Most format advice stops at the container — square or vertical, image or video, how many seconds. That matters, and the platform data on it is at the bottom of this page, but it is the smaller half of the question. The larger half is what kind of argument the ad is making, and advertising has no standard vocabulary for that at all, so this page builds one. Part three of the ad creative series.
Start with statics, then promote the winners
For a small budget this is the sequence I recommend every time, and the reasoning is about learning rate rather than taste.
Statics are faster and cheaper to produce than video, which means you can put more distinct messages into the market for the same money. Reuse anything that already exists — organic posts that performed are free creative — or build a quick set of three to five statics carrying genuinely different angles. Not three colourways of the same claim. Three different arguments.
That set tells you which hook, angle or story the market responds to, and it tells you quickly. Then turn the proven statics into video: a slideshow-style edit built from creative that has already earned its place is the cheapest next step available, and it starts from a known-good message rather than a guess.
The mistake is running the sequence backwards — commissioning video first, discovering the message was wrong, and having spent the production budget finding out.
The 43 archetypes
Grouped by the job the creative does in the buyer's head, because grouping by layout produces nothing useful — a two-column grid can be a comparison, a feature set or a review wall. Each entry lists its most native aspect ratios and its funnel stage: top, middle or bottom.
Pattern interrupt and native — 10 archetypes, mostly top of funnel
These win by not looking like ads. The feed has trained everyone to skip polish, so the creative borrows the visual language of the surrounding organic content instead. Highest ceiling on cold traffic, and the highest risk of misfiring for premium or regulated brands — the format has to be a plausible thing a person would actually post.
- Ugly / anti-design — 4:5, 1:1 · top. Deliberately unpolished: a raw screenshot, a marker circle, a default system font, no grid. Polish is the pattern; breaking it buys a half-second, and the roughness reads as a person rather than a brand.
- Meme format — 1:1, 4:5 · top. A recognised meme layout carrying your category's specific pain in the caption. Borrowed recognition — the format is pre-decoded, so all the attention lands on your line. Decays fast as the meme ages.
- Notes-app confession — 9:16, 4:5 · top to middle. A screenshot of the phone's notes app holding a blunt first-person list. The format is the internet's established grammar for an unguarded admission, so the copy inherits candour it did not earn.
- Direct-message exchange — 9:16 · top. A two-person thread arriving at a recommendation. Stages word-of-mouth: the claim comes from a friend in the fiction, and the reader completes the referral themselves.
- Post screenshot — 1:1, 4:5 · top to middle. A short social post, yours or a customer's, screenshotted as the whole ad. Works best when the post genuinely exists and has engagement on it.
- Forum thread — 4:5, 9:16 · middle. A question post with the top reply visible. Mirrors how people research purchases now. Never fabricate threads that imply real users said something.
- Search-bar prompt — 9:16, 4:5 · top to middle. The high-intent query typed into a search field with the answer below. Shows the buyer their own question in their own words.
- Comment section — 4:5, 9:16 · middle. A stack of real comments. The disagreement in a real thread reads as more credible than unanimity.
- Notification stack — 9:16 · top to middle. A lock screen stacked with bookings, orders or new leads. Shows the outcome as a felt experience rather than a number.
- Photo-dump carousel — 9:16, three to seven slides · top. A casual sequence paced like a personal post. Earns a swipe instead of demanding a click; the last slide does the selling.
Proof — 7 archetypes, middle and bottom of funnel
The strongest of these are the ones where the evidence is something a real customer actually said.
- Testimonial card — 4:5, 1:1 · middle to bottom. One quote, attributed, treated as the whole frame.
- Review wall — 4:5, 1:1 · middle to bottom. Many short reviews tiled together; volume is the argument.
- User-generated screenshot — 4:5, 9:16 · top to middle. Customer content reproduced as-is, with the roughness left in — that is exactly the signal that makes it work.
- Press or editorial clipping — 4:5, 1:1 · middle. Borrowed authority. Use a publication that actually covered you.
