When Facebook ads that used to bring in enquiries or sales go quiet, the cause is almost always one of six things: tracking that has stopped recording results, a recent edit that sent the ad set back into learning, an ad people have seen too often, an audience with nobody new left in it, a dearer auction, or a change on the page the ad sends people to. Each leaves a different pattern in Ads Manager, so checked in the right order they are usually quick to tell apart.
This post is for UK business owners and in-house marketers who run their own Facebook and Instagram ads, or who want to test the explanation an agency has given them.
Check that results really fell
Before hunting for a cause, make sure there is something to explain. Four quick checks rule out the cases where the ads are fine and only the report has changed.
- Is the account still spending? A declined card, a reached account spending limit, an ad rejected on review or a restricted ad account all stop delivery, sometimes with no more than a small notice in Account Overview. If spend fell to zero on one day, start there.
- Do your own records agree? Count the enquiries, bookings or orders that actually reached you over the same weeks. If Meta reports fewer but your own count has not moved, the measurement has slipped, not the ads. Some gap is normal, and the post on why GA4 and Meta report different conversion numbers explains where it comes from.
- Are the periods comparable? Use two periods of equal length starting on the same weekday, under the same attribution setting. Ads Manager reports a conversion against the date the ad was shown or clicked, and the usual 7-day click, 1-day view attribution setting lets conversions arrive for up to a week afterwards, so the latest few days always look weaker than they will.
- Is it more than normal variation? An ad set producing a handful of leads a week swings between good and poor weeks with nothing wrong. Compare four weeks against the four before. A fall that holds for a month, or a sharp drop on one date, is worth investigating.
Find which number moved
Cost per result is built from three numbers: CPM (what you pay for a thousand impressions), link click-through rate (the share of people who click) and conversion rate (the share of clickers who enquire or buy). Add those columns in Ads Manager and put the good period beside the poor one.
- CPM rose, the other two held: you are paying more to reach people, so look at the auction, the season and audience size.
- Click-through rate fell: people are scrolling past, so look at creative fatigue and frequency.
- Clicks held, conversion rate fell: the problem sits after the click or in the tracking.
When all three shift together, look for an account edit or a tracking change on the date it began, because either can move every number at once.
Tracking has stopped recording results
I check this first because it is quick to confirm and does double damage. When Meta stops receiving the event an ad set is optimised for, the reports go quiet and the delivery system loses the signal it uses to choose who sees the ad, so a reporting fault can become a real performance problem.
The giveaway is results falling sharply on one date while link clicks and landing page views carry on. The usual culprits on UK sites:
- a new cookie banner, or a consent plugin update, that now blocks the Meta Pixel even after a visitor accepts
- a redesign that dropped the pixel code or moved the thank-you page a custom conversion relied on
- a form that now submits without loading a new page, so the thank-you page event never fires
- a disconnected integration, such as a shop platform’s Meta app or a Conversions API connection with an expired access token
To confirm it, open Events Manager, select your dataset and look at the last 28 days of activity for the event your ads optimise for. A cliff on one date points at whatever changed on the site that day. Then make a test enquiry yourself and watch for it in Test Events; my post on testing whether conversion tracking records real enquiries and sales covers the full check.
One UK point matters. Under PECR the pixel should load only after a visitor agrees to marketing cookies, so some missing conversions are expected and lawful. The answer is a banner that records consent properly and a Conversions API setup that sends the events you are allowed to send more reliably, with server-side events following the visitor’s choice rather than working round it.
A recent edit sent the ad set back into learning
After launch, or after a significant edit, Meta’s delivery system spends a while working out who responds before it settles. During this learning phase costs run higher and results swing, and the Delivery column shows “Learning”. An ad set that cannot gather enough results to finish shows “Learning limited”.
Edits that restart learning include changing targeting, the optimisation event or the bid strategy, adding a new ad, making a large budget change and pausing for a week or more. Frequent small edits do the most harm: an account adjusted after every poor day can spend most of its life in learning.
Ads Manager keeps an activity history for every campaign, ad set and ad, logging who changed what and when. Line those changes up against the date results fell; it is also the quickest way to see what an agency or colleague did.
If learning is the cause, make one considered change and leave it alone. If the ad set is stuck in “Learning limited”, it needs more results per week: combine small ad sets with similar audiences, or optimise for a more frequent event closer to the click, such as add to basket instead of purchase, until volume returns.
People have seen the ad too often
An ad that worked for months can wear out. Creative fatigue shows as frequency (the average number of times each person has seen the ad) climbing, link click-through rate sliding, and cost per result rising while CPM barely moves. Break results down by week and look at each ad on its own, not the ad set as a whole.
