Facebook Ad Creative Strategy: The Optimal Meta Media Mix for ROAS and Reach

Every recommendation here is tied to a number, from our own account data or from Meta's published guidance. By the end you will know what to produce, in what ratio, how many pieces to run for your budget, and how to judge the mix once it is live.

How we ran this study

Two data sources sit behind every number in this article.

Our own accounts: we measured the trailing 365 days of live spend across our active Meta accounts at the individual-ad level, classifying each ad as video or static by whether it recorded video plays, then aggregating spend, return on ad spend, click-through, CPM, cost per thousand reached, and ThruPlay by asset type. About 60 percent of those accounts pass clean purchase data back to Meta, so we report return on ad spend only for those and judge the rest on reach and engagement.

The 83-brand study: we analyzed the live Meta Ad Library entries of 83 leading DTC brands across food and beverage, supplements, pet, beauty, apparel, and hardgoods. For each brand we separated produced creative (video, image, carousel) from catalog and dynamic ads, because the Ad Library shows every product variant as its own card and that inflates catalog counts. The Ad Library only shows currently active ads, so this captures what leading brands are running right now. It does not include paused or historical creative.

Two limits to keep in mind. The Ad Library has no spend or performance data, so brand-level figures only tell us what each brand keeps live. Cost per thousand reached overstates unique reach when summed across many ads, so we use it only to compare asset types within one account.

Finding 1: leading brands run video-led produced creative

Among produced creative across the 83 brands, video outnumbers static image by roughly four to one, and carousel barely registers.

The split shifts by category, and that matters when you set your own target. Health and supplement brands run almost entirely video; apparel and hardgoods lean video for the demo; food, beverage, and beauty carry the most static of the categories we measured.

There is a second layer under all of this. Catalog and dynamic ads, which Meta calls Advantage+ Catalog or DPA, made up about 57 percent of all live ad cards across these brands. That number looks huge until you remember one dynamic ad shows up as dozens of product cards in the Ad Library. Treat catalog as a separate, always-on retargeting engine fed by your product feed. Run it in parallel and do not count it toward your produced-creative mix.

Finding 2: video scales, static holds its per-dollar edge

Video won click-through in nearly every one of our accounts and took the majority of budget almost everywhere. That is video doing its job: buying reach and clicks on cold audiences. Return per dollar looked different. Over the trailing year, static matched or beat video on return on ad spend in most of our largest, cleanly-tracked accounts, and the two were close in the rest.

The pattern holds across sectors. A specialty-food account returned 3.4x on static against 2.7x on video. A coffee account: 3.6x static, 2.5x video. A superfoods account: 12.0x static, 7.4x video. In each case video still carried more of the budget, because it was doing the prospecting, while static returned more per dollar.

For some brands the ranking flips. A personal care account and a premium wine account ran video-led and had video edge out static on return, and a couple of accounts came out dead even. The ranking is account-specific because video and static do different jobs: video buys reach and clicks, and static tends to convert at a better cost per result. A healthy account keeps both live and leans on each where it wins.

The practical rule: lead the budget with video because it scales, hold a static layer at about a quarter to a third of production because it is usually your cheapest cost per result, and move budget between them based on which is hitting your KPI.

The full picture: ten cleanly-tracked accounts

We then widened the return analysis to ten cleanly-tracked accounts over the trailing 90 days, about 1.3 million dollars of spend. The Finding 2 pattern held and sharpened. Video won click-through in all ten and took most of the budget in nine. Return per dollar leaned the other way, with a lean static-and-catalog layer matching or beating video in seven. Five accounts, picked to span the range, show the shape.

Two things keep this honest. Our classifier splits ads by whether they logged a video play, so the static column also holds carousel and catalog ads, and part of that per-dollar edge is catalog retargeting working as designed. And the two largest prospecting accounts, the seafood and oral-care brands, both had video win outright. Across all ten accounts the split is the same: video buys the reach and the clicks, a static-and-catalog layer defends your cost per result, and running only one of them leaves return on the table.

Finding 3: for awareness, the cheapest reach depends on the account

If your KPI is reach, brand lift, or retail support, return on ad spend is the wrong scorecard. The right metrics are CPM, cost per thousand reached (the closest Meta proxy for household penetration), frequency, and ThruPlay. On those metrics the picture is account-specific.

In one awareness account, static bought impressions at less than half the CPM of video ($1.78 against $3.58). In another, the two were about even and video was slightly cheaper. Static holds no attention though, because a still has nothing to play, so video carried all of the ThruPlay and the higher click-through in both. So an awareness buy uses both: buy reach on whichever format is cheaper for that account, and use video to hold attention and build the memory.

