a hook on its own.

- Lead with a face, a reveal, or a sharp visual contrast. A talking head with
  a clear expression beats a wide product shot.
- Avoid third-party app UI in the frame. CapCut overlays, TikTok timers, and
  app chrome read as reposted content and tank cold distribution (see
  `originality-and-audio.md`).
- The aspect ratio matters for grid framing too. Reels appear on the profile
  grid in a 3:4 crop. A composition centred for 9:16 video can crop badly on
  the grid. Frame the hook in the upper two-thirds.

## On-screen text timing

Match the on-screen text to the spoken beats, but don't transcribe.

- Hook text appears on the first frame and holds for two to three seconds.
- Escalation beats get short text labels, three or four words each.
- The payoff line gets its own bold beat, ideally the cleanest piece of text
  in the video.
- The send-prompt CTA at the end gets its own frame and its own text. Don't
  pile it on top of the payoff.

If on-screen text is constant noise that changes every half-second, the eye
gives up reading and engagement drops. Use text as punctuation, not as a
caption track.

## Anti-patterns

- **"Hey guys, welcome back to my channel."** The single fastest way to fail
  the three-second cliff. Cut everything before the first interesting word.
- **Logo-first intros.** Branded animations and bumper cards burn the cliff
  on someone else's voice (yours, but yesterday).
- **Vague benefit hooks.** "Level up your skincare." The hook has to name a
  concrete result, a specific viewer, or a real stake. See `viral-hooks` for
  archetype options.
- **Hook that only works with audio.** If a sound-off viewer can't tell what
  the Reel is about in two seconds, you've spent your reach on people who
  scrolled.
- **Clickbait without payoff.** A withheld reveal that never reveals tanks
  sends and trains the algorithm that your account makes empty promises.
- **Trying to hook three different audiences at once.** A muddy hook to a
  broad audience reads as nothing to anyone. Pick one viewer per Reel.

## Quick checklist

Run any candidate Reels hook through `assets/reels-hook-checklist.md` before
publishing or trialling.

# Sends playbook

A DM share carries roughly three to five times the weight of a like when
Instagram decides whether to push a Reel to non-followers. That makes sends
per reach the single most important metric on Reels for cold growth.
Likes per reach still helps you re-reach existing followers in main feed, but
it does not buy you new audience the way a send does.

The whole point of this playbook is to build the Reel so the natural reaction
is "I should send this to [specific person]." That reaction has to start
forming in the hook, not in the CTA.

## Why sends carry the most weight

A share is a costly action. Liking a Reel takes a half-second tap. Sending a
Reel takes a real decision: who needs this, why, and is it worth the social
cost of the send. Instagram treats that cost as a quality signal because it's
hard to fake. A Reel that earns sends has earned an endorsement from one
person to another, which is what the recommendation system is trying to
predict at scale.

This is also why send-prompts work where like-prompts don't. "Like this if
you agree" is a free action and the algorithm pattern-matches the CTA as
engagement bait. "Send this to the friend who buys every serum" is a
specific instruction that triggers a specific behaviour for a specific
person.

## Structures that earn sends

These are the recurring shapes of high-send Reels. Pick one and build the
script around it. Don't stack three.

1. **Named-persona favour.** The payoff is information that a specific
   recognisable person in the viewer's life needs. The Reel is the favour.
   Example: "If your friend keeps mixing retinol and vitamin C, send this
   before they wreck their skin barrier." The viewer sends because they
   actually know that friend.

2. **Receipt for a running argument.** The Reel ends an ongoing disagreement
   between the viewer and someone they know. Example: a Reel proving a
   common skincare claim is wrong becomes ammunition in a back-and-forth.
   The viewer sends because they win the argument by sending.

3. **Tagged callout.** The hook directly names a behaviour the viewer
   recognises in someone else. Example: "POV: your friend who has tried
   every new mascara." The viewer sends because the Reel is the callout
   they wanted to make themselves.

