Creative fatigue is the first suspect, not the algorithm
Start with frequency, not with the budget. A coffee shop targets a 3–5 km radius around Salmiya, Hawally or Kuwait City, so your reachable audience is usually 60,000 to 200,000 people — small. At 150–300 KD a month you burn through that audience in two to three weeks. Open Meta Ads Manager, set the window to the last 7 days, and add the frequency column. Under 2.0 is healthy. Between 2.0 and 2.5 you are on the clock. Past 3.0 in a seven-day window the creative is finished: CTR falls, CPM climbs, and your cost per result doubles while nothing else changed.
Refreshing means a new hook, not a new caption. The same video with different text is the same ad to the algorithm and to the customer who already scrolled past it yesterday. A café at this radius needs new creative every 10 to 14 days — call it three to four fresh assets a month, and every 7 to 10 days in Ramadan or Hala February. Film them on a phone: the milk pour in close-up, the 6 am opening, the new cold drink, a regular ordering their usual. Keep the first two seconds different every time. Run three ads at once and kill whichever crosses 2.5 first.
Before you blame the creative, confirm the drop is real. A café's numbers swing by day of week and by season. Fri–Sat behave nothing like Sun–Wed, so never compare a weekend to a weekday. Compare the last 14 days to the 14 before them, not yesterday to last Tuesday. Any window with fewer than about 50 results is noise, not a trend. If the 14-day view still shows cost per result up 40% or more, you have a real problem worth chasing. If it moved 10%, that is normal variance and touching it will only make things worse.
Ready to start your Shopify store?
Start a free trial and try the platform for yourself.
Then the calendar, then the edit you forgot you made
If frequency is fine, check the date. Kuwait's ad auction has a rhythm and cafés sit in the middle of it. CPMs climb hardest in Ramadan, in the week before Eid, and through Hala February, because every restaurant, retailer and mall is bidding on the same feeds. A CPM of 2.5–3.5 KD in a quiet month can hit 5–7 KD in those windows. Nothing is broken — you are paying peak rates for the same eyeballs. July and August are the mirror image: half the neighbourhood is travelling, CPMs fall, but foot traffic falls with them, so cost per result can still look ugly.
Next, open the change history on the ad set. Meta resets the learning phase whenever you touch targeting, budget, placements, the optimisation event or the creative, and the set then spends 3 to 7 days relearning at a worse cost. A budget jump from 5 KD to 15 KD a day does it. So does a manager fixing a typo in the primary text. If the drop starts on the same day as an edit, that is your answer: revert if you can, or leave it completely alone for a full week and judge it after. Change one thing at a time, then wait.
Check the boring plumbing too. A rejected ad inside an otherwise live campaign quietly halves delivery. A pixel that stopped firing after a website or theme update makes purchases and views disappear, so the algorithm optimises blind and the dashboard shows a crash that never happened in the shop. Test it in Events Manager and confirm events are still arriving from the last 24 hours. Same for a payment link or a Google Business profile somebody edited. Verify tracking before you rewrite strategy — roughly half the sudden drops we audit in Kuwait turn out to be measurement, not marketing.
If the ads are fine, the leak is after the click
Ads deliver clicks; the click has to land somewhere that converts. Look at where you send people. An Instagram profile with no menu in the highlights, a WhatsApp number nobody answers after 9 pm, or a Talabat page showing your shop as closed all read on the dashboard as ads that stopped working. Delivery apps also take 20–30% and keep the customer's details away from you. If online ordering is part of the plan, put a proper ordering page on your own store with KNET checkout and send paid traffic there, so you keep the margin and can retarget buyers later.
Only after all of that should you accept the honest answer: demand moved. A new café opened 400 metres away, the mall you sit in changed its anchor tenant, a road closed, or your 3 KD latte is now priced against a 1.75 KD one across the street. You will see it in the shop before Ads Manager shows it — same footfall but a smaller ticket, or fewer regulars at the same hours. Ads cannot fix a demand shift; they only pour more traffic into the same result. Fix the offer, the hours or the location first, then turn the spend back up.