There is no single safe frequency figure; it depends on audience size, campaign age and how varied the ads are. Local businesses feel it first. A salon or trades business advertising within a few miles of its premises has a small pool of people, so the same faces see the same ad quickly. The reach and frequency calculator shows how fast your budget will cycle through an audience of your size.
The fix is creative that looks new to the person scrolling. A different headline on the same image rarely counts; a fresh angle on the offer, a short vertical video in place of a static image or a new opening line on a video usually does. The delivery system uses the ads themselves to decide who to show them to, so several genuinely different ads give it more to work with than five versions of one idea.
The audience has run out of new people
Fatigue is about the ad; audience saturation is about the people. Once most likely buyers in an audience have seen your offer and acted or decided against it, even fresh creative gives a short lift that fades within days. Reach levels off while frequency keeps rising.
Narrow audiences hit this first, especially retargeting, whose size depends on traffic from other channels. If organic search or a Google Ads campaign dropped a month ago, the pool of recent visitors shrinks and the retargeting ads weaken weeks later with nothing changed inside Meta. Visitors who decline marketing cookies are not added to website audiences either.
The fix is more people to reach: widen the radius or age range where the offer genuinely suits them, test broad targeting with Advantage+ audience, run a prospecting campaign that refills the retargeting pool, or refresh the customer list behind a lookalike. If you upload a customer list, check your privacy notice tells customers their details may be used this way.
The auction has become more expensive
Sometimes nothing is wrong with your account. You buy impressions in an auction against every advertiser chasing the same people, so when they spend more, your CPM rises. In the UK the clearest case is mid-November to Christmas, when Black Friday and gift advertising push costs up across the board, even for businesses selling nothing seasonal. Valentine’s Day and Mothering Sunday can have a smaller version of the same effect.
Your own demand moves too: garden services slow in winter, and many B2B firms see enquiries thin out in August and over Christmas. If the account has a year of history, compare the same weeks last year before deciding something has broken.
The Meta Ad Library shows every active ad a Facebook Page is running, so you can see whether a rival has started advertising heavily to your local audience. If CPM is up but click-through and conversion rates have held, the ads are working in a pricier market; decide whether the result is still worth the cost rather than rebuilding campaigns that work.
Something changed after the click
If clicks hold steady but enquiries fall, follow the customer’s path. Open the ad on your own phone, complete the form or checkout, and check the enquiry or order actually reached you. This turns up forms that stopped sending after a plugin update, cookie banners covering the submit button on small screens, pages slowed by new images, ads pointing at URLs that now redirect or error, offers that changed without the ad being updated, and featured products that sold out.
A falling ratio of landing page views to link clicks suggests the page has slowed or fails to load for some visitors. Run through the landing page checklist for paid traffic before you touch the ads.
The order to work through
| What you see | Likely cause | Where to confirm it |
|---|---|---|
| Spend at or near zero | Billing, spending limit, rejected ad or account restriction | Account Overview and delivery status |
| Results fell on one date, clicks did not | Tracking stopped recording | Events Manager and Test Events |
| Costs jumped straight after a change | Ad set back in learning | Activity history and Delivery column |
| Frequency up, click-through rate down | Creative fatigue | Weekly results for each ad |
| Reach flat, new ads fade within days | Audience saturation | 30-day reach and frequency, retargeting audience size |
| CPM up, click and conversion rates steady | Seasonal or competitive auction pressure | Same weeks last year, Meta Ad Library |
| Clicks steady, conversion rate down | Landing page, offer or form | Your own test enquiry on a phone |
Work from the top. The early rows are quick to rule out, and when one of them is the cause, nothing lower down will help until it is fixed.
What to avoid while you investigate
- Changing several things at once. You will not know which change helped, and each significant edit restarts learning.
- Switching everything off. Ad sets paused for a week or more restart in learning. If money is clearly being wasted, pause the weakest ad or ad set, not the account.
- Duplicating ad sets to “reset” them. Copies compete for the same people and split the results each has to learn from.
- Raising the budget to push through. More money behind a struggling ad set usually buys more of the same results at a higher cost each.
- Loading the pixel before consent. It breaks PECR, which the ICO enforces; the fix belongs in the consent and tracking setup.
- Moving the whole budget elsewhere overnight. If Meta looks like the wrong fit for what you sell, weigh it properly first; comparing Facebook Ads with Google Ads sets out how to decide.
When to bring someone in
If the cause is still unclear after the table, or the fix involves the site’s tracking and consent setup, a second pair of eyes can save weeks of guessing. Have the activity history, the Events Manager graph for your main conversion and your own enquiry or sales count for the last two months ready. Diagnosing drops like this is part of the Facebook and Instagram ads work I do for UK businesses, and you can book a consultation to talk through what you are seeing.