# A practical creative testing workflow: brief,… — Cut Marketing > Plan one creative question, preserve the brief, review generated media and compare advertising reports without mistaking missing data for zero. [Home](/)/[Blog](/blog)/Creative testing # A practical creative testing workflow: brief, approval and observed results Plan one creative question, preserve the brief, review generated media and compare advertising reports without mistaking missing data for zero. 2026-11-03 4 min read Cut Marketing Team ![Original AI-generated green bottle study used to illustrate a creative brief](/samples/iterframe-field-bottle.webp) A useful creative experiment leaves a record of three things: what you intended to change, what you actually published and what the report observed. A folder of generated images covers only part of that job. Cut Marketing connects those records. Its current workflow includes product and brand briefs, generated creative, review decisions, an approved export pack, CSV or manual results, and linked follow-up experiments. Advertising delivery and spend remain in your ad account. ## Start with one question and a saved brief Write the product description, intended audience and the benefits you can support. Add actual product references when the selected model supports them. Keep the brand voice, visual rules, approved claims and prohibited wording explicit. For a reusable bottle, a question could be: “Does an opening question lead to lower observed cost per product-page visit than a product introduction?” This is a hypothesis to test, not a promised result. Choose the opening hook, opening visual or call to action as the brief element to vary. Creating two treatments is often easier to review than creating six before the question is clear. Cut Marketing supports two to six variants; this is a product limit, not a recommendation about statistical power. The experiment keeps a snapshot of the product and brand brief. Later edits to the product do not rewrite that record. ## Separate media from ad copy Inspect the script and the visual generation prompt before generating. Cut Marketing provides copy for the ad separately from its generated image or video. A generated visual is not a guarantee that the exact script, caption or headline appears in the media. For a copy comparison, use the same approved visual across treatments where possible. The reuse action links the existing output without another generation charge. Add each treatment’s exported copy when configuring the ads. Regenerating a different image for each hook can introduce visual differences beyond the copy question. For approved still images, [Compose image ad](/blog/exact-ad-copy-on-ai-images/) can place the exact hook, optional body and CTA into a downloadable PNG. Video exports continue to provide the original media and separate copy. For a visual comparison, inspect the actual outputs for unintended changes to the product, framing or context. Keeping a model and aspect ratio constant does not make generative outputs identical. ## Approve the exact render you plan to use Review product shape, packaging, readability, people and hands, and whether the scene implies an unsupported claim. Request changes if something is wrong. A successful generation means a file was produced; it does not mean the creative is ready to publish. Each approval in Cut Marketing refers to a particular completed render. Editing the treatment or generating another output resets its approval. The export pack becomes available once every current variant has been approved. The pack contains the media, copy, variant IDs, ad names, a results template and the saved brief. When a product page was supplied, its links include campaign and variant UTM parameters. Google documents manual campaign tagging in its [campaign URL guide](https://support.google.com/analytics/answer/10917952?hl=en). Keep the variant ID in the advertising platform’s ad name so a later CSV can match it. ## Import one comparable report Use one row per variant for the selected period. Cut Marketing accepts CSVs with English or Korean headings and can match an exact variant ID, an ad name containing that ID or a unique variant label. It previews matched rows and reports skipped unrelated rows before saving. Keep these details with the report: - Reporting period and time zone. - Currency, without an assumed exchange-rate conversion. - Source and attribution window. - Spend, impressions and clicks. - Conversions and revenue when the source reports them. An empty purchase field does not mean no purchases happened. Keep it empty if it was not reported. Do not add two overlapping weekly exports together. Cut Marketing stores each import separately and displays one report at a time. CTR is clicks divided by impressions. CPC is spend divided by clicks. CPA requires reported conversions above zero; ROAS requires reported revenue and spend above zero. Where a denominator or reported input is missing, a rate may be unavailable. These calculations describe the imported observations. ## Turn an observation into the next test A lower observed CPA or CPC can identify a treatment worth investigating. It does not isolate the effect of creative from audience, delivery, sample size or attribution. Google’s [attribution overview](https://support.google.com/analytics/answer/10596866?hl=en) explains how attribution models allocate credit along a path to an important action. Keep that reporting context when interpreting a comparison. Cut Marketing labels sparse or incomplete comparisons and does not declare causal significance. From a reported candidate, create a follow-up experiment that retains the approved treatment and the original brief. The parent experiment remains linked, so the next decision has a traceable starting point. The follow-up carries the actual approved starting render and fixes its generation settings. Copy alternatives can share that image without another generation charge; visual alternatives still need their own renders. Every treatment in the new experiment needs review before the full pack can be exported. [Start an experiment](https://hookframe-app.ksj00073.workers.dev/workspace) or [browse original creative examples](/showcase/). ## Keep reading [![AI-generated pale mint and plum compact camera with a cream woven strap](/samples/studies/daylight-camera.webp) Production studies ### Product image to video: a camera study Follow a fictional camera from an Indigo image to MiniMax H3 motion and a SeedVR2 1080p file. Inspect exact prompts, full-frame videos and measured delivery. ](/blog/product-image-to-video-camera-study)[![AI-generated sage ceramic coffee mug in warm sunlight on pale stone](/samples/studies/cloud-mug.webp) Creative production ### AI product video upscaling: review the source before 1080p export Inspect two actual MiniMax H3 and SeedVR2 product-video pairs. Compare full frames, measured dimensions and generated details before approving an upscale. ](/blog/ai-product-video-upscaling)[![AI-generated fictional green steel bottle on blue stone, beside curved ivory paper with open space on the left](/samples/field-bottle-wide-nano.webp) Creative production ### How to review an AI product video before export Use an actual bottle study to check product shape, camera movement, exact ad copy and approval. A practical review before handing off an AI product video. ](/blog/review-ai-product-videos) ## Make the next ad in minutes. Start with a product image or a short brief. The studio shows the credit cost before you generate. [Open studio](https://hookframe-app.ksj00073.workers.dev)