The scenario
The brand sells everyday sneakers, casual shoes and a school-shoe range for kids, on a Shopify store with KNET, Tabby and cash on delivery at checkout, shipped by a local courier across Kuwait. A small team, an Instagram following built on styling reels, and a catalogue priced comfortably inside pocket-money and back-to-school budgets.
Traffic was healthy and add-to-cart was fine. The problem showed up after the order left the warehouse: about a fifth of pairs came back, split almost evenly between wrong size and refused at the door, and every ad in rotation was a styling shot that never once mentioned a measurement.
- Monthly revenue band
- pending client sign-off
- Average order value
- pending client sign-off
- Fulfilment
- Own stock, local courier, next-day across Kuwait
- Team
- Founder plus two part-time stylists who shoot the reels
Style sold the click. Size lost the order.
When we pulled the return log against the ad account, the pattern was immediate: every ad in the last quarter was a styling shot — a pair on a marble step, a pair on a moving foot, a discount card — and not one of them mentioned a measurement. The return reasons, read in the customer's own words, were almost all about size: runs small, looked bigger in the photo, ordered my usual size and it didn't fit.
The brand's read had been that a fifth of returns was just normal for online footwear, and in some categories that might be true. What it was actually paying for was cash-on-delivery orders refused at the door because the buyer never trusted the size in the first place, plus the courier trip, the repacking, and the customer who quietly never orders again. None of that showed up in cost per purchase, which is the number the ad account was optimising for.
The ads were not underperforming. They were winning the wrong argument — convincing someone the shoe looked good, when the real decision at checkout was whether it would fit. The account did not have a creative-quality problem; it had never once tested an ad that answered the size question, so every winning ad was quietly manufacturing a return.
What we did — the creative angles
The losers are here on purpose. A test with only winners was never a test.
The centimetre measurement, stated on camera
Winner"Measure your foot in centimetres before you order — here's how."
- Format:
- Vertical, 15 seconds, phone camera, a ruler against a bare foot on the floor
Sizing was the single most common return reason, and no ad had ever shown the customer how to check it. Leading with the measurement turns the ad into a tool the buyer uses at the moment of doubt, and it disqualified the wrong size before the order was placed rather than after it arrived. It was the clearest winner on hook rate and it held cost per purchase as spend increased.
True to size or runs narrow, said plainly
Winner"This one runs narrow. Order half a size up."
- Format:
- Vertical, 12 seconds, text-on-screen over a foot sliding into the shoe
The second most frequent DM question, and a pure trust objection: shoppers who had been burned once by a model that ran small. Saying it out loud did more for the checkout rate of the retargeting audience than any discount the brand had run, because it removed the last reason a saved item never became an order.
The authenticity unboxing
Winner"Original box, original tags — here's the invoice."
- Format:
- Vertical, 20 seconds, unboxing with the supplier invoice and authentication card shown on camera
Kuwait's sneaker market is full of copies, and "is this real" was sitting unanswered in the comments of every ad. Showing the paperwork instead of claiming authenticity in a caption worked because it gave the buyer something to check, and it was the single best performer for cold audiences on TikTok.
Back-to-school durability drop test
Neutral"Dropped, kicked, still standing — school shoes shouldn't need babying."
- Format:
- Vertical, 16 seconds, outdoor durability test filmed with a child actor on a school playground
Timed for the September rush, this leaned on the parent's real worry — will it survive a term of actual use — rather than how the shoe looked in a lookbook. It performed solidly through the back-to-school window and flattened the moment the season passed: a calendar-anchored angle, not an evergreen one.
The styling reel
Lost"New drop, straight from the box."
- Format:
- Vertical, 14 seconds, aesthetic styling shots set to music, no voice, no text
This was the brand's default format before testing, and it kept losing for the same reason every time: it looked good and answered nothing. It pulled cheap likes and saves but the worst return-adjusted cost per purchase in the round, because the people it attracted were shopping the aesthetic, not the fit — and some were exactly the buyers who later refused the parcel at the door.
National Day colourway
Neutral"Kuwait colours, a limited run before the 25th."
- Format:
- Vertical, 15 seconds, product reveal in national flag colours with a countdown overlay
A genuinely strong week around National Day and Liberation Day, riding the same wave every seasonal brand in Kuwait chases, and flat outside that window. Worth keeping on the annual calendar rather than the always-on rotation, shot and live two weeks ahead of the date, not the week of it.
