The scenario
The brand distributes a mix of Korean and European skincare lines in Kuwait, direct to consumer, on a Shopify store with KNET, Tabby and cash on delivery at checkout. Three staff run the account day to day, sourcing from a handful of overseas distributors and holding stock locally so orders placed before the evening still arrive the next day. It competes on the same shelf as Boots and Sephora at the Avenues, with a following built mostly on TikTok reviews rather than paid reach alone.
The advertising problem had been quietly building for a year. Meta and TikTok had rejected or limited the reach of a growing list of ads, and each rejection was treated as a one-off — reword the caption, resubmit, move on. Nobody had stepped back to ask why the same account kept tripping the same wire, or noticed that every rejected ad shared the same structure: a claim about what would happen to the customer's skin, dressed up in slightly different words each time.
- Monthly revenue band
- pending client sign-off
- Average order value
- pending client sign-off
- Fulfilment
- Local stock, own courier, next-day across Kuwait
- Team
- Three staff covering sourcing, orders and content
Every rejection was treated as a paperwork problem. It was a creative problem.
When we pulled the account's disapproval history, the pattern took minutes to see and had apparently taken a year to notice. Nearly every rejected or limited ad made some version of the same promise: skin that looked calmer, brighter or smoother after use, usually shown as a split-screen or a slow zoom on the buyer's face. The brand's response each time had been to soften the wording — swap 'clears' for 'helps with', swap a number of days for 'soon' — which platform review systems read exactly the same way, because the image and the implication had not changed.
Underneath the compliance issue was a second problem the rejections were hiding: even the ads that did get approved were not winning. A before-and-after photo is an easy brief to write and a hard one to make honest, and split-screen claims about someone else's face read as generic to a Kuwaiti audience that has seen the same format from a dozen brands and has learned to distrust it on sight. The account had spent a year fighting the platform over a format that was not converting well even when it survived review.
Nobody on the team had ever built an angle sheet from something other than the product itself — no read of the WhatsApp questions, no read of the reviews, nothing about the fact that half of the reorders were happening two to three weeks before Ramadan, when routines change and skin gets more sensitive to fasting, heat and less sleep. The account did not have a wording problem. It had built its entire creative library on the one claim every platform is built to catch, and it had nothing else to say.
What we did — the creative angles
The losers are here on purpose. A test with only winners was never a test.
The split-screen face claim that started the rejections
Lost"Calmer skin in two weeks — see the difference."
- Format:
- Vertical, 15 seconds, split-screen face close-up, day one against day fourteen
This was the format the account had leaned on for a year, and it is the exact shape platform reviewers are trained to catch: a specific before-and-after on skin, paired with a timeframe. It kept getting flagged, and on the runs that did get through, cost per purchase was high because the claim reads as generic in a market that has seen the same split-screen from a dozen brands.
An influencer's own before-and-after, reshared as an ad
Lost"This is what three weeks of the routine did for me."
- Format:
- Vertical, 20 seconds, reposted organic content boosted as a paid ad
Organic content from a real customer felt safer to the team because it was 'her word, not ours', but boosting it as an ad puts the brand behind the same claim in the platform's eyes regardless of who is speaking. It was rejected twice and then quietly abandoned, having spent budget on appeals rather than on anything that could actually run.
The morning routine, filmed in the order it is applied
Winner"Cleanser, serum, moisturiser, SPF — in that order, every morning."
- Format:
- Vertical, 18 seconds, phone camera in a real bathroom, no face close-up
The single most common WhatsApp question in two years of chat history was not about results — it was 'what order do I use these in'. Answering that question on camera, with hands and bottles rather than a face, sells the routine without making a claim about what it does to anyone's skin, and it approved cleanly on every platform on the first submission.
The texture, up close, with no face in frame
Winner"This is what a gel-cream feels like versus a heavy cream."
- Format:
- Vertical, 10 seconds, macro shot of the product being swatched on the back of a hand
Texture is the thing a customer genuinely cannot know from a product page, and describing it makes no claim at all about outcome. This performed especially well on TikTok, where the audience is used to ASMR-style product content and reads it as information rather than as an advertisement, and it doubled as an honest answer to the humidity-versus-air-conditioning question Kuwaiti buyers ask about heavier creams in summer.
The batch code, shown on camera
Winner"Here is exactly where to find the batch code on your box."
- Format:
- Vertical, 12 seconds, close-up on packaging, pointing to the printed code
Authenticity, not efficacy, was the second most common worry in the brand's inbox — Kuwaiti buyers of imported skincare are used to counterfeits sold cheap on social platforms, and a lot of them check for this before checking anything else. A short ad that just teaches the customer how to verify her own box addresses the actual objection stopping the sale, and it never mentions the skin at all.
Why she reordered, in her own words
Winner"I stopped buying it for a month. Then I went back."
