Black Friday ads for electronics and furniture: Sume Formats

For electronics and furniture, two Sume catalog Formats do most of the work: sume-product-usage-demo shows a product being used, and sume-before-after builds a reveal. Adobe's forecast, reported 2026-09-29, puts electronics at $63.3 billion, up 5.9%, and furniture at $33.4 billion, up 7.3%.

Match the Format to the product

The catalog describes these Formats by job, not by industry, so the match below is a suggestion to test, not a claim from the docs.

What to supply

Give the Format real product photos through attachments and describe the use in instruction: who uses it, where, and the one benefit to show. For furniture, supply a photo of the empty room for a before-and-after, and keep the instruction clear that the furniture's size and color must not change.

Generated video can get proportions wrong. For furniture, compare the output to the real dimensions, since a sofa that looks smaller than it is will lead to returns.

Run it safely

Start with one product, set generation_spend_cap_usd, use an Idempotency-Key per product and review the artifact. The receipt's usage.debited_usd_micros gives the cost to compare across Formats. When a recipe works, reuse the same instruction text for the next item.

The category numbers are Adobe's forecast for the US, and they say nothing about demand for your product. Use them to time your production, not to size your budget.

Pick one hero per category

Resist the urge to make ten versions of every item. Choose one hero product per category, make a strong clip for it and spend the remaining effort on the checking step. A short list of clips you have reviewed beats a long list you have not, and the receipts from the hero runs tell you what the rest of the range will cost.

Keep a record

Electronics and furniture ads are judged on accuracy, so the notes you keep matter. Keep a short log beside the files: the Format slug, the exact instruction, the Idempotency-Key and the final cost from the receipt. When something works, you can repeat it on demand, and when something fails, you can see what changed between the two runs.