An ecommerce video ad strategy is a rule about what changes between funnel stages, not a rule about how many videos you own. Prospecting creative has to earn attention from someone who has never heard of you, so it opens on a problem or a result and holds the offer back. Retargeting creative talks to someone who already saw the product, so it opens on the objection and puts the offer in the first three seconds. Run 8 to 20 variations live per active campaign, weighted heavily towards prospecting, and refresh on a cadence set by your category rather than by the calendar.
How should creative change by funnel stage?
The mistake is running the same 30 second product film at every stage with a different audience attached. Each stage is a different conversation, and the opening, the length and the position of the offer all move.
| Stage | What the viewer already knows | Opening (first 2 seconds) | Length | Where the offer goes | Live variations |
|---|---|---|---|---|---|
| Cold prospecting | Nothing | Problem, result or an unusual visual. No brand, no logo | 6 to 15s | Last 2 seconds, or omitted entirely | 6 to 12 |
| Warm interest (video viewers, engagers) | Recognises the product category | The specific use case, named out loud | 15 to 25s | Around the 60 percent mark | 3 to 5 |
| Retargeting (viewed product, no add to cart) | Saw the product page | The objection: price, fit, sizing, shipping, returns | 8 to 15s | First 3 seconds | 3 to 6 |
| Cart abandoners | Chose a variant, hesitated | The product itself, in use, plus the specific reason to finish | 6 to 10s | First 2 seconds and repeated at the end | 2 to 4 |
| Returning customers | Owns it | The adjacent product or the refill logic | 10 to 20s | Middle, framed as a reorder not a discount | 2 to 3 |
The pattern is that certainty buys you time. A cold viewer will not sit through a setup. A cart abandoner already wants the thing, so the discount code can go on screen at second one without wasting the hook.
One thing worth being blunt about: a logo in the first two seconds of a prospecting ad is a wasted second. Nobody in a cold audience is looking for you. Put the logo where the viewer already cares, which is after the result has landed.
Why one creative per product stops working
Catalogue-shaped thinking produces one video per SKU, and it fails for a structural reason. A single creative per product means every product's performance is a single point of failure, and when that creative fatigues the whole SKU goes dark. There is nothing to rotate to.
Our 2026 fatigue benchmark, which is a synthesis of published platform and agency figures rather than our own experiment, puts CTR decline at 15 to 20 percent in the first two weeks, then a cliff in week three of 45 to 70 percent below launch. Meta internal research in the same report shows CTR dropping 45 percent after a fourth exposure to the same creative. A one-creative-per-SKU account hits that fourth exposure in days on a broad audience.
The second failure is diagnostic. If a SKU underperforms on one video, you have learned one thing about one video, not anything about the product. Three arguments running at once tell you whether the problem is price, proof or audience.
How many variations per SKU?
Work backwards from spend, not from catalogue size. A SKU carrying real prospecting budget needs 4 to 6 distinct arguments in rotation. A long-tail SKU carrying a few hundred euro a month needs one or two, and should mostly be sold by the winners of the hero SKUs' tests.
Concretely, for a store with 40 products where 6 drive most of the revenue:
- 6 hero SKUs at 4 to 6 arguments each: 24 to 36 variations
- 10 secondary SKUs at 2 arguments each: 20 variations
- 24 long-tail SKUs: cover with 3 to 4 category-level ads, not per-product ads
That is roughly 50 to 60 live files, not 40. If you want to size a set against your own spend and refresh interval instead of guessing, the ad variation calculator does that arithmetic.
The refresh cadence is the part most stores get wrong, and it is category-specific. Days to a 40 percent CTR decline in our benchmark:
| Category | Days to a 40 percent CTR decline | Practical refresh |
|---|---|---|
| Food and beverage | 9 | New creative into rotation weekly |
| Fashion | 12 to 14 | Weekly on the hero SKUs, fortnightly elsewhere |
| Beauty and DTC | about 18 | Fortnightly |
| Electronics | about 21 | Every two to three weeks |
| B2B SaaS | about 28 | Monthly |
A food and beverage brand and an electronics brand can run the same funnel structure and still need completely different production rates. Nine days is the fastest measured figure in the report, and it means a snack brand needs new arguments in rotation before most teams have finished briefing the last batch. The report's own operational conclusion is that throughput, not talent, is the bottleneck: the number of unique concepts shipped per month predicts campaign longevity better than the quality of any single ad.
What about seasonal and promo creative?
Treat promo creative as a separate set with its own lifecycle, not as an edit of your evergreen ads.
Two rules that hold up. First, a promo ad and an evergreen ad should not share a hook. If your Black Friday creative is your best evergreen creative with a price sticker on it, you have fatigued both at once, and the evergreen version comes back in December already burned. Second, build the promo set before the window opens: a two week fashion sale outlasts the 12 to 14 days it takes creative to lose 40 percent of its CTR, so a single sale ad decays inside its own promotion.
A workable split is 70 percent of promo budget on the offer stated plainly, 30 percent on the product argument with the offer as a closer. The second group is what you keep after the sale, because it still works at full price.
Where the footage for all of this comes from
Fifty live files at a fortnightly refresh is a lot of editing, and supply is where most stores stall.
This is what we built Genyad for, so treat the next paragraph as a product note. You upload the footage you already own once, it transcribes and tags every clip, and each variation is a fresh script, shot selection, voiceover, caption set and export drawn from that library rather than a re-cut of one timeline. One standard variation is one credit, editing and re-exporting cost nothing, and exports come in 9:16, 4:5, 1:1 and 16:9 at 1080p with no watermark. Our ecommerce video ads page has more on the approach.
If a SKU has photography but no video, stills can be animated into short motion clips with our image to video ad generator at 2 credits per second. A 6 second clip is 12 credits, which is worth spending on a hero product and not worth spending across a 400 item catalogue.
Two things Genyad does not do, because they matter to this strategy. It does not import products from a URL and it does not render templates from a product feed or CSV, so a per-SKU automated build across a large catalogue is not what it is for. If feed-driven per-product rendering is the actual requirement, a template renderer like Plainly is the right shape of tool. Genyad also does not publish to Meta or TikTok directly, so uploading and structuring the campaign stays your job.
Frequently asked questions
How many video ads does an ecommerce store need live at once?
Most active campaigns need 8 to 20 variations in rotation. For a store, that count is per campaign rather than per catalogue: concentrate 4 to 6 distinct arguments on the SKUs carrying real budget and cover long-tail products with category-level ads instead of one video each.
Should retargeting ads be shorter than prospecting ads?
Usually yes, and the reason is the offer position rather than attention span. A retargeting viewer has already seen the product, so the persuasion job is narrow: handle one objection and state the offer immediately. That fits comfortably in 8 to 15 seconds, where a cold prospecting ad often needs the setup.
How often should ecommerce creative be refreshed?
By category. Our benchmark report puts days to a 40 percent CTR decline at 9 for food and beverage, 12 to 14 for fashion, about 18 for beauty and DTC, and about 21 for electronics. Set the refresh interval just inside that number for your category rather than using one cadence across the account.
Can I generate video ads from product photos if I have no footage?
Yes, within limits. Animating stills into motion clips costs 2 credits per second in Genyad, so a 6 second clip is 12 credits. That is fine for a hero product and not economical catalogue-wide, and real footage of the product in use still outperforms animated stills in prospecting.
Does a promo ad count as a new variation?
Only if the hook changes. Adding a discount overlay to an existing winner does not reset its fatigue curve, because the viewer reacts to the same opening and the same claim. Build promo creative with its own hooks so the evergreen version survives the sale.