- "As seen in" logo bar — 1:1, 1.91:1 · middle to bottom. Publication marks in a row, often as the base of another layout.
- Case-study stat — 1:1, 4:5 · middle to bottom. One outcome number, the client type, and nothing else.
- Founder letter — 4:5 · middle. First-person text as the creative. Works when there is a real person to be.
Comparison and positioning — 6 archetypes, middle of funnel
These do not argue that you are good; they argue that the alternative is worse. Powerful when you are replacing an existing habit, spend or vendor, because the buyer already has the reference point loaded. Naming a competitor invites both legal exposure and their retargeting — the process-based versions get most of the effect with none of that.
- Us-versus-them matrix — 4:5, 1:1 · middle. Two columns, five to seven rows, ticks against crosses. The grid does the arguing, and it is scannable at thumbnail size.
- Before / after split — 4:5, 1:1 · middle. One frame divided: chaos against resolution. Regulated in health, finance and beauty — check the platform policy first.
- Price anchor — 4:5, 1:1 · middle to bottom. Their cost struck through beside yours. The first number sets the scale the second is judged against.
- Old way / new way — 4:5, 1:1 · top to middle. Compares two processes rather than two brands. All the contrast of a competitor comparison with none of the exposure, because it attacks a status quo nobody defends.
- Myth versus fact — 1:1, 4:5 · middle. A widely-held belief struck out, the correction beneath. Being told you are wrong about something is unusually hard to scroll past.
- "X meets Y" — 1:1, 4:5 · top. Positions a new product at the intersection of two things the reader already understands. The fastest way to explain something genuinely new in one line.
Educational and utility — 8 archetypes, top and middle of funnel
These give something away before asking for anything. They earn saves and shares, which is a real distribution advantage. The failure mode is being so useful that nobody needs the product — the information should make the reader more aware of the problem, not solve it.
- Cheatsheet card — 1:1, 4:5 · top to middle. A dense reference card. Optimised for the save, not the click, and a saved ad gets seen more than once.
- Checklist — 4:5, 9:16 · top to middle. The reader audits themselves against it in real time, and a gap they find is a need you did not have to assert.
- Listicle — 4:5, 9:16 · top to middle. A stated count promises a finite read, and it survives cropping better than most layouts.
- How-it-works diagram — 1:1, 4:5 · middle. Three or four steps with arrows. Buyers rarely doubt the benefit; they doubt what it will cost them to get there.
- Feature anatomy callout — 4:5, 1:1 · middle. Product centre-frame with pointer lines to labelled benefits. The exploded-diagram convention reads as documentation, which is trusted differently from advertising.
- Red arrow and circle — 1:1, 4:5 · top to middle. A screenshot with one crude ring around the number that matters. The video-thumbnail vernacular for "look here", and legible at thumbnail scale.
- Stat or data-visualisation card — 1:1, 4:5 · top to middle. A chart as the hero, with the takeaway as the headline. Only works if the shape of the line is obvious in half a second — label the conclusion, not the axes.
- Calendar or schedule — 4:5, 1:1 · middle. What happens in week one, week two, week four. Pre-empts "how long will this take".
Product-forward — 7 archetypes, all stages
The oldest family and still the workhorse: show the thing. Lowest creative risk and lowest ceiling, which makes them the right control to test everything else against. For software the product is an interface, so the craft moves into which single screen you crop to.
- Hero on white — 1:1, 4:5 · middle to bottom. Product alone, generous margins, one short line. The default control ad.
- In-context lifestyle — 4:5, 9:16 · top to middle. The product where it is actually used, with a person using it. The setting qualifies the audience faster than targeting does.
- Interface screenshot hero — 4:5, 1:1 · middle. The real interface cropped tight on the one screen carrying the promise. Crop hard — a full dashboard at feed size reads as noise.
- Hook plus benefit stack — 4:5 · middle. Headline across the top third, product in the middle, three bullets underneath, in the order a skim actually happens. Boring, and it keeps winning.