4. **Useful list with a memorable name.** A Reel with a clear takeaway and a
   name viewers can repeat ("the three-product face") earns sends because
   the viewer thinks "I want to share this list with [person] later." The
   name does the heavy lifting in the DM thread.

5. **Validation send.** A Reel that confirms a feeling the viewer's friend
   has been having. Example: "If your skincare obsessed friend feels guilty
   about throwing out half-used serums, this is the permission slip." Sends
   land as "see, you're not alone."

6. **Concrete how-to for a specific situation.** A Reel that solves a small
   specific problem someone the viewer knows actually has. Generic how-tos
   don't earn sends. Specific how-tos for a recognisable person do.

## Reframing the CTA

A strong CTA names the recipient as well as the action. Compare:

- Weak: "Follow for more." (generic, low value, engagement-bait pattern)
- Weak: "Save this." (better, but no targeted recipient)
- Weak: "Share with a friend." (no friend in particular comes to mind)
- Better: "Send this to the friend who buys every new serum."
- Better: "Send this to your sister who keeps trying retinol on broken skin."
- Better: "Send this to whoever swears by their 12-step routine."

The pattern is: send this to [the specific person who fits this Reel].
Specificity is the lever. A vague "tag a friend" might earn one in twenty
sends. A named persona earns a send from anyone who can picture that person.

Place the CTA at the end of the Reel, in voice and in on-screen text, with
its own beat. Don't bury it under the payoff visual.

## Send-suppressing anti-patterns

- **Generic CTAs.** "Like and follow," "double-tap if you agree," "drop a
  heart." These are engagement-bait patterns and they spend the closing
  seconds on a low-weight action.
- **CTA that asks for nothing in particular.** "What do you think?" earns
  comments at best, no sends.
- **Payoff that solves a problem nobody specific has.** If you can't name
  the person who needs this Reel, the viewer can't either.
- **Educational Reels with no shareable container.** Three good tips with
  no organising name don't get sent. Three tips called "the three-product
  face" do.
- **Hook that promises a payoff the body never delivers.** Viewers who
  feel cheated don't send. That single failure compounds across the Reel's
  whole life.
- **Closing on the creator instead of on the share.** "Thanks for watching,
  see you next time" replaces the most valuable moment with throat-clearing.

## Diagnosing low sends in Insights

If share rate or sends per reach is below your normal baseline:

1. Check the CTA. Was it a named-persona send-prompt, or a generic ask?
2. Check the payoff. Is it useful to a specific recognisable person, or to
   a vague "anyone interested in this"?
3. Check the hook. Did it signal a shareable payoff in the first three
   seconds, or just curiosity for curiosity's sake?
4. Check the topic legibility. Is the niche clear without audio?
   (See `caption-and-search.md`.)

Most low-send Reels fail at the named-persona test, not at the production
quality test. Recut the CTA and the hook before reshooting.

## Working with the rest of the skill

The sends playbook depends on the hook making a shareable promise
(`reels-hook.md`), the body delivering on it (see retention in
`viral-short-form`), and the caption reinforcing the named-persona angle
(`caption-and-search.md`). Sends are the metric, but they're earned by the
whole script. The CTA alone doesn't carry them.

# Trial Reels

Trial Reels are Instagram's built-in way to test a Reel on non-followers before
it ever touches your follower feed or your grid. If the goal of the Reel is
cold reach (the usual goal), this is the default first move. A flop costs you
nothing. A winner gives you a confirmed cold-audience signal before you publish
to everyone.

## When to use one (and when to skip)

Default to Trial Reels when all of the following are true:

- The Reel is meant to reach non-followers (most Reels).
- The account is public and has at least about 1,000 followers (eligibility).
- You're not certain the hook lands. Trials are cheapest exactly when you're
  least sure.
- You want a clean test on a new format, hook style, topic, or length.

