The best ad was already in the account
Before generating anything with AI, I listed a client's old ad videos. The footage that beat everything was unpolished phone video nobody was using.
A heritage snacks brand asked for new ad creatives, and my instinct was the modern one: generate. Write prompts, spend credits, produce something polished.
Before doing that, I made one API call and listed every video and image the ad account had ever uploaded. Twenty-five videos, sixty-five images, accumulated over eighteen months. Inside them: about three and a half minutes of real, brand-approved, vertical footage with sound — sitting unused while the account ran catalogue ads.
Then I pulled the spend data, and the finding got sharper.
The ad that was quietly winning
The single ad taking over 99% of the winning campaign’s budget was a sixteen-second phone video. Not produced. Not colour-graded. Real packets on a real counter, shot by someone standing in front of them. The copy underneath it was one line about childhood.
The most valuable asset in a 40-campaign account was first-party footage that cost nothing, while money went out monthly for polished work that performed worse.
Nostalgia is not heritage
The framing difference is worth being precise about, because both ads drew on the same brand history and pointed in opposite directions.
The brand’s own ads sold heritage — decades of trust, the company’s age, its credentials. The winning ad sold nostalgia — the customer’s own childhood: walking to the shop with pocket money, the two-rupee packet. Heritage is the brand’s résumé. Nostalgia is the viewer’s memory. One asks for respect; the other returns something the viewer already owns.
Why it wins mechanically, not just emotionally
The nostalgic ad collected comments and shares. The catalogue ads collected clicks.
That difference shows up in the auction before it shows up anywhere else. In the same account, in the same month, the shareable campaign paid a ₹55 CPM and the catalogue campaign paid ₹144. Our own polished ads had the better click-through rate — 3.75% against 1.65% — and were shared approximately never, so we paid multiples per thousand impressions to reach the same people.
The conclusion I keep coming back to: attention is priced before the click. Click-through rate tells you the ad works on the person who sees it. Shares decide how much you pay for the seeing.
The practice
Two habits came out of this, and I now apply both before any creative work:
- Audit the account’s own media library first. One API call. The footage a brand has already approved, already paid for, and already tested against a real audience is a better starting point than anything a model will invent.
- AI generation has to earn its place — it’s for the scene that doesn’t exist and cannot be photographed. Reproducing a real product it has never seen is exactly what it is worst at.
The least glamorous creative strategy I know is also the one with receipts: find what people already shared, and make more of that.