Ads Not Converting
Facebook ads not converting? How to tell if it's tracking, delivery or your site
Three comparisons, one for each place a Meta campaign breaks, so you fix that one thing instead of rebuilding the campaign.

Short answer
- When Meta ads spend but stop converting, one of three things broke: tracking, delivery or the website. Check each with one comparison before you rebuild anything.
- Tracking: if your store's orders held steady over the same 14 days that Meta's results fell, Meta most likely lost the signal while the sales still happened.
- Delivery: read spend, CPM and frequency by day and placement over a full week, because Meta may spend up to 75% over a daily budget on some days.
- Site: compare link clicks with landing page views, as Meta itself suggests. A widening gap means people leave before the page loads.
- An ad set usually leaves learning after about 50 results in the week after its last significant edit, and frequent budget changes can send it back.
Facebook ads that keep spending but stop converting have broken in one of three places: tracking, delivery or your website. Each one shows when you put one report against another, so this guide walks through the three comparisons in order, starting with your store's own orders.
We run a performance agency and build AdCtrl, which reads Meta, GA4 and Tag Manager from Claude or ChatGPT, so some steps below have a route through it. If Meta and GA4 simply never agree on the count, that's a different problem, covered in why conversions don't match. If the ads get no impressions at all, see Meta ads not delivering.
Checked 12 November 2026 against Meta's Business Help Centre and developer docs1.
Why are my Meta ads not converting?
Meta ads stop converting when Meta stops seeing sales that still happen, when Meta starts spending the budget differently, or when people click and then don't buy on the site. Find out which one before you change anything, because each has a different fix and usually a different person who can make it.

The triage in one picture: compare store orders first, then delivery by day and placement, then each step from click to purchase. The first pair that broke is the branch to fix.
Many "suddenly stopped converting" threads skip this and go straight to new creative, a new audience or a rebuilt campaign. A rebuilt ad set usually needs about 50 results in a week to leave learning2, and it does nothing for a broken pixel or a slow checkout.
Is it tracking, delivery or the site?
The quickest way to tell tracking, delivery and site problems apart is to compare two numbers that normally move together and see which pair stopped. Each row below is a symptom, the comparison that confirms it and who usually fixes it.
| What you see | Branch | Compare | Who usually fixes it |
|---|---|---|---|
| Store orders steady, Meta results down | Tracking | Meta results vs store orders, same dates and attribution setting | Whoever runs the pixel and Conversions API, often a developer |
| Spend, CPM or frequency moved first | Delivery | Spend, CPM and frequency by day over a full week | The person running the ad account |
| One placement took the budget | Delivery | The same metrics by placement | The person running the ad account |
| Delivery column shows Learning limited or Creative fatigue | Delivery | Delivery column and Last significant edit | The person running the ad account |
| Link clicks steady, landing page views down | Site: load or redirect | Link clicks vs landing page views | Developer or site host |
| Landing page views steady, checkouts down | Site: page, offer or checkout | Landing page views vs checkouts initiated vs purchases | Site or store owner |
Each symptom has a pair of numbers that should move together. The pair that broke names the branch and the person who can fix it.
Source: Meta Business Help Centre on landing page views, the learning phase, creative fatigue and performance fluctuations1, 2, 3, 4; the pairings are our method.
Run the rows top to bottom. Tracking comes first because a tracking break makes every result-based number lie: if Meta can't see the purchases, results, cost per result and the learning phase all look broken while spend and clicks look normal.
Did Meta stop seeing sales your store still gets?
When your store's orders hold steady on the same days that Meta's results fall, Meta has most likely stopped seeing sales that still happen. That points to tracking, and changing the ads won't fix it. The first two checks need only store access; the Events Manager checks are for whoever manages the pixel.
Line up the dates first. Meta counts a result inside the ad set's attribution setting, for example within 1 or 7 days after a link click, or 1 day after an impression5.
Your store counts an order on the day it was placed. Compare the same dates, in the same time zone, and keep the attribution setting in mind. The full rules for comparing the two counts are in why conversions don't match, so here we only use the trend.
