The short answer
With a small Facebook ads budget, spend less on volume and more on research. On our board of 2,443 winning Meta ads, 526 pages have only one or two winners, and those ads carry a median estimated spend of $244K, more than the ads of pages with ten or more. One good ad is enough. Static images with a problem/benefit headline reach the top tier 18% of the time.
The short answer: fewer concepts, borrowed from what already works
A small budget can't buy many tests. So each test has to start closer to a winner. That's the whole strategy, and the data on our board backs it up.
We looked at 2,443 live Meta ads from 749 advertiser pages, each with at least $50k in estimated spend. Then we compared the pages that have one or two winners with the pages that have ten or more.
526
pages with only 1 or 2 winning ads
$244K
median est. spend of their ads
18%
top-tier share, image ads with a problem/benefit headline
2.6%
top-tier share, AI-generated video
The top tier is Ad Radar's top 10% by winner score (live 21+ days), so about 10% is the average. Spend figures are estimates modeled on public engagement data.
One winner is enough
| Advertiser page | Pages | Ads | Median est. spend | Median days live |
|---|---|---|---|---|
| 1 or 2 ads on the board | 526 | 656 | $244K | 105 |
| 10 or more ads on the board | 47 | 941 | $220K | 135 |
Pages with one or two winners have ads that are just as big as those from pages with ten or more. The large advertisers have more ads, run them longer and reach the top tier more often (11% against 7%, partly because they launch more variants, which our winner score rewards). But the individual ad of a small advertiser isn't smaller.
What this means for a small brand: you don't need a creative machine. You need one ad that works and the budget discipline to keep feeding it. Once you have it, new hooks on the same body are cheaper than new concepts. See iterations vs new concepts.
If you have a cat older than seven, you'll want to hear this.
- Est. spend
- $1.9M
- Days live
- 308
- Format
- Video, 130s
The only ad from this page on our board. A fresh cat food brand, a voiceover over footage, sent to the homepage. Over $1.8M in estimated spend in about ten months.
The cheap formats that win
Production cost matters when the budget is small. Here's how the formats compare.
| Production format | Ads | Reach the top tier | Median est. spend |
|---|---|---|---|
| Footage + voiceover | 177 | 14% | $345K |
| UGC selfie | 78 | 13% | $592K |
| Founder or expert | 145 | 12% | $296K |
| Product demo | 97 | 8% | $301K |
| AI-generated | 77 | 3% | $229K |
| All image ads | 475 | 13% | $173K |
| All video ads | 1,968 | 9% | $236K |
Footage + voiceover is the standout for small teams. It needs stock or simple product footage and a script, no on-camera talent. It reaches the top tier more often than any other format with a meaningful sample. More in footage + voiceover ads.
AI-generated video is the trap. It's the cheapest to make and the weakest on the board, and it tends to get pulled sooner. The founder or expert format costs you nothing but time, if you're willing to be on camera.
Image ads: the best ratio of cost to result
Image ads reach the top tier more often than video (13% against 9%), though each one carries less spend. Split by opener, one combination stands out:
| Image ad opener | Ads | Reach the top tier | Median est. spend |
|---|---|---|---|
| Problem/Benefit Opener | 119 | 18% | $191K |
| Specific Number Opener | 68 | 13% | $147K |
| First-Person Story | 155 | 11% | $202K |
| Question Opener | 51 | 10% | $185K |
| Offer-First Opener | 64 | 9% | $139K |
A static image that names the problem or the benefit in the headline is the best cheap bet we can measure. You can make five of them in an afternoon. See static image ads.
Breathe cleaner air, noise-free and filter-free
- Est. spend
- $9.6M
- Days live
- 127
- Format
- Image
- Variants
- 7
A static image whose headline is three benefits in six words. Past $9.6M in estimated spend with 7 variants. No video, no creator.
The Nurses' Favorite Shoes of 2024
- Est. spend
- $1.3M
- Days live
- 631
- Format
- Image
- Variants
- 3
An image with a nurse's testimonial as body copy, sending traffic to a listicle. Live for 631 days. The headline names the audience, the body does the selling.
Borrow the structure, not the ad
The cheapest research you can do is to find the structure that already works in your category and build your own version. Not copying the ad: copying the shape. The opener type, the angle, the proof, the landing page.
Openers vary more than you'd think in how often they reach the top tier:
| Hook pattern (all formats) | Ads | Reach the top tier |
|---|---|---|
| Specific Number Opener | 411 | 13% |
| Warning/Shock Opener | 156 | 10% |
| First-Person Story | 438 | 10% |
| Problem/Benefit Opener | 714 | 9% |
| Offer-First Opener | 255 | 8% |
| Question Opener | 366 | 7% |
A number in the first line ("If you have a cat older than seven", "3 washing tips") outperforms an open question. If you only have budget for three concepts, the first should open on a specific number from your product or your customer. For the method, see how to reverse-engineer a winning ad.
The budget math
Meta's help center says an ad set exits the learning phase after about 50 optimization events within 7 days. That number sets your floor.
| Target cost per purchase | Weekly budget to reach ~50 purchases | Per day |
|---|---|---|
| $20 | $1,000 | ~$143 |
| $40 | $2,000 | ~$286 |
| $60 | $3,000 | ~$429 |
If your budget is below that line, you have three honest options:
- Run one ad set. Every extra ad set splits your events. Put your 3 to 5 ads in one ad set and let delivery pick.
- Optimize for a higher-volume event (add to cart, initiate checkout) while you find your first winner, then switch to purchases.
- Accept a slower read. Judge ads on two weeks instead of one, and change less in between. A significant edit restarts learning.
What you shouldn't do is spread $100 a day across six ad sets and twenty ads. Nothing gets enough signal and every result looks like noise. Our guide to creative testing budget goes further on the split between testing and scaling.
A small-budget plan for the next 30 days
- Research week. Pull the 20 longest-running ads in your category with six-figure estimated spend. Note the opener type, angle and landing page of each. Look for the shape that repeats.
- Build three concepts from the top two shapes. At least one static with a problem/benefit headline, one footage + voiceover video, and one founder or customer story.
- Launch in one ad set, broad audience, purchase optimization if your budget clears the table above.
- Week three: cut and iterate. Keep the best ad. Write three new openers for it. Leave the body alone.
- Week four: put most of the budget on the winner and one new concept from the research list.
For what "proven" looks like in spend terms, Facebook ad spend benchmarks has the full distribution. The research step is where a tool like Ad Radar saves the most time: filter by category, days live and estimated spend, and read the scripts. Plans start at $49 every four weeks.
Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.