- Feature grid — 1:1, 4:5 · middle to bottom. Breadth as the argument. Poor on cold traffic, where breadth reads as unfocused.
- Packaging and unboxing — 4:5, 9:16 · middle to bottom. Sells the moment of receipt rather than the object.
- App-store style — 9:16 · middle. Device frame with a caption above, repeated as a set. A format people are already fluent in reading.
Offer and urgency — 5 archetypes, bottom of funnel
These convert people who are already convinced and have not moved. They work almost exclusively on warm audiences and they wear out fastest. Hold them for retargeting and real deadlines.
- Discount or offer card — 1:1, 4:5 · bottom. The offer as the entire creative. No interpretation required, and it reads fully at thumbnail size.
- Offer stack — 4:5 · bottom. Everything included, itemised, each line with a value and a total underneath. Direct-response heritage, still effective, easily overdone.
- Guarantee or risk reversal — 1:1, 4:5 · bottom. Moves the risk to the seller. Often the last objection standing on a retargeting audience.
- Countdown or deadline — 9:16, 1:1 · bottom. Only run it against a deadline that is real. A permanent countdown teaches the audience to discount every future one.
- Receipt — 4:5, 9:16 · middle to bottom. The cost comparison rendered as an itemised receipt. A receipt is a document, not a claim, so the arithmetic feels audited rather than argued.
The container data
This is the part the platforms have actually measured, and the lifts are large enough that ignoring them is expensive.
The resolution technique is the one people skip, and it is the largest single lift in the set. Uploading at 720p or better costs nothing.
On Google, adding a single vertical asset to a campaign is worth about 35% more conversions on Demand Gen, 40% higher view-through rate on video view campaigns, and about 45% better viewability on video reach campaigns, per Google's own before-and-after data. Same footage, different crop.
What to read next
- User-generated content and creator ads — the format most of these archetypes borrow from.
- Ad creative testing — how to run a three-to-five static set properly.
FAQ
Should I start with static ads or video? Statics, then promote the winners. Statics are faster and cheaper to produce, so the same money puts more distinct messages into the market and you learn faster. Build three to five carrying genuinely different angles — not three colourways of the same claim, three different arguments. Then turn the proven ones into video, where a slideshow-style edit starts from a known-good message rather than a guess. The mistake is running the sequence backwards and spending the production budget discovering the message was wrong.
Why group formats by the job they do rather than by layout? Because grouping by layout produces nothing useful — a two-column grid can be a comparison, a feature set or a review wall. What predicts where a creative belongs and whether it survives a cold audience is the kind of argument it is making, and advertising has no standard vocabulary for that, so this page builds one. The 43-archetype catalogue is Bytown's own working taxonomy, not a standard, and the funnel placements are starting points rather than rules.
Which container change is worth the most? Uploading at 720p or better is the largest single lift in TikTok's set at +312% conversion, and it costs nothing. Read the rest of that data with care: it was published in December 2021 and is dated. A 2026 Socialinsider analysis of more than six million brand videos puts peak engagement at 15 to 30 seconds, which does not agree with the same chart's 21-to-34-second finding. Where two datasets five years apart disagree, the newer one is describing a platform the older one did not.
What does adding a vertical asset actually buy on Google? Per Google's own before-and-after data, about 35% more conversions on Demand Gen, 40% higher view-through rate on video view campaigns, and about 45% better viewability on video reach campaigns. Same footage, different crop.
Sources on this page
- The 43-archetype catalogue is Bytown's own. It is a working taxonomy, not a standard, and the funnel placements are starting points rather than rules.
- TikTok Marketing Science, "Creative That Drives Conversions", December 2021. Machine-learning analysis of thousands of auction ads. Dated; read with the 2026 comparison noted above.
- Socialinsider, 2026 analysis of more than six million brand videos, on video length against engagement.
- Google Ads Help, vertical asset performance, Google internal data, global, 2024. Before-and-after comparison on adding a vertical asset.
