Across our agency clients, the average spend per Meta ad has fallen 74% since 2025. On TikTok, it’s down 69%. In most cases, budgets haven’t reduced. Ad volumes have roughly quadrupled.
How we got here
Meta’s Advantage+ and TikTok’s Smart+ campaigns are built to ingest as many creative variants as possible. More genuine variation gives algorithms more material to work with, and the platform logic is reasonable. The challenge sits at the brand end.
Algorithms reward creative input, so teams produce more. Generative AI has accelerated this considerably: resizing assets that once took days now take hours. Brands that shipped 10 creatives a month are now shipping 80. Creative volume has become a proxy metric, measurable and increasing, and easily mistaken for progress.
Before long, brands are managing libraries of hundreds of active ads, each backed by a fraction of what it needs to prove itself.
The signal problem
Meta’s own research puts creative as the driver of 56% of ad performance, ahead of targeting and bidding strategy combined. But creative only delivers that impact when it gathers enough signal to be understood.
Meta requires 50 optimisation events at ad set level before the algorithm stabilises. With 20 creatives sharing that budget, each asset gets roughly 2-3 purchase events of data. At a £60 CPA, reaching that threshold for a single creative in isolation would require £3,000 of spend behind it, the average ad is getting a tenth of that.
Meta labels creatives within Advantage+ campaigns accordingly: Best, Good, Learning, Low. In fragmented accounts, most assets sit in Learning indefinitely, not because the creative is ‘wrong’, but because it never gathered enough signal to reach a verdict.
The commercial cost
The downstream effects are predictable. Performance feels volatile because the data underpinning it is thin. Creative teams spend more time producing and less time understanding what actually works. Real winners get buried in a portfolio too large to read clearly.
YouTube offers a useful comparison. When an asset costs real money to produce, brands are more deliberate about what they make and more committed to making it work. They run fewer creatives, put proper budget behind each one, and make clear decisions about what stays and what gets replaced. The result is better signal and more confident optimisation.
What good looks like
The brands navigating this well operate creative as a system. Each sprint, they generate insight, build strategy around it, create with genuine diversity, measure with enough budget to get clean data, and optimise based on what actually happened. The system gets smarter over time because every decision feeds the next one.
Their frameworks answer four questions: what gets tested – a small pool of ideas that are meaningfully different from each other; what gets enough budget to be judged fairly; what gets killed when it isn’t working; and what gets scaled decisively when it is.
If your spend-per-ad trend is heading in the same direction, it’s worth understanding why. It may explain more about your results than any individual campaign decision you’ve made this year.
