The best ad ideas to try next are ones that already moved someone else's numbers: scrappy product videos, sharp problem angles and offers shown on the ad.
Below are 25 ideas from real results. In our database, a catalog flip-through made in 10 minutes beat an agency video on TikTok (1.32 vs 0.29 ROAS), burnout and ADHD angles cut an app's Meta CPA from €57.15 to €23.85, and a bundle template took catalog ROAS from 1.96 to 4.49 (vendor-reported).
Quick answer
If you only try three ideas this month:
- A low-effort native video of the product in use (Baker Creek, Petit BamBou).
- A new angle built on a named pain (the mindfulness app, Booksy, Ambrook).
- An offer shown on the ad: a bundle, strikethrough price or free delivery (Twillory, Latico, Raymour & Flanigan).
How we ranked them (evidence filter)
How we picked: we pulled ROASArchive results where the creative, angle, format or offer was a named part of the change and a number was attached.
We have not run these ads ourselves; every number comes from whoever published it.
We ranked by how many times better the main cost or return metric got, so a 50% CPA cut and a doubled ROAS both count as 2x.
Entries 22-25 have no before-and-after cost number (a spread across ads or a single result), so they sit at the bottom.
Most are before/after case studies where several things changed at once, not split tests.
Tool vendors' stories about their own products are marked vendor-reported.
Ranked table
| Rank | Test | Change | Result | Who ran it |
|---|---|---|---|---|
| 1 | Baker Creek (TikTok) | Quick catalog flip vs agency video | ROAS 0.29 → 1.32 | scandiweb |
| 2 | Mindfulness app (TikTok) | Bold text + trending audio | D14 ROAS 26.92% → 71.9% (peak) | AdQuantum |
| 3 | Ambrook (Meta) | Product-first ads, "tax prep" angle | Cost per qualified lead −62% | Flighted |
| 4 | Booksy (TikTok) | Pros talking about no-shows | CPA −60% | inBeat |
| 5 | Magic Story (Meta) | Micro-persona angles | Cost per subscriber ~−60% | Flighted |
| 6 | Mindfulness app (Meta) | Burnout and ADHD angles | CPA €57.15 → €23.85 | AdQuantum |
| 7 | Petit BamBou (TikTok) | Creator + real app footage | CPI −57% | Admiral Media |
| 8 | Twillory (Meta) | Bundle and review templates | ROAS 1.96 → 4.49 | Marpipe (vendor-reported) |
| 9 | TAO Clean | "3 reasons why" UGC | CPA −50% | Hustler Marketing |
| 10 | Unnamed brand (Meta) | €10 price cut, same ads | CPA €170.34 → €93.51 | @SamPiliero |
| 11 | The Sill (Facebook) | Men and women split creative | ROAS 158% → 274% | Hawke Media |
| 12 | Unnamed brand (Meta) | ~20 new angles | CPA $101.29 → $60.50 | @SamPiliero |
| 13 | Raymour & Flanigan (Meta) | Free-delivery callout | ROAS +54% | Marpipe (vendor-reported) |
| 14 | NeuroNation (Google) | Progress-bar visual | Cost per purchase −34% | Admiral Media |
| 15 | Sleepy Tie (Meta) | Organic videos as ads | ROAS +48% | Flighted |
| 16 | Quay (Snapchat) | Dynamic product ads | ROAS +40% | Common Thread Collective |
| 17 | Honeydew Care (Meta) | Captioned recut for sound-off | CPL −26% | Flighted |
| 18 | Latico (Meta) | "20% Off" strikethrough | ROAS 2.66 → 3.52 | Marpipe (vendor-reported) |
| 19 | Bedding brand (Meta, Google) | 15% off Labor Day | Blended ROAS 10.11x → 13.17x | @SamPiliero |
| 20 | Taylor Stitch (Meta) | Lay flats vs models | CPA −18% | Marpipe (vendor-reported) |
| 21 | Inshallah (Meta) | AI-translated UGC | CPI 15% below €2.52 | Admiral Media |
| 22 | Grüns (Meta) | Ads by avatar | ROAS 2x–5x by avatar | @SamPiliero |
| 23 | Owner (Meta) | Free tool instead of demo | $60–$140 per qualified lead at $1M+/mo | Flighted |
| 24 | Ecommerce brand (Meta) | Statics only | $32.98 per purchase | @runtraffic |
| 25 | Ecommerce brand (Meta) | Old-footage b-roll mashup | $58.61 CPA on $107k | @alexgoughcooper |
The 25 ad ideas, one by one
This list is about what to make; for what ads looked like, browse our ad examples.
More angle tests and the best ad hooks each have their own page.
1. Shoot the scrappy version of the product video
The change: scandiweb shot a quick flip-through of Baker Creek's seed catalog for TikTok, made in about 10 minutes.
The result: 233 purchases at 1.32 ROAS, vs 1 purchase at 0.29 for the agency's polished video. Split test; spend per video not shared.
Why it likely worked: It looked like a post, not an ad, and showed the product fast.
Try it if: Your polished video is losing (1.32 ROAS is still thin after costs).
