If YouTube Shorts is part of your media plan, do not buy “one vertical cut.” Buy a system of vertical assets that can be recombined, deployed, and compared. A 9:16 export solves file geometry. It does not decide how the opening earns attention, when the product appears, how the brand enters the story, which proof shot can be swapped, or whether the call to action works with the Shorts interface.
That distinction matters for global brands working with manufacturers in China. The most efficient time to create testable options is while the product, set, talent, and camera are already together—not after the master film has been locked.
The platform rules already point to a production decision
Google Ads’ current Shorts asset guidance recommends vertical video and says 9:16 assets are prioritized for Shorts brand campaigns. Landscape and square assets can still serve, but a landscape video may appear with blurred space above and below.
There is no single universal Shorts duration. Google lists 6–60 seconds for Video Reach campaigns and 10–60 seconds for formats including Video Action, App, and Performance Max. Longer uploads are possible, but only the first 60 seconds play in the Shorts feed.
The interface timing also changes by campaign type. The Shorts call-to-action button can appear at three seconds for Performance Max, App, and Demand Gen campaigns, and at 10 seconds for Video View and Video Reach campaigns. That means the reveal, brand cue, and verbal close should not be designed without knowing the media setup.
A useful production brief therefore starts with the campaign objective and the question to be tested—not merely “9:16, 15 seconds.”
Automation can change the frame; it cannot define the meaning
Google Video Enhancements can generate vertical and square versions from a landscape asset. This is useful for filling format gaps and creating candidates quickly, particularly when a brand already owns clean, well-composed footage.
But automated reframing cannot reliably know which hand movement proves the product, which expression lands the joke, which subtitle must remain clear of the interface, or whether the brand should enter at second four or second nine. Geometry is not narrative.
A 2026 Wolt Shorts experiment published by Think with Google tested brand introduction at four, nine, and 12 seconds alongside different humour treatments. The result should not be generalized into “every brand must appear at second four.” The transferable lesson is that brand timing was treated as a controlled variable, rather than a subjective argument after the edit.
Platform automation is well suited to format adaptation and candidate generation. Human teams still own product truth, performance, pacing, cultural judgment, and brand meaning.
The ONCE 5×3 Shorts procurement matrix
The “5×3” framework does not mean buying 15 finished videos. It means defining five variable families and requiring three layers of delivery, so one production can support repeated learning.
The five variable families are:
- Hook: Prepare at least two openings that can stand alone—an action, tension, contrast, question, or result.
- Product proof: Capture operations, materials, before-and-after states, or use situations as independent shots, not only as fragments inside one master edit.
- Brand entry: Let packaging, sound, a recognizable product shape, or a character introduce the brand at testable moments; do not rely only on a corner logo.
- Reason modules: Film each major selling point as a compact block that can be recombined for different audiences.
- Action path: Coordinate picture, voice, captions, and the platform CTA instead of adding a generic purchase line in the final second.
The three delivery layers are:
- Reusable source library: Clean audio, text-free masters, safe framing, and consistent asset names.
- Modular edit masters: Swappable hook, proof, branding, and CTA sections rather than one irreversible timeline.
- Test-ready exports: Campaign-appropriate durations, safe captions, sound variants, version IDs, and a variable manifest.
For brands producing near their supply chain, ONCE can connect this planning to video production workflows in China, so product access and technical review happen before footage leaves the factory region.
Put the variable system into the brief and contract
A workable Shorts brief should specify six items:
- Objective and campaign type: reach, views, action, or a multi-objective plan;
- Native composition: the 9:16 hero frame, eyelines, hand actions, product visibility, and interface-safe zones;
- Variable plan: whether the round isolates one variable or deliberately tests a combination;
- Duration set: which 6–60 or 10–60 second versions are primary and which are exploratory;
- Automation permissions: whether the platform may crop, shorten, add voice-over, or create new variants;
- Measurement identity: how each version ID will appear in the ad account and review sheet.
Acceptance should go beyond resolution. Preview every version on a real phone. Check that the decisive product action remains visible, captions avoid interface controls, the core message survives sound-off viewing, sound adds meaning when enabled, and the visual is ready when the CTA appears.
A compact ID such as H2-P1-B4-C1 can identify the second hook, first proof module, brand entry at four seconds, and first action path. Naming is not bureaucracy. It is what allows media results to inform the next production decision.
If the same footage also needs to support product questions or AI-assisted shopping, the related ONCE analysis on preparing product video for YouTube shopping agents explains how visual proof and structured product facts can work together.
Evaluate cost per testable hypothesis, not cost per file
Instead of comparing only “price per video,” use a more decision-relevant internal measure:
Cost per testable creative hypothesis = attributable production cost ÷ independently deployable, clearly identifiable hypotheses.
A hypothesis counts only when the version can run on its own, the changed variable is clear, and the result can be traced back to that variable. Ten bitrate exports are not ten hypotheses. A caption-colour change may not deserve a separate test.
If the budget is small and Shorts is a temporary placement, automated reframing may be sufficient. If the product involves complex actions, talent, multiple markets, or ongoing paid media, designing a 5×3 asset system before the shoot is usually more controllable than repeated rescue edits and reshoots.
The cheapest production is not necessarily the one with the lowest price per finished file. It is the one that turns each shoot day into more usable, testable, and reusable creative choices. Share your campaign objective, product claims, and target markets with ONCE to receive a Shorts asset plan.
Production examples and next steps
Turn the checklist into a brief with clear assets, decisions and delivery versions. Compare the realme GT7 product film as a finished-work reference.
Review Video production services. From brief to final film. and realme GT7: Graphene Ice-Cooling Body Product Demo. To discuss your project, send the current assets, decision owners, required versions and review schedule through the English project enquiry form.
