The single most expensive creative mistake in ecom paid is over-investing in production polish on too few creatives. One $20K VSL running for six months loses to eighteen $1,500 creatives running for two weeks each, every time. The audience saturates on a single creative inside 14-21 days at meaningful spend.
ad units per two-week batch is the cadence that compounds
Six scripts, three hook variants per script
The compounding pattern is six scripts × three hook variants, shipped every 14 days. Each batch teaches the platform new audience patterns. Top performers from each batch get re-cut into derivative variations the next batch. The creative cadence is the asset, not any single creative.
Why velocity beats polish
Performance Max consumes creative the way Smart Bidding consumes conversion data: it needs variety to learn what works for which audience. A single high-polish creative gives the model one data point. Eighteen rougher creatives give the model eighteen data points across angles, hooks, and treatments. The model gets smarter faster on volume.
Past a baseline of competence (lit decently, audio clean, talent watchable), the eighteen-unit batch outperforms the polished singleton on cost-per-conversion by 30-50% in the accounts we've measured. Production polish is real, but it's a second-order optimisation. Variety is first-order.
When polish does matter
Brand-building creative for high-AOV verticals (jewellery, furniture, premium electronics) does benefit from polish because the buyer is judging the brand, not the offer. But even then, you ship the polished anchor creative AND the eighteen-unit batch. The batch is what feeds the model.
The six-script framework
Each batch needs six distinct scripts so the model can find which angle resonates with which audience. We run the same six categories on every account, regardless of vertical, and let the data sort which script wins for the brand.
Each script gets three hook variants (3-second openers). Total 18 ad units per batch.
| Script type | What it does | Best on |
|---|---|---|
| Demonstration | Shows the product in action against a clear before/after | YouTube in-stream + Demand Gen feed |
| Transformation | Customer journey from problem to result | YouTube long-form + Demand Gen story |
| Founder story | Why the brand exists, what it solves | YouTube long-form, brand-building |
| Customer proof | UGC-style testimonial with native handheld feel | Demand Gen feed + YouTube Shorts |
| Comparison | Side-by-side vs the obvious incumbent | YouTube in-stream + Search-paired Demand Gen |
| Educational | Quick explainer on the category itself, brand soft-mention | YouTube in-stream skippable + Demand Gen feed |
The three-hook variants
Each script gets three opening hooks. The first three seconds of any video ad determine whether it gets watched. Three hooks per script lets the model find which opener works for which audience without rewriting the body.
- Pattern interrupt: visual or audio surprise that breaks the scroll. Works on cold feed audiences.
- Direct address: "If you've ever struggled with X, this is for you." Works on intent-warm audiences.
- Curiosity gap: poses a question the body answers. Works on educational + comparison scripts.
The 14-day production cadence
Days 1-3: script + hook ideation
Six scripts written, three hook openers per script. Total 18 ad units in the production queue. Brief the talent and props in parallel.
Days 4-9: production
In-house or partner studio. The same talent and props can produce 18 units in a week if pre-planned. Same shoot day; different angles and hooks.
Days 10-12: edit + delivery
Each unit cut to YouTube specs (16:9 + 9:16 vertical) and Demand Gen specs (square + portrait). Captions burned in for sound-off compatibility.
Day 13: campaign upload
All 18 units uploaded into asset groups across YouTube, Demand Gen, and PMax video. Budget allocated to discovery; Smart Bidding decides the winners.
Day 14: review last batch
Top 4 ad units from the prior batch get re-cut into 6 derivative variations for the next batch. The system compounds: every winner produces multiple variations.
Don't run a creative more than 3 weeks
Even winners decay. Past 21 days at meaningful spend, even your top creative loses 30-50% of its day-1 efficiency to audience saturation. Always ship the next batch before the previous batch decays. Cadence is the asset, not any single creative.
What this costs and what it returns
Eighteen creatives every two weeks at $1,500 each is $39K/month in production. That's the upper bound; in practice we see brands run this for $20-25K/month using in-house talent and a partner studio for editing. The return: top-of-funnel scale that's otherwise unavailable. Most accounts can't push past $50K/month on Google + YouTube without this cadence because the creative side starves before the bid side.