The exchange promise, on its own
Lost"Wrong size? We swap it free within three days."
- Format:
- Vertical, 10 seconds, text-on-screen policy statement over a plain background
A reasonable idea that lost as a standalone ad: a policy statement with no visual proof reads as a claim, not evidence, and it barely moved hook rate on cold audiences with no reason yet to care about an exchange. The same line worked far better as the closing seconds of the measurement video.
What we did — the optimizations, in order
Cross the return log against the ad account
We pulled the last quarter of ad spend and the last quarter of return reasons into one sheet, tagged every ad by angle and every return by cause, and matched them against each other rather than reading either list alone.
Why: Cost per purchase alone hides a return sitting three weeks downstream. Joining the two lists showed the styling-led ads were quietly the most expensive creatives in the account once refused-at-the-door orders were subtracted.
Build the angle sheet from fit questions, not the lookbook
We read a year of DMs, size-exchange requests and product reviews and pulled the eight questions that came up most, then wrote each one as the customer's own sentence, measurement-led rather than style-led.
Why: The brand already had every objection sitting in its inbox; it had simply never turned one into an ad. Anchoring the sheet to real fit questions gave the founders a shared list to shoot against, instead of guessing at a new look every month.
Shoot proof, not polish
Ten pieces in two weeks: a ruler against a bare foot, an unboxing with the invoice on camera, an outdoor durability test, each cut with three hook variants, filmed on a phone rather than booked into a studio.
Why: A studio shot photographs well and answers nothing. Proof formats — a measurement, a document, a stress test — carry the evidence the buyer needs at the moment of doubt, and cost less to produce than the lookbook shoots they replaced.
One testing campaign per platform, everything else frozen
One campaign per platform, one ad set per angle, equal budgets, five to seven days, with audience, offer and landing page identical across every ad set in the round.
Why: Sizing content changes buyer behaviour in ways a normal creative test does not — it can lower orders and returns together. Freezing everything but the creative is the only way to read that trade-off honestly.
Read hook, then hold, then click, then return-adjusted cost
We judged each ad in that order and added a fifth number the brand had never tracked: the share of orders from each ad that came back within fourteen days, matched by ad ID rather than campaign.
Why: A creative that wins on cost per purchase and loses on returns is not a winner, it is a return with a delay. Tracking the return share per ad let us tell a genuinely good hook apart from one that was just fast at placing a doubtful order.
Promote winners, retire the styling default
The three winning proof-led ads moved into the scaling campaign, the styling reel was retired rather than merely paused, and the exchange-policy line was folded into the closing seconds of the measurement video instead of running alone.
Why: Keeping a known loser live out of habit is how an account slides back to where it started. Retiring it and folding its one useful line into a stronger ad changes what is actually spending money, not just what exists in the account.
Set a September and pre-Eid cadence, not a constant one
Six new pieces every two weeks through the school-shoe season in August and September, dropping to two a month outside it, with the National Day colourway shot and approved three weeks ahead of the date rather than the week before.
Why: Footwear demand in Kuwait moves in named windows — back to school, Eid, National Day — and a flat, always-on cadence either wastes budget in quiet months or arrives too late for the peak ones. Matching cadence to calendar is what makes the testing budget worth spending year-round.
What changed
The table above carries the headline number, and the shape matters more than the size: cost per purchase moved down while spend went up, and it moved because the share of spend sitting on proof-led creatives went from a minority to a clear majority — not because one ad went viral. Read against the return log rather than the ad platform alone, that shift is worth more than the raw number suggests.
The pattern across winners was consistent: every winning ad answered a doubt about fit or authenticity before it asked for the sale, and the two clearest losers were the two that only showed the product looking good. The brand's instinct going in had been to make prettier styling content; what actually moved the number was answering, in the first two seconds, the exact question a buyer was about to ask a courier at the door.
There was a second effect the team noticed before we pointed it out. Because the measurement and fit ads pre-answered the size question, exchange requests started arriving with the right size already stated rather than as an open-ended complaint, which made the exchange itself faster to process even before any return-rate number moved.
What we would do next
Put the same measurement content on the product page itself, not just in the ad. Right now a buyer who clicks through from the winning ad lands on a page with a size chart in inches and no note on fit, which reopens the exact doubt the ad had just closed.
Build the Eid and National Day proof-led variants a month ahead rather than the week before, and add a return-share column to the weekly ad report as standard, so a cheap order that comes back stops looking like a win in the dashboard the founders actually read.