2. Rebuild TikTok ads with bold text and trending audio
The change: AdQuantum relaunched a mindfulness app on TikTok with burnout angles, bold text and trending audio.
The result: D14 ROAS up to 71.9%, from 26.92%.
A peak, with bidding rebuilt too. Agency before/after.
Why it likely worked: Bold text works with the sound off; familiar audio feels native.
Try it if: Your TikTok ads are reused Meta videos.
3. Swap lifestyle ads for product-first ads with one urgent angle
The change: Flighted replaced Ambrook's lifestyle ads with product-first Meta ads and a "tax prep" angle.
The result: Cost per qualified lead −62% while spend grew 181% month over month. Agency before/after; bidding also changed.
Why it likely worked: A deadline (taxes) gave B2B buyers a reason to act now.
Try it if: Your software ads explain the brand more than the product.
4. Let niche creators talk about their daily headaches
The change: inBeat had barbers and nail techs post TikToks about no-shows, run from their accounts.
The result: CPA −60% while paid media scaled 6x. Agency before/after.
Why it likely worked: Pros trust other pros, and the pain was instantly recognizable.
Try it if: You sell to a trade and your ads are written by marketers.
5. Write a separate angle for each micro-persona
The change: Flighted wrote Magic Story angles for micro-personas such as gift-buying grandparents.
The result: Cost per subscriber roughly −60% in 60 days as spend rose from $20k to $150k+ a month. Includes Black Friday. Agency before/after.
Why it likely worked: Each buyer type saw its own reason to buy.
Try it if: One product has several distinct buyers, especially gift buyers.
6. Lead with a named problem, not an aspiration
The change: AdQuantum swapped the same app's self-improvement message for burnout and ADHD angles on Meta.
The result: Burnout funnel CPA €23.85, vs €57.15 before.
Best funnel only, with bid tests running too. Agency before/after.
Why it likely worked: "Burnout" is a problem people feel today. "Self-improvement" is a vague wish.
Try it if: Your product fixes a specific pain but your ads sell a dream.
7. Film real app footage with a native creator
The change: Admiral Media switched Petit BamBou's French TikTok ads to creator Spark Ads using real in-app footage.
The result: CPI −57%. One viral video (40.6 million views) drove much of it. Agency before/after.
Why it likely worked: Viewers saw exactly what they'd get, from a person rather than a brand.
Try it if: Your app ads use mockups or animated screens.
8. Put a bundle deal or a review on catalog images
The change: Twillory tested bundle and review templates over its Meta catalog ads at equal spend.
The result: ROAS 1.96 (plain) → 4.49 (bundle) and 4.41 (reviews) over ten days. Vendor-reported by Marpipe.
Why it likely worked: A plain product photo gives no reason to buy today.
Try it if: You run catalog ads with untouched product shots.
9. Try a "3 reasons why" UGC script, then remake the winner
The change: Hustler Marketing's winning TAO Clean UGC ad was a "3 reasons why" script, later remade with a new creator.
The result: CPA −50%, 3.2 ROAS in April 2024, 49.4% hook rate. Agency before/after.
Why it likely worked: A numbered list promises a short, clear payoff.
Try it if: You need a first UGC script to brief creators with.
10. Test a lower price before new ads
The change: A brand cut its price by €10, same ads and budget (shared by @SamPiliero).
The result: CPA €170.34 → €93.51 (−45%), ROAS 2.5x → 4.5x over 10 days. Before/after; margin not shared.
Why it likely worked: Sometimes the offer, not the creative, is what's broken.
Try it if: Your ads get clicks but not sales and you have margin to test.
11. Split creative by gender
The change: Hawke Media gave men and women their own The Sill creative and copy.
The result: ROAS 158% → 274%, CPA −41% while spend grew 18x. Year-over-year agency comparison.
Why it likely worked: The two groups bought plants for different reasons.
Try it if: Your reviews show men and women buy for different motives.
12. Launch a batch of new angles in a tired account
The change: @SamPiliero's team launched about 20 angles for two avatars in one Meta campaign.
The result: Cost per purchase $101.29 → $60.50 (−40.27%) over two 30-day periods. Before/after.
Why it likely worked: New angles reach buyers the old one missed.
Try it if: CPA has crept up and you keep remaking the same idea.
13. Add a free-delivery callout to product ads
The change: Raymour & Flanigan added "Free Next Day Mattress Delivery" to its product ads.
The result: ROAS +54% vs the control design. Vendor-reported by Marpipe.
Why it likely worked: It answered the biggest objection right on the image.
Try it if: Free or fast shipping only shows up at checkout.
14. Build the visual around visible progress
The change: Admiral Media rebuilt NeuroNation's Google App ads around a progress bar.
The result: Cost per purchase −34% (French Android) while spend rose 48%. Agency before/after.
Why it likely worked: A progress bar shows the outcome in one glance.
Try it if: You sell improvement over time: learning, fitness or habits.
15. Run your best organic videos as ads
The change: Flighted launched Sleepy Tie's proven organic videos as Meta ads in groups of three.
