How to brief and QA ad creative so you get it right the first time

Half of this is a brief that a builder can act on without guessing. The other half, which almost nobody writes down, is the QA order: what to check, in what sequence, and at which moment, because the same note costs ten times more one stage later.

Briefs and QA/By Badal Kariwal/01 Sep 2026/13 min read

What is in here

What actually has to be in a brief?
Five fields most templates leave out
When does a note stop being cheap?
The QA half: meaning first, mechanics last
What is the Stranger Test?
Words and picture make one claim, and you have to read them together
What QA cannot do for you

The short answer

A brief is actionable when it names the reference and what about it you mean, separates the assets that exist from the ones you are assuming, states the one claim you can defend, states the sentence the ad may never say, and names one person who signs off. QA is the other half and it has an order: meaning first, structure second, mechanics last, each checked at the earliest stage it can be checked. A note given one stage late costs many times what the same note cost one stage early.

What you get out of this

A one-page brief template you can copy, including the five fields most templates omit
The QA order, and why checking mechanics before meaning wastes the pass
The five written questions of the Stranger Test, in full
A timeline of when each note is cheap, from your side of the table
The message-logic rule that kills good-looking ads: words and picture make one claim

The same brief, written two ways

Compare

The same brief, written two ways
Dimension	What most briefs say	What a builder can act on
The reference	A link, or a mood board	Yes. A link plus the sentence naming what about it you mean
Assets	A shared drive folder	Yes. Two lists: what exists, and what is being assumed
The message	Three benefits and a tone of voice	Yes. One claim you could defend in writing to a regulator
Constraints	Brand guidelines attached	Yes. The sentence this ad may never say, and who says so
Audience	Women 25 to 44, urban	Yes. Where they are and what they have already seen this week
Approval	The team will review	Yes. One name, one deadline, and what happens if they are away
Success	High engagement	Partly. Three things that would make this a kill

The left column is not lazy. It is what most templates ask for, which is why it keeps producing round two. The right column is the same information with the guessing removed.

01What actually has to be in a brief?

Seven fields, and it fits on one page. If you briefed your last project with two lines and a mood board, that is not a mistake, it is what everyone has been shown counts as a brief. The problem is arithmetic rather than effort: every field you leave out is a decision somebody else makes on your behalf, silently, at the point where changing it is most expensive.

The one-page brief, as we ask for it

Copy this

THE BRIEF                                    one page, no attachments

1  THE REFERENCE
   Link, plus the sentence saying what about it you mean.
   "This one. Not the mood. The way it holds one object for four
    seconds and cuts on the sound."

2  THE ASSETS            two lists, kept apart
   HAVE    product stills (14 files), a 40s phone clip of the
           unboxing, logo in vector, the licensed font
   ASSUME  nothing lives here. Anything not in HAVE gets sourced,
           generated or refused, and the build notes say which

3  THE ONE CLAIM         the thing you can defend in writing
   "Stays on for twelve hours, including in the shower."

4  THE NEVER             the sentence this ad may never say, and why
   "Never the word cure. Regulator, not taste."

5  THE MOMENT            where the viewer is, not who they are
   "In bed at 11pm, sound off, has seen four competitor ads
    this week and can name none of them."

6  THE DECIDER           one name, one deadline
   "Priya. Thursday 18:00. If she is away, it waits."

7  THREE THINGS THAT WOULD MAKE THIS A KILL
   1.  ...
   2.  ...
   3.  ...

Written in a client's own words, not in agency language. If a field is genuinely unknown, write UNKNOWN in it rather than deleting it. An unknown that is written down gets asked about. An unknown that is missing gets guessed.

The field people fight aboutNumber seven. Naming three things that would make you kill the ad is uncomfortable to write and it is the single highest-value line in the document, because it turns a taste argument in week two into a rule agreed in week zero.

02Five fields most templates leave out

Briefing templates are widely available and mostly fine. What they consistently omit are the five fields that carry the risk, because each one asks the client to commit to something rather than describe something. These are the five, with the specific failure each one prevents.