- 40 → 40 example: store orders a day, week 1 vs week 2
- 40 → 12 example: Meta results a day, same two weeks
In this made-up example the store kept taking about 40 orders a day for two weeks, while Meta's results fell from about 40 to about 12 a day in the second week. The sales kept coming while Meta stopped seeing them, so that's tracking. If both had fallen together, the sales really dropped, and delivery or the site is where to look next.
Read Meta's results by day.
Manual: in Ads Manager, set the date range to the last 14 days, split by day, and note the attribution setting each ad set uses.
With AdCtrl: write "Show Meta spend and results by day for the last 14 days, per campaign."
Count orders in your store for the same days.
Manual: in your store admin, export or filter orders by day for the same 14 days and time zone. Do the same for GA4 purchases if you track them.
With AdCtrl: write "Show GA4 purchases by day for the last 14 days." Store orders stay in your store admin, because AdCtrl has no store connection.
Check that Events Manager is still receiving purchases.
Manual: in Events Manager, open your pixel and select Overview. It counts the events Meta received before they're deduplicated, discarded because of consent controls and other policies, or processed. Connection Method shows the channel each event came through. Events sent through the Conversions API should be verifiable within 20 minutes, and the Event Freshness tab (under Event Details for each event) shows the average delay on a scale from Real Time to Weekly.
Check deduplication.
Manual: when the pixel and the Conversions API both send the same purchase, Meta needs a way to tell one purchase sent twice from two different purchases. That's deduplication. Open Event Details for Purchase, then the Event Deduplication tab. The tab shows the percentage of events deduplicated from each source; higher is better, and Meta shows a warning when the rate is too low.
Check Event Match Quality.
Manual: Event Match Quality is Meta's score for how well the customer information sent with your server events can match those events to a Facebook account. It runs from 1 to 10, and Meta suggests aiming for 6.0 or higher where possible6.
If orders held and Meta's results dropped on a specific day, look at what changed on the site that day: a new theme, a checkout app, a consent banner or a tag edit. The tracking checker shows which pixel, Tag Manager and consent tags a page loads, so you can compare before and after. The fix sits with whoever manages the pixel, and the Conversions API setup has its own guide in Meta Conversions API.
If Shopify has paid orders your Analytics report is missing, use the Shopify GA4 purchase guide to trace the checkout before changing the ads.
Did delivery change: spend, CPM, frequency or placement?
Meta can spend the same budget very differently from the weeks that converted: more on some days, at a higher CPM, on a placement that doesn't sell, or with the ad set back in learning. Read a full week, not a single day. Meta says results broken down by hour swing up and down and average out over time, and suggests reading breakdowns over a full week4.
- 75% how far over a daily budget Meta may spend on some days
- 7× the daily budget, Meta's weekly spending ceiling
- ~50 results an ad set usually needs in the week after its last significant edit to leave learning
Meta is gradually introducing this daily budget flexibility to some Ads Manager accounts. On those accounts a day far over budget with no sales can be normal, as long as the week stays under seven times the daily budget. A week with the same spend and half the sales is a real change.
CPM alone is a weak verdict here. When an ad set optimises for conversions, cost per thousand impressions may not be a good indicator of performance, so read it next to results and cost per result. If CPM itself jumped, why your Meta CPM is so high shows how to tell whether the market, the audience, the ad or the goal moved it.
Split spend by day for the whole week.
Manual: in Ads Manager, use the Breakdown menu to split the campaign by day. Look at spend, impressions, CPM, frequency and results together.
With AdCtrl: write "Show Meta spend, impressions, CPM, frequency and results by day for the last 14 days."
Split the same week by placement.
Manual: use the Breakdown menu again to split by placement. Look for a placement that took a large share of spend with few results.
With AdCtrl: write "Which placements on Meta have high spend and no results?"
Read the Delivery column and the last edit.
Manual: in the ad set view, read the Delivery column for Learning or Learning limited, and add the Last significant edit column to see when you last made a change that could reset learning2.
For anyone pulling this through the Marketing API: the Insights API calls the two placement breakdowns publisher_platform, which platform the ad ran on (Facebook, Instagram or Audience Network), and platform_position, where within it, such as the Facebook desktop feed or Instagram mobile feed7.
The Meta ads MCP guide compares the other ways to pull these reports into an AI chat.
If