Start by checking whether performance actually fell
This sounds patronising and it is not. Meta reports conversions against the day of the click or view, not the day of the purchase, and it keeps filling those numbers in for days afterwards. Look at the last three days and you are reading an incomplete row.
Compare a settled period with an equivalent settled one — a full fortnight against the previous full fortnight, ending several days ago. Also check that you are comparing like with like: a period containing Dashain is not comparable to one that does not, and a week with a public holiday behaves differently on both sides of the market.
If the drop survives that comparison, it is real. Now work down the list.
Cause 1: you edited the ad set and reset the learning phase
This is the most common self-inflicted cause, and the most frustrating, because it is caused by trying to help.
When you make a significant edit to an ad set — budget, audience, optimisation event, placements, creative — Meta re-enters the learning phase. Delivery becomes unstable and cost per result usually rises while the system re-learns. If you edit again two days later because results look bad, you reset it again. It is entirely possible to keep a campaign permanently in learning and conclude the platform has stopped working.
The fix is discipline rather than cleverness: make changes in batches rather than daily, and give a change enough time and enough conversions to be judged. Where you want to test something properly, make a new ad set rather than editing the one that is performing.
Cause 2: the audience is exhausted
A small audience seen repeatedly stops responding. Frequency is the number to look at — Meta reports it directly — and the pattern is recognisable: frequency climbing, click-through falling, cost per result rising, all at once and gradually rather than overnight.
This is structural, not a creative problem, and it is guaranteed with a small retargeting pool or a tightly-defined interest audience in a market the size of Nepal. Broadening the audience, capping frequency, or rotating in genuinely different creative all help. Making the same ad in a different colour does not, because it is the same ad to the person who has seen it nine times.
If you want to work out in advance how fast a given budget will exhaust an audience, the reach and frequency calculator does that arithmetic.
Cause 3: the creative is worn out
Distinct from audience fatigue, though they look similar. Creative fatigue shows as a drop in click-through rate on one specific ad while others in the same ad set hold steady. Audience fatigue drags everything down together.
The practical difference matters: creative fatigue is fixed by new creative, audience fatigue is not. Check per-ad rather than per-campaign before deciding which you have.
Cause 4: tracking broke and nobody noticed
This one is dangerous, because performance did not fall — your measurement of it did, and every decision you take afterwards is made on false numbers.
It happens after a site change, a new consent banner, a theme update, a developer removing what looked like an unused script, or a change to the thank-you page URL that the conversion event was watching. The signature is unmistakable once you know it: clicks and spend steady, conversions falling off a cliff on a specific date.
Test it rather than assuming. Submit your own form and confirm the event fires and arrives. If your conversions are dated to a single day when everything changed, suspect tracking before you suspect the market — and a tracking audit is cheaper than a month of decisions made on broken data.
Cause 5: competition and seasonality
You are bidding in an auction. When more advertisers enter it, or the ones already there raise their budgets, your costs rise without you doing anything. Festival periods, election periods and the run-up to any major shopping event all do this, and they affect everyone in the auction at once.
The way to tell this apart from a problem in your account is to look at CPM. If your cost per thousand impressions rose while your click-through rate and conversion rate held steady, the auction got more expensive and your ad is fine. If CPM held steady and click-through fell, the problem is yours.
That single comparison separates most “the platform is broken” cases from genuine account problems, and it takes a minute.
Cause 6: the landing page changed
The ad is only half the journey. A slower page, a new cookie banner covering the form, an extra required field, a broken layout on phones — any of these lower conversion rate while every ad metric stays identical.
The signature here is clicks holding steady while conversions fall. Open the page on an actual phone, on mobile data rather than office wifi, and try to complete the form yourself. It is remarkable how often that is the whole investigation.
A short diagnostic order
Run these in sequence and you will usually have the answer inside an hour:
- Re-compare settled periods, not the last three days.
- Check whether CPM rose. If yes, suspect the auction.
- Check click-through per ad. Falling on one ad is creative; falling across all is audience.
- Check frequency. Climbing means the audience is exhausted.
- Submit your own form and confirm the conversion arrives.
- Check your own edit history against the date the drop started.
Step six finds it more often than anyone likes to admit.
What not to do while you are diagnosing
Do not turn everything off and rebuild. You lose the learning, the history and the ability to work out what actually happened, and you will be back in the same position in three weeks without knowing why. Change one thing, wait long enough to judge it, and write down what you changed and when — a simple dated log of account changes is the most useful diagnostic document most advertisers do not keep.
If you would rather have someone else work through it, I can look at the account and tell you which of these it is.