The fields that carry the risk

Five tabs, five omissions

The reference, and the sentence saying what you mean by it

A reference with no sentence attached is a mood board, and a mood board briefs a feeling rather than a film. The sentence has to name a structural thing: the number of shots, the length of the held beat, where the cut lands, what the sound does at the cut.

Prevents: a rebuild of the vibe that passes every language check and misses the whole structure.
Test: could a stranger read your sentence and pick your reference out of five similar films?

Costs you: Two minutes and one specific opinion

Each of these exists because we shipped the failure it prevents. None of them takes more than two minutes to fill in, which is the annoying part.

The reference field is the one worth the most and the one most often filled in badly, so it has its own piece: the reference video is the specification, including how to take a census of one before anybody prompts anything.

03When does a note stop being cheap?

The moment the thing your note is about has been built. That sounds obvious and almost nobody acts on it, because reviewing paper feels like a waste of a busy person's time and reviewing a finished film feels like the real meeting. It is the exact opposite. Here is the same note arriving at six different moments.

One note, six moments, six prices

Click along it

"Actually, the claim is the twelve hours, not the ingredients"

Nothing exists yet. The note is a sentence swap in a document, and every downstream decision is made from the corrected version.

Cost of this note here: one sentence

Read the cost column and nothing else. Every gate in a production pipeline exists to drag a note leftward into a stage where being wrong is affordable.

A note is not expensive because it is a big note. It is expensive because of the moment it arrives.
the whole argument for reviewing paper you find boring

The ad came back wrong. Which stage should have caught it?

Answer two questions

Run this on the last creative that needed a second round. The ending you land on tells you which document or which gate to fix, rather than which person to be annoyed with.

The useful thing about that tree is where it does not end. None of the endings is fix the ad. Three of the five are fix a document or fix a gate, because a second round is almost always a decision that was made too late rather than a hand that slipped.

04The QA half: meaning first, mechanics last

Most QA checklists you can find are mechanics: aspect ratio, safe zones, file size, caption spelling. All necessary, all last. Running them first is how a technically immaculate ad that says nothing gets shipped, because a green checklist feels like an answer. Check whether the thing works, then check whether it is broken.

The order is not a preference. A mechanical fix applied to an ad that says nothing is the most wasteful work available in this job, and it feels productive the whole time you are doing it. Check the meaning first, so you find out whether the mechanics are worth fixing at all.

The QA pass, in the order that saves the most

Tick as you go - it remembers

1. The combined claim. Watch the first two shots. Write down the one claim the words and the picture make together. If it disparages the product, the brand or the buyer, stop here. Nothing downstream fixes it.2. The question. What question does this pose by second three, and at what second is it answered? If you cannot name one, there is no film yet, only an assortment of shots.3. The leak. Scrub the first third. Does any frame show the answer before it has been earned? A finished product in shot two ends the film before it starts.4. The shuffle. Reorder the beats in your head. If they still make sense in any order, it is a montage, not a story. Passing looks like the reordered version becoming actively false.5. Sound at frame zero. Is there something on the soundtrack in the first frames, and is it an event rather than a fade? Then mute it and check the film still works, because most of the feed will watch it that way.6. Type, on the delivered file. Every line readable at 360 pixels wide, against the worst pixel behind it, held long enough to read. Judge the export, never the timeline.7. The set. Put it beside the other three creatives in the batch. If they share more than two of argument, subject, world, energy, point of view and borrowed genre, you have one ad four times.8. Spec, last. Aspect, safe zones, full bleed, loudness, captions, the platform's own rules. Cheap to check and cheap to fix, which is exactly why it goes at the end.

About twenty minutes, run cold, on a phone, at arm's length, with sound. Give it to somebody who did not build the ad and did not read the brief. Their answers and yours diverging is the useful output.

Item six has a floor rather than an opinion behind it. Contrast is measured against the worst local background under each glyph run rather than an average, because an average of a dark doorway and a lit white blouse describes no pixel anybody is looking at. The numbers are in the legibility floors for captions on a phone.

05What is the Stranger Test?

It is a written pass, done cold, before anybody who knows the project sees the cut. You watch the whole film at 360 pixels, the width it actually occupies in a feed, and you answer five questions in writing. Writing them down is not ceremony. A thought you have not written is a thought you will revise to match the film.