- Format:
- Vertical, 22 seconds, unscripted customer testimonial, no product shot of skin
The customer talks about how it feels to use, how it fits into a routine she was already running, and the fact that she went back to it after trying something cheaper — never about what it did to her skin, because we did not ask her to say that. It reads as a reorder story rather than a transformation story, and it was the strongest performer on hold rate across the whole test.
An ingredient explainer, read like a label
Neutral"What's actually in the bottle, in plain Arabic."
- Format:
- Carousel, five cards, one ingredient per card with a plain-language note
Educational and safely compliant, but it tested as neutral: informative enough to hold attention, not specific enough to a Kuwaiti buyer's actual worry to convert at the rate of the routine or the authenticity angles. It earns a place in retargeting for the buyer who has already decided to trust the brand and wants the detail, rather than in the cold audience it was built for.
What we did — the optimizations, in order
Read the disapproval history like a diagnosis, not a queue
We pulled every rejected and limited ad from the past year and tagged each one by the specific claim it made, rather than treating the rejections as a wording issue to resolve one appeal at a time.
Why: A wording fix addresses the symptom the platform happened to catch. Tagging the underlying claim showed us the account had exactly one creative idea repeated in different clothes, which is a production problem, not a compliance one.
Build the angle sheet from the inbox, filtered for what is safe to say
We read two years of WhatsApp messages and reviews for the questions that were not about outcome — routine order, texture, authenticity, how it holds up against Kuwait's summer humidity and winter dryness — and built eight angles from those.
Why: This is the step that makes compliance and performance the same problem instead of a trade-off. Every angle here is both platform-safe and a real question the brand's own customers were already asking, so nothing on the sheet needed the founders' instinct for what might slip past review.
Shoot for the reason someone buys, never for the result on a face
Every brief for the production sprint specified what the camera was allowed to show — hands, bottles, packaging, texture on a surface — and what it was never allowed to show, which was a face used as evidence of a change.
Why: Writing the constraint into the brief up front is what stops a well-meaning editor from reaching for the easiest shot on set, which is always a face. It also means every deliverable is reviewable in one pass instead of going back for reshoots after a rejection.
One testing campaign per platform, frozen everywhere else
One campaign per platform, one ad set per angle, equal budgets, five to seven days, with the audience, the offer and the landing page held identical across all seven angles.
Why: Removing every other variable is what let us say with confidence that the routine and authenticity angles outperformed the ingredient explainer, rather than one of them simply landing in front of a warmer audience by chance.
Read hook, then hold, then click, then cost
Each of the seven angles was judged in that order, and we wrote down the step it succeeded or failed at rather than a single pass-or-fail verdict on the whole ad.
Why: The ingredient carousel and the routine video both got approved, both had a decent hook, and only reading further down the funnel showed that one held attention and converted and the other did not. Stopping at approval tells you nothing about whether the ad sells.
Promote the winners, and write down why the losers lost
The routine, texture, batch code and testimonial angles moved into a scaling campaign, and every rejected or underperforming angle went into a shared document with the specific reason it did not survive.
Why: A team that only keeps a record of what worked will eventually reshoot the split-screen face claim in eighteen months, because nobody remembers why it was banned. Writing the loss down is what makes the lesson survive staff turnover.
Plan the Ramadan and wedding-season variants around the routine change
Because reorders already clustered before Ramadan and before wedding season, we scheduled routine-led variants for both four weeks ahead — a lighter evening routine for fasting skin, a build-up routine for brides — rather than shooting them the week they were needed.
Why: The demand shift around Ramadan is not a guess; it showed up directly in the reorder data. Having the seasonal variant ready before the season starts is the difference between meeting existing demand and discovering it a week too late.
What changed
The table above carries the numbers once the client confirms them, and the shape worth stating now is that two things moved together rather than one causing the other by accident: the account stopped losing ads to rejection, and cost per purchase came down as spend that used to sit on disapproved or throttled creative moved onto ads that were both compliant and better at selling.
The clearest read came from comparing the two loser angles against the four winners. Every angle that survived and scaled showed a process — a routine, a texture, how to check a box — or let a real customer speak for herself without being asked about her skin. Every angle that lost either made a claim about someone's face or repackaged someone else's claim as an ad. That is not a coincidence specific to this brand; it is closer to a rule for the category.
The ingredient carousel's neutral result was as useful as the winners. It confirmed that education alone does not carry a cold audience the way a routine or an authenticity angle does, which told the team exactly where to spend the next production sprint's budget rather than splitting it evenly across every safe idea.
What we would do next
First, extend the same rule to the product pages, because a page that still leans on split-screen imagery undoes the discipline the ads now carry the moment a customer clicks through. That is a page audit, not a new production sprint.
Second, build a standing review of every disapproval within a week of it happening rather than at the end of a quarter, so a drifting pattern gets caught at ad three rather than ad thirty.