By the time a piece is finished you have watched it forty times. You know what's coming. You supply context automatically. You have entirely lost the ability to see it the way the only person who matters will see it — once, cold, on a phone, half-distracted.
This is a structured last pass designed to work around that. It takes about forty minutes and it catches things no amount of re-watching will.
The conditions matter more than the checklist
Do this properly or don't do it.
On the device it'll be watched on. Phone, held at normal distance, not a colour-graded monitor. Most content is watched on a screen the size of a hand.
At the volume it'll be watched at. Which is frequently zero. If the piece doesn't work muted, you need to know now.
In the feed if you can. Upload it unlisted or private and view it in the actual product surface. The context around a piece is part of the piece.
After a break. Minimum a few hours, ideally overnight. Familiarity fades slowly and there is no shortcut.
Once, without stopping. No scrubbing back. You get one pass, like everyone else.
Pass 1 · The cold watch
Watch it start to finish. Don't take notes during — notes turn you back into an editor.
Immediately afterwards, write down three things from memory:
- What was it about, in one sentence?
- What's the strongest moment?
- Where did your attention go?
Compare those to what you intended. The gaps are the findings.
If your one-sentence answer doesn't match the sentence you wrote when you pressure-tested the idea, something moved during production. That happens constantly and it's usually worth fixing.
Pass 2 · The first ten seconds
Watch only the opening, three times.
Second 0–3. What is on screen? Is it moving? Is there a face? Does anything happen, or is it a logo animation eating the most valuable three seconds you have?
Second 3–10. Has a question, tension or promise been established? Would someone who has never heard of you have a reason to stay?
Muted. Does the opening work with no sound? On feed placements, that's the default state, not an edge case.
The hook is the highest-leverage ten seconds in the piece and the part most likely to have been assembled last, in a hurry.
Pass 3 · The attention audit
Now watch again and mark every moment your attention drifts — even a second. Especially the ones you've trained yourself to sit through.
Then look at what's on screen at each mark. Almost always one of:
- A point already made
- A transition longer than it needs to be
- Setup for a payoff that's too far away
- Someone being polite instead of saying the thing
- B-roll running past its usefulness
The fix is nearly always a cut, not an addition. If you have retention data on comparable pieces, this is where to check it — the drop-off point in a retention curve is the closest thing to free attribution you'll get.
Pass 4 · The brief check
Open the creative brief and read it against the cut.
- Does the piece do the one thing the brief asked for?
- Is the audience the brief named the audience this speaks to?
- Is the single takeaway present, and is it unmissable rather than merely present?
Drift is normal and not automatically bad — some drift is the work improving. What matters is that it's noticed and accepted rather than accidental. If the piece is now doing something different, say so explicitly and decide whether that's the piece you want.
Pass 5 · The publishing surface
The cut is not the whole product. Check what the piece will actually appear as.
Thumbnail at real size. Shrink it to the size it appears at in a feed. If the text is unreadable or the subject is unclear, it's a bad thumbnail regardless of how it looks at full size — see thumbnails that read at feed size.
Title as a promise. Does the piece deliver what the title implies? An overclaimed title buys a click and spends trust.
First line of the description. Usually the only line most people see.
Captions. Burned in or platform-generated, and are the platform-generated ones accurate? Auto-captions mangle product names and jargon reliably.
Aspect ratio per placement. A landscape cut in a vertical slot is a letterboxed compromise, and it's a shoot-day decision you can't fully fix here.
End state. What's on screen when it stops, and what does it ask for? A piece that ends with nothing asks for nothing.
Pass 6 · The stranger test
Send it to one person outside the project. Not for approval — for two specific questions:
- What do you think this is about?
- What would you do after watching it?
Give them no context. Context is the thing you're testing for.
Their answer to question 1 is your real message, whatever you intended. If it differs from your brief's one-line takeaway, the piece isn't communicating what you think.
One outsider is enough for this. You're not sampling an audience — you're checking whether the piece stands alone, and one person who's never seen it will find that out.
What to do with what you find
Sort every finding into three buckets before you touch the timeline:
Fix now. Cheap, high-impact, mostly cuts. A slow open, a dead thirty seconds, an unreadable thumbnail. Do these.
Fix if there's time. Real but marginal. Be honest about whether "if there's time" means never.
Log for next time. Anything structural — a decision made at brief or shoot stage that can't be repaired in post. Write it down somewhere the next brief will actually see it. This is the bucket that improves your work over a year, and it's the one that evaporates because nobody records it.
Where Acumin fits
Pre-Flight runs a structured read on a cut before it goes out, and thumbnail review checks legibility at feed size. Both are reads against your own history and your category — they'll flag a slow open or a thumbnail that won't survive scaling.
What they can't do is pass 6. A tool has the full context by definition; the thing you're testing is whether the piece works without it. Send it to a person.
How to use this today
Take the last piece you published. Watch the first three seconds on your phone, muted.
Then find its retention curve, if you have one, and look at where people left. Go to that timestamp and watch what's there.
That's a free version of pass 3 on work you've already shipped — and whatever you find will probably be true of the piece you're about to publish too.
Related: How hooks work, structurally is pass 2 in depth. Giving feedback without destroying the edit is how to hand these notes to someone else without wrecking the thing.