Skip Trial Reels and publish directly when:

- The Reel is primarily for your existing followers (a behind-the-scenes, a
  thank-you, a community-only joke). Followers don't see trials.
- The Reel is time-sensitive in a way that needs to hit the grid right now (a
  product drop, a live event, a calendar moment).
- You're under the 1,000-follower threshold. The toggle isn't available yet.
  Focus on publishing and on the 0 to 1K slog.

## How the mechanic actually works

- Toggle "Trial" before you publish. The Reel goes out only to non-followers.
- Your followers see nothing in feed, nothing in the Reels tab, nothing on
  your grid. The Reel is invisible to people who already know you.
- At roughly the 24-hour mark you get a read: views, likes, comments, shares,
  plus a comparison to your previous trials. This is the cold-audience verdict.
- At roughly the 72-hour mark Instagram may auto-share the Reel to your
  followers if it judges performance strong. You can also manually publish to
  everyone at any point, including immediately.
- If you do nothing and Instagram doesn't auto-share, the Reel stays a trial.

## What a trial actually confirms

A trial is a non-follower test. The signal it gives you is "did this Reel earn
distribution from people who don't know you." That's the signal that maps to
cold reach on the public Reel.

A trial does not tell you:

- Whether your existing followers would have engaged. That's a different
  audience with different expectations.
- Whether the Reel would have looked good on your grid. The grid is its own
  composition problem (see `caption-and-search.md` for thumbnail notes).
- Whether the comments would have looked the same with followers in the mix.
  Follower comments often anchor the early thread.

Treat a trial as a hook-and-topic test for cold audiences. If it converts,
publishing is a low-risk decision. If it doesn't, the hook or the topic was the
wrong bet for non-followers, not necessarily a failure of the whole idea.

## Reading the 24-hour numbers

Compare against your own prior trials, not against your follower-feed posts.
The audience is different and so the baseline is different.

- **Views trending higher than your usual trials**: the seed test expanded.
  The hook held the first three seconds well enough.
- **Sends and saves higher than usual**: the payoff is shareable and the topic
  is legible. This is the strongest "publish to everyone" signal.
- **High views, low sends and follows**: the Reel got reach but didn't make
  the topic legible. People watched, scrolled, didn't think to share or follow.
  Rewrite the on-screen text and CTA before re-trying.
- **Low views even on the trial**: the hook failed the three-second cliff or
  the topic is too narrow for cold audiences. Reshoot the first three seconds.

See `insights.md` for the full metric playbook. The trial dashboard is a
subset of those numbers, scoped to non-followers.

## The publish decision

You have three real options at the 24 to 72 hour mark:

1. **Share to everyone now.** Numbers are clearly above your trial baseline.
   Don't wait. Sustained early velocity matters and a published Reel can keep
   compounding.
2. **Let Instagram auto-share at 72 hours.** A passive yes. Useful when the
   numbers are okay but not obvious, and you want Instagram's own judgement
   on whether to push it.
3. **Leave it as a trial.** The numbers underperformed. Don't publish. Learn
   from the trial and recut, or move on. Nothing of this Reel will count
   against your follower-feed metrics or your grid.

## Anti-patterns

- Trialing only winners you already love. Trial Reels are most valuable on the
  drafts you're least sure of. If you only trial the safe ones, you learn
  nothing.
- Treating a trial flop as a verdict on the topic forever. Often it's the hook
  or the first frame. Recut the opening three seconds and re-trial before
  abandoning the idea.
- Trialing too rarely. The dashboard's "compare to prior trials" view gets
  more useful with a meaningful sample. Run trials regularly on your weaker
  drafts and the baseline starts to mean something.
- Trialing a Reel and then publishing the same cut at the wrong time.
  Auto-share decisions and your own re-publishes both restart the clock. Pick
  a publish window with intent.

## Quick decision sheet

See `assets/trial-reels-decision.md` for a one-page sheet you can fill out
per Reel.