Built for the Feed, Not the Shoot

The Brief

What the Meta Feed Was Really Asking For

The shift inside Meta over the last two years is not subtle, but its implications are easy to miss. The platform has steadily moved away from rewarding the single best-performing creative and towards rewarding the brand that can sustain the highest volume of variations in market at any given time. Meta’s own creative guidance now centres on creative volume, hook diversity, and continuous variant testing as primary drivers of campaign performance.

For a brand operating at Sparx’s pace, this turns into a simple but uncomfortable math problem. The feed wants twenty creatives this week, not one. It wants ten hook variations, not three. It wants the next test live before the previous test has finished resolving.

The question Sparx’s team was facing was not whether their creative was good. It was whether their creative production model could move at the speed the feed was now demanding.

The Problem

Why Production-Led Creative Couldn’t Keep Up

A shoot-led creative cycle is an extraordinary thing. It produces work with depth, texture, and intention that nothing else quite replicates. It is also, by design, a cycle built for one or two big executions per campaign, not for the twenty variants the algorithm now wants to test inside a fortnight.

Three structural realities make the gap difficult to close inside a traditional production model. Output is slow because every variant needs the same set, talent, and post-production cycle. Testing is limited because each test is expensive enough that brands self-edit which variants are worth shooting. And competitiveness erodes because the brands building creative on a different cost-and-time curve start to dominate the variant economy on the platform.

This is not a failing of any team. It is a structural mismatch between an old production model and a new platform reality and Sparx’s team named it first.

The Strategy

We Stopped Optimising the Shoot. We Replaced It

Working alongside Sparx’s digital and D2C team, we did not try to make the shoot faster, cheaper, or more efficient. We replaced it as the primary engine of creative production.

The system we built together is AI-led at the point of creative generation. Instead of designing campaigns around what could be photographed, we designed them around what could be imagined, generated, iterated, and tested at the speed the feed rewards.

The Three Principles the Engine Was Built On

Three principles sat under every decision.

01. Speed over production dependency: an engine that could produce a new creative in hours instead of weeks was structurally more valuable than one that produced a better single creative in months.

02. Visual impact over realism: on a feed where attention is won in the first two seconds, an exaggerated, surreal, scroll-stopping frame consistently outperformed a polished but conventional one.

03. Volume and variation over single executions: the brand that tests fifteen hook variations will out-learn the brand that tests three. Every quarter. On identical spend.

Inside the AI-Led Creative System

The engine produced six distinct creative formats, each calibrated for a specific moment in the feed.

• Hyper-Reality Motion put Sparx footwear in gravity-defying, physically impossible movement engineered to break scroll inside the first frame.
• Impossible Product Moments built surreal scenarios where the product appeared, transformed, and interacted in ways no shoot could replicate.
• Product-in-Action in Amplified Worlds layered real product usage walking, styling, daily wear over cinematic, exaggerated environments, holding relatability and aspiration in the same frame.
• Hook-First Videos were designed around their opening two to three seconds, where the entire creative decision had to be earned.
• High-Variation Concepts produced multiple creative routes across visuals, hooks, and narratives so testing could run continuously.
• Feed-Native Edits were built short, fast, and mobile-first, designed for the way Meta actually serves and rewards content.

What 4.2% CTR Actually Means

The system, in market, delivered a 4.2% CTR for Sparx on Meta a number that earns its weight in context. It is not the result of a single high-performing creative. It is the result of an engine that could sustain enough variation, at enough volume, at enough velocity, for the algorithm to find the winners on its own.

AI turned content from a bottleneck into a growth lever.

“Meta ads is a constant cycle of testing and learning. Consistently generating creatives has helped in finding winning creatives, AI just made the process faster and easier to scale.”

Vishal Pathania · Head of Digital and D2C, Sparx (Relaxo)

The Funnel That Made the Number Possible

The headline number is a system outcome. Underneath it sits a funnel built deliberately around what each stage of the audience actually needed.

• Top of funnel – Discovery. High-impact AI videos built to grab attention and earn the first scroll.
• Middle of funnel – Consideration. Product-in-use visuals, leaning into styling, comfort, and everyday wear to convert curiosity into intent.
• Bottom of funnel – Conversion. Offer-led creatives carrying clear pricing and a direct call to action.

Same creative engine. Three distinct creative jobs. One coherent performance system

The Horizon: Volume Is the New Creative Metric

For most brands, the question of AI in creative is still phrased as a choice whether to use it, whether it is ready, whether it is on-brand. Sparx’s engagement reframes the question entirely. The brands building AI-led creative systems now are compounding two advantages simultaneously: the cost-and-time advantage of generating creative at scale, and the learning advantage of testing more variants per quarter than any production-led brand can match.

A year from now, that gap will be difficult to close without rebuilding the production model itself.

The Sparx system continues to evolve new formats, new hooks, new applications of the underlying engine across the brand’s broader Meta presence. The frontier has moved. The work continues.