Give it to somebody outside the project if you can. If you cannot, wait a night and do it in the morning. The version of you that knows what shot four was meant to do is not a stranger, and cannot be turned into one by trying hard.

The Stranger Test, all five items

One screen at a time

Setup

Watch it cold, at feed size, once

Full cut, at 360 pixels wide, with sound, on a phone. Once only. The second viewing is worthless for this purpose because you now know what happens, and the whole test is about a viewer who does not.

Item a

What question does this pose by second three?

In the words a stranger would use. If your answer is a description of the product rather than a question, the film has no engine and the rest of the test will confirm it.

Item b

When is that question answered?

Give a second, not a range. A question answered at second one is not a question. A question never answered is a mood piece, which is legitimate for a brand film and expensive for a performance ad.

Item c

Does any earlier frame leak the answer?

Scrub the setup frame by frame. Generated cuts leak constantly, because nothing in a generator knows which of your frames is the payoff. The finished thing arriving in shot two is the classic version.

Item d

What does each beat do to the question?

Every beat should raise it, sharpen it, or answer it. Write one verb per beat. Beats with no verb are the ones to cut, and there are usually two of them in any film that feels slightly long.

Item e

What claim do the words and picture make together?

Read from the customer's chair, not the brief's. This is the item that most often produces a kill, and it is the subject of the next section, because the way it fails is not obvious until you have seen it happen once.

1 / 6

It takes about four minutes on a fifteen second film. The output is five written sentences, and they go in the delivery notes whether or not they are flattering.

06Words and picture make one claim, and you have to read them together

We killed a film for a line that was fine on its own. The copy was fear-shaped, aimed at the difficulty of the task, and it sat over a shot of the product kit. Read the way a customer reads, words and picture made a single sentence, and that sentence was: do not buy this. Nothing about the craft was wrong. The combination was wrong.

Read it aloud as one sentence, assembled the way a customer assembles it: the line on screen, then the thing underneath the line. If that sentence would make you hesitate about buying, it is a kill, and no amount of grading rescues it.

Four things that pass a copy review and fail message logic

Flip them

All four are real. Three of them are ours, and the fourth was in a file we were asked to fix rather than to build.

That last card is worth a paragraph on its own. A rebuilt static set went out with zero geometric breaches and was killed entirely on semantics, which is why there is now a comprehension pass whose only job is to answer what is this ad saying, and is every label pointing at the thing it names. Fresh eyes, before the client's. There is more on running those checks in deterministic creative QA.

07What QA cannot do for you

A clean pass is a floor, not a verdict. Fewer than half of the films that clear every automated gate here survive the first human viewing, and that number is not a defect in the gates. It is what they are for. They clear the mechanical floors so that the remaining argument is about whether the ad works, which is the only argument worth having.

The second limit is worse and it generalizes past advertising. A lint is a floor, not a target, and when a check fails you fix the thing rather than the number. The soundtrack that taught us that, tuned green and shipped anyway, is in sound is 40% of the ad.

Four films that went through this order

Ours, briefed and checked this way

Ephoria - campaign ad

Hotel client - campaign reel

Beauty device - campaign ad

Skincare - studio demo

Paper first, sound committed before the first cut, the Stranger Test written down before packaging. The order did not make any of these good. It made it possible to spend the review on whether they were.

If you only adopt one thing from this page, adopt the order. Brief the five committing fields, then check meaning before mechanics, and move every note as far left as it will go. The rest is detail, and the detail is covered in paper before pixels and in the pre-launch checklist for generated footage.

The eight terms this page runs on

Search the vocabulary

Filter terms

8 terms

The one claimBrief

The single statement of fact the ad exists to deliver, written in a form you could defend to a regulator. Everything else in the film may be atmosphere. Three benefits and a tone of voice is not one claim.

The neverBrief

The sentence this ad may never say, written with who says so beside it: regulator, platform, legal or taste. A negative specification is faster to check than a positive one and prevents a re-render.

The deciderBrief

One named person with one deadline and a stated fallback if they are away. A schedule with an unnamed approver in it is a schedule with a hole where the turnaround promise should be.

