Creative strategist · Media buyer · DTC

150 million plus views on ads I made.

I find the angle, write the hook, make the ad and buy the media.

See the ads
58Mviews on one ad
$2M+ad spend managed
$4M+revenue for DTC brands
7 yrsof paid social

Selected work

Every ad here ran on my own budget, mostly for my own brands. That means fewer ads than an agency reel, and no result I didn't pay for. Each one comes with the process behind it.

Soft Pet Paws · Reel “Are you trimming your dog's nails right?” Watch the ad ↗
Soft Pet Paws · Dog nail grinder · USA
“Are you trimming your dog's nails right?”
Angle
The fear of hurting your dog, instead of the product's features.
What I did
Reworked my own viral ad: tested new hooks and reordered the scenes.
Result
3.2× ROAS against 2.1× for the original. Ran for a year.

The process behind it

  1. 1. Starting pointA viral ad with an average returnThe original had 58M views but sat at 2.1× ROAS. People watched it; not enough of them bought.
  2. 2. AngleFear of hurting the dogOwners weren't doubting the grinder. They were afraid of getting it wrong. The new ad starts there.
  3. 3. StructureQuestion, problem, product, actionOpened on the question, then tested hooks and reordered the scenes until the story held.
  4. 4. Result3.2× ROAS for a yearRetired when it fatigued and the audience moved to shorter, faster UGC.
ClearSmile · Teeth whitening · Greece
“Does it really work?”
Angle
A skeptic's test, instead of the 50% offer every other ad led with.
What I did
Audited the account, then put the best-holding visual over the best-selling creator.
Result
4.86× ROAS and 459 shares. Account ROAS went from 2.68 to 4.84.

The process behind it

  1. 1. AuditTwo ads, strong in opposite placesThe UGC video turned 4.84% of clicks into buyers but only 23% kept watching. The reveal held 54% of viewers and sold half as well.
  2. 2. DecisionCombine themUse the reveal to earn the first seconds and a person to make the sale.
  3. 3. BriefReveal over the creator's questionThe purple gel reveal plays while she asks “Does it really work?”, then her story continues. No offer line.
  4. 4. Result4.86× ROAS on $3,003Hold rate stayed at 52%, 459 shares, and cost per purchase fell from $11.53 to $6.76.
AI-created videos

Generated, not filmed.

The angle and the hook still come first. AI changes how fast I can test them: a new version in hours, not a shoot.

  • “Okay. Last one.” Soft Pet Paws · UGC-style owner story
  • “I'm a groomer. Half my clients only come in for nails.” Soft Pet Paws · Expert testimonial
  • “So this is your dog's quick.” Soft Pet Paws · Vet explainer with animation

How I make them

  1. 1. Angle and scriptWritten before anything is generatedSame audit and angle work as a filmed ad. The script decides every shot.
  2. 2. Start frameOne image that sets the sceneA generated still of the product and setting. I fix details by editing that image, not by starting over.
  3. 3. VideoCheap drafts first, then the finalShort draft clips to find what works, then the full-quality run of the one that does.
  4. 4. Upscale and editFinished like any other adUpscaled, cut and captioned, then tested against the filmed ads on the same targets.

Tools: Claude, Nano Banana, Seedance, Runway, Topaz, Artlist, Adobe Premiere.

Method

Winning ads aren't luck. They're a system.

Every brief is a hypothesis about the customer. The account tells me which ones were right.

  1. 01Customer researchReviews, ad comments, support tickets and forums, read for the customer's own words: what they want, what they fear, and what they tried before.
  2. 02Competitor and category researchWhat the category runs in the Ad Library and on TikTok. Ads that stay live for months are working; angles nobody is using are the opening.
  3. 03Account auditPerformance read by angle, hook and format, not ad by ad. Hook rate, hold rate, click-through and conversion show where each creative wins and where it leaks.
  4. 04Angle mapPersonas against angles against awareness stage. It shows what's been tested, what's missing, and what to make next, in priority order.
  5. 05Briefs built to be testedOne hypothesis per brief. Three to five hooks on the same body, so a result says which idea won, not which edit.
  6. 06Test, learn, iterateBudgets and kill rules set before launch. Winners get new hooks, formats and creators; every learning goes into a library the next brief starts from.
06 → 01Then it starts again. Every round of tests makes the next brief sharper.
Work with me

Show me your ads.

Tell me about the brand and what's running now. I'll come back with what your best ad is good at, what it's missing, and what I'd make next.

Email me Message me on LinkedIn
alex.ntasis@hotmail.com · Barcelona