The ads did not break — the drop got old
Start with the calendar, not the ad account. Pull every piece you are actively spending on and write down the date it landed. In Kuwait a fashion drop has a working life of roughly 21 to 30 days before the people most likely to buy it have already seen it and decided. If more than half your spend is sitting on SKUs older than 45 days, nothing you change inside Ads Manager will save it. The metric that gives it away is new-customer purchases falling while returning-customer purchases hold steady. That is not a targeting problem or an algorithm change. That is an audience that has finished shopping this collection.
If the drop is genuinely new, look at frequency next. The reachable fashion audience in Kuwait is small — a few hundred thousand people who actually buy — so seven-day frequency climbs fast. Under 2.0 is healthy. Between 2.0 and 3.0 you are on borrowed time. Above 3.0 your cost per purchase will run 40 to 70 percent worse than week one, and no bid change fixes it. Pair it with hook rate: if three-second views fall under 20 percent of impressions, the opening frame is burnt. Same product, new first three seconds, filmed on a different girl in a different Salmiya flat usually resets the numbers on its own.
Fix the cadence, not the campaign. A brand spending 40 to 80 KD a day needs three to five genuinely new creatives a week, not three crops of the same photo. Shoot in batches: one afternoon gets you a month of content if you plan the outfits in advance. Keep two ad sets on the proven creative and one always testing, capped at 20 percent of daily budget. Kill a test once it has spent roughly three times your target cost per purchase with nothing to show. Most brands who say their ads stopped working are still running creative that is eight weeks old.
Ready to start your Shopify store?
Start a free trial and try the platform for yourself.
A sold-out size looks exactly like a broken ad account
This is the one owners miss most. The ads are still working, the traffic is still coming, the add-to-carts are still happening — but M and L are gone. For womenswear in Kuwait, medium and large are usually 55 to 70 percent of units sold. Lose them and your conversion rate halves while spend continues at full speed, and the account reads like a targeting collapse. The number that separates the two is add-to-cart rate versus cart-to-purchase rate. If add-to-carts per session are unchanged but the share of carts that convert falls from a normal 25 to 35 percent down under 15, you have an inventory problem wearing an advertising costume.
The fix is structural, not tactical. Your ads should never be able to send paid traffic to a size that does not exist. A store on Shopify syncs stock to the Meta and TikTok catalogue automatically, so a sold-out piece drops out of dynamic ads within hours instead of burning budget for a week. Then add a rule you check every Sunday: any SKU where the two best-selling sizes are out gets pulled from active creative the same day. And set a floor — once a hero piece is under 30 percent of its original stock, stop scaling it and move the budget to the next piece.
Then check the auction, and the pixel last
Kuwait's ad auction is seasonal and brutal about it. A fashion account that normally pays 2 to 5 KD per thousand impressions will pay 7 to 10 KD in the last ten days of Ramadan and the week before Eid, because every brand in the country is bidding at once. Hala February and back-to-school do a milder version of the same thing. July and August are the opposite trap: CPM drops because half of Kuwait is travelling, so the cheap traffic looks great and buys nothing. Compare this week's CPM to the same week last year, not to last month.
Tracking comes last because it is the rarest cause, but it is the most expensive when it happens. Open your store's real order count for the last seven days and put it next to what Meta reports. A gap under 15 percent is normal attribution noise. A gap over 30 percent means the pixel or Conversions API stopped firing, usually right after a theme edit, an app install, or a domain change. Check event match quality too — under 6.0 out of 10 and the algorithm is optimising half blind. On the catalogue side, a feed with disapproved items or missing prices quietly kills dynamic retargeting.
Work the list in order and you will find it in one afternoon. Age of the drop, then frequency and hook rate, then sizes, then CPM against last year, then tracking. Do not change bids, budgets and creative on the same day — you will never know which one moved. Change one thing, give it three to four days and at least 50 results before you judge it, and write down what you changed with the date. Brands that keep that log stop having mystery drops within two or three months, because the same three causes keep repeating.