Stranger TestQA

A written pass done cold at 360 pixels, answering five questions: what question the film poses by second three, when it is answered, whether an earlier frame leaks it, what each beat does to it, and what claim words and picture make together.

Message logicQA

The rule that words and picture are judged as one combined claim, read from the customer's chair. Negative words over the product are a negative claim about the product, however good the shot is.

Comprehension checkQA

A pass whose only job is to answer what this ad is saying and whether every label points at the thing it names. It exists because a set of ours passed every geometric check and was killed entirely on meaning.

Mockup gateStage

A cheap ratified artifact between approved paper and any full-cost build. For a film: styleframes, one motion test of the riskiest beat, and a sketch of the bed. Minutes of compute, never hours.

Dwell floorFloor

The minimum time a line of type stays fully opaque, computed from its own word count rather than chosen. Fades do not count toward it, because text at partial opacity is not readable text.

House vocabulary. Take it into your own process if it is useful. Having one agreed name per idea matters more than which names you pick.

Questions people actually ask

Open what you need

What should be in a creative brief for video ads?

Seven fields on one page: the reference plus the sentence saying what about it you mean, two asset lists kept apart for what exists and what you are assuming, the one claim you can defend in writing, the sentence the ad may never say and who says so, where the viewer is when they see it, one named decider with a deadline, and three things that would make you kill it.

How do I QA a video ad before launch?

In this order: the combined claim the first two shots make, the question the film poses and when it is answered, whether any early frame leaks the answer, whether the beats survive being shuffled, sound in the first frames, type checked on the exported file at 360 pixels, the ad set beside its siblings, and platform spec last.

Run it cold, on a phone, with somebody who did not build it and did not read the brief.

Why does my agency keep needing more rounds?

Usually because the notes are arriving after the thing they are about has been built. A note on the story is minutes at the paper stage and a rebuild at the first cut, and the same note feels identical to the person giving it. Ask to review the story on paper and the mockup before the full build, even when both feel too rough to be worth a meeting.

What is the Stranger Test?

A pre-delivery pass where somebody watches the full cut cold at 360 pixels and answers five questions in writing: what question does this pose by second three, when is it answered, does any earlier frame leak the answer, what does each beat do to the question, and what claim do the words and picture make together. The writing is the point, because an unwritten answer bends to match the film.

How many rounds of revision should a video ad need?

Fewer than you are getting, and the fix is upstream rather than in the revision policy. Most rounds are a decision that was never made in the brief being made late by argument. Name the decider, name the one claim, and name three things that would make you kill it, and rounds tend to collapse to one substantive pass plus mechanics.

Should the brief include a reference video?

Yes, and with a sentence attached saying what about it you mean. A reference on its own briefs a mood, and a mood gets rebuilt as a vibe. Name the structural property you actually want copied: the shot count, the length of the held beat, where the cut falls against the sound, where the type lands.

Every field you leave blank gets filled in by somebody with less information than you, at the most expensive moment available. One page, seven fields, written before anybody opens a timeline.

Where the numbers came from

W3C. Understanding Success Criterion 1.4.3: Contrast (Minimum) - the 4.5:1 ratio the legibility check measures against
BBC. Subtitle Guidelines - reading-rate guidance behind the idea of a minimum time on screen per word
Motion. Creative Benchmarks, 550,000+ ads across 6,000+ advertisers - used for the point that most ads never get meaningful spend

Every figure above links to the place it was published. Numbers marked as ours are measured inside this studio and we say so where they appear. We do not print a statistic we cannot point at.

Badal Kariwal

Runs Sutra Haus, a one-person ad studio that has shipped over a thousand finished creatives - film and stills - for DTC brands and hotels. Writes here about what the work actually taught him, including the parts that failed. The person who reads your brief is the person who builds the work. Send him something to make.

Test the brief, not the studio

Send a link and one sentence. Get a finished ad back.

One product link and one line saying what about your reference you actually mean. We build one finished cut from it, free, inside three days, yours to run whether or not we ever work together. It is the cheapest way to find out whether your brief says what you think it says.

Replies within a day. Ad within three.