The result: ROAS +48%. Agency before/after; bid caps and checkout changed too.
Why it likely worked: Organic views are a free pre-test of attention.
Try it if: You post on TikTok or Reels but your ads don't use it.
16. Switch to dynamic product ads
The change: Common Thread Collective ran Snapchat Dynamic Product Ads for Quay beside regular campaigns.
The result: ROAS +40%. Side-by-side test in Q4 peak season.
Why it likely worked: Each viewer saw products close to what they'd browsed.
Try it if: You have a big catalog and a few hand-picked product ads.
17. Recut UGC with bold captions for sound-off viewing
The change: Flighted recut Honeydew Care's top organic TikToks with bold text overlays for Meta.
The result: CPL −26% in the ASC campaign the week it scaled. One-week agency before/after.
Why it likely worked: Most feed viewers watch muted.
Try it if: Your best videos rely on someone talking.
18. Show a strikethrough discount on the image
The change: Latico added a "20% Off" strikethrough price to Meta catalog images in a 14-day test.
The result: 3.52 ROAS vs 2.66 for all other May ads (+32%). Vendor-reported by Marpipe.
Why it likely worked: The saving is visible before the click.
Try it if: Your sale ads look the same as full-price ads.
19. Run a simple holiday sale
The change: @SamPiliero's team ran a 15% off Labor Day sale for a bedding brand.
The result: Revenue +102.1% to $66,218.62 and blended ROAS 10.11x → 13.17x. Before/after; holidays lift sales anyway.
Why it likely worked: A familiar date ends the waiting.
Try it if: You never discount and want to see if a modest sale pays.
20. Shoot products in a setting without a model
The change: Taylor Stitch tested styled "environmental lay flats" against on-model images.
The result: 18% lower CPA for lay flats. Vendor-reported by Marpipe.
Why it likely worked: A product in context reads well in a small feed image.
Try it if: You sell apparel or home goods and only test model shots.
21. Translate a proven ad with AI instead of reshooting
The change: Admiral Media AI-translated a winning English UGC ad into French for Inshallah.
The result: CPI 15% below a native French creator ad's €2.52. 10-day test in one ad set.
Why it likely worked: The script was already proven; only the language changed.
Try it if: You have winners in another language.
22. Build ads around different avatars and compare
The change: @SamPiliero shared Grüns' Meta ads built around separate customer avatars.
The result: ROAS ranged from 2x to 5x by avatar. Campaign breakdown, no control.
Why it likely worked: The spread shows how much the angle choice matters.
Try it if: You've only written ads for one type of buyer.
23. Swap a demo ask for a free tool
The change: Flighted replaced Owner's "book a demo" ask with an AI Grader lead magnet.
The result: Spend grew to $1M+ a month at $60–$140 per qualified lead. Agency case study; results vary.
Why it likely worked: A free tool is an easier first yes than a sales call.
Try it if: Demo requests are expensive or scarce.
24. Try a statics-only campaign
The change: @runtraffic ran an ecommerce Meta campaign with only static image ads.
The result: $32.98 cost per purchase on $253k. No video comparison.
Why it likely worked: Statics are cheap, so you can test more offers.
Try it if: Video production is your bottleneck.
25. Cut old footage into a b-roll mashup
The change: @alexgoughcooper shared a Meta video cut from old footage as a b-roll mashup.
The result: $58.61 CPA on $107k of spend. Single ad, no comparison.
Why it likely worked: It's new to the viewer even if the clips aren't.
Try it if: You have unused clips and no shoot budget.
What the winners have in common
Native beat polished
The scrappy catalog flip, the in-app creator footage, Booksy's pros and the recycled organic posts all looked like posts, not ads.
The angle mattered more than the edit
Grüns saw a 2x-5x ROAS spread by avatar, and the mindfulness app more than halved CPA by naming burnout.
The offer belongs on the ad
Bundles, strikethrough prices and free delivery all worked on the image itself.
The €10 price cut shows the offer sometimes needs fixing more than the ad.
Ideas came in batches
@SamPiliero's team tried about 20 angles at once. Most ideas lose; volume finds the winner.
What to test next
Pick one idea from each group: one format idea (1, 7, 15 or 25), one angle idea (3, 5, 6 or 12) and one offer idea (8, 10, 13 or 18).
Then run it as a creative test with a fixed budget and a single success metric.
- Write down the metric first. CPA, ROAS or cost per lead, not CTR alone.
- Change one thing per ad. Most results above changed several things at once, so their cause is unclear.
- Iterate on the winner, as TAO Clean did with a new creator.
Compare your plan with what advertisers test most.
For platform-specific ideas, see our TikTok ad examples and UGC ad examples.
For more ideas before you brief creators, an ad library helps.
Foreplay is a swipe library with creative briefs; Atria adds creative analytics and AI briefs to an ad library.
We haven't used either. Compare them in Foreplay vs Atria or see more ad spy tools.
Both have affiliate programs, so we may earn a commission if you sign up through our links.
Results vary by product, margin and market, so test before you scale.