The framework: match the asset to the job

There is no universal ratio because the mix is a set of jobs, and each objective and funnel stage hires a different one. The mix shifts on two axes.

By objective, for a conversion goal: produce video-led with a deliberate static layer, tuned by category. These targets track the category video shares in Figure 2, adjusted for what returned best in our accounts.

By funnel stage: the lead format changes as the audience warms. Cold traffic leads with the formats winning click-through in our accounts, UGC and founder video, and warmer traffic shifts to the closers.

Within the video share, UGC video was the click-and-scale workhorse in our data, founder and testimonial pieces returned above their spend share, and product demos carried the hardgoods accounts. Within the static share, brief offer, testimonial, product-on-white, and before-and-after, because those become the long-running winners.

How many creatives, and how many for your budget

Two questions hide inside "how many creatives," and mixing them up wastes budget.

Per ad set: give the algorithm enough distinct concepts to explore without splitting your budget too thin. Meta's guidance is a standing rotation of 3 to 5 concepts per ad set, with at least one fresh asset every 1 to 2 weeks, per its creative fatigue recommendations. Distinct means a genuinely different idea. Meta's system groups near-duplicate creatives together, so a color swap or a one-word change does not add a new concept.

Per month, scaled to budget: Meta does not publish a number here, so the table below is our operating benchmark. New-concept volume should climb with spend, at roughly one new concept per $1,000 to $1,500 of monthly spend at smaller budgets and tapering as you scale.

Two of our own numbers back this up. Across our accounts, healthy creative programs kept about 30 percent of live creative launched within the trailing 30 days. And the strongest performers in the 83-brand study kept a small evergreen core, single winning creatives, some live for 300 to 950 days, running underneath a steady stream of new tests. New volume is how you find winners, and the evergreen core keeps you from rebuilding the account from scratch every month.

Refresh cadence and creative fatigue

Meta defines creative fatigue as the drop in response when an audience sees the same creative too many times (Meta Business Help). Meta's own analysis puts money on fixing it: adding fresh, diverse creative to fatigued ad sets improved conversion rate by about 8 percent on average in high-fatigue cases (Analytics at Meta).

The most common mistake we see is refreshing on frequency alone. Frequency is a lagging indicator. By the time it spikes, response has usually already slipped. Refresh on the compound signal: frequency rising, a clear drop in engagement from its first-week baseline, and cost per result climbing, all at once. Meta suggests adding a fresh asset every 1 to 2 weeks; as a tiered default we run top-of-funnel every 2 to 3 weeks, mid-funnel every 3 to 4, and retargeting every 4 to 6. Short-form video placements tend to fatigue faster than static feed, so watch them sooner.

Does creative diversity actually feed the algorithm?

A claim goes around that you should keep some of every asset type live, even ones not currently delivering, because their presence feeds the algorithm. A Meta rep may even tell you a version of it. The mechanism behind it is real, but the conclusion people draw from it is wrong, and acting on the wrong version costs money.

Start with what is true. Meta's ad-retrieval system, Andromeda, narrows tens of millions of eligible ads down to a few thousand candidates before the auction, and Meta describes handling the growing volume of creative as a core engineering problem it built the system to solve (Meta Engineering, 2024). Each genuinely distinct concept is another way for the system to match you to the right person, which is why diverse, active creative helps. Meta's guidance agrees: it recommends diversifying creative concepts to combat fatigue and convert new audiences.

Now the part people get wrong. The benefit comes from distinct, active creative that the system can actually test. A non-delivering asset earns nothing by sitting in the account. Worse, near-duplicate ads get grouped into a single entity, so piling on lookalike variants adds no new match surface. And when your own overlapping ads chase the same people, they bid against each other in the auction, which inflates CPMs and stacks frequency. Low-relevance ads also get throttled in delivery, so keeping them live does not feed anything. What actually helps is a set of distinct, active concepts across formats the system can test and choose from. If a rep tells you otherwise, ask for the specific Meta document first. What to actually do: keep 5 to 10 genuinely distinct concepts live per active campaign, pause any ad still spending past your target cost per acquisition without converting, and never keep a dead ad running just to have that format on the account.

Frequently asked questions about Facebook ad creative strategy

How to run it

Set the KPI first, because it decides the mix. For return on ad spend, run video-led with a real static layer, and give static enough budget to show its efficiency. For reach, widen with the cheaper-reach format and hold with video. Give the algorithm a handful of genuinely distinct concepts, scale new volume to your budget, refresh on the compound fatigue signal, and keep an always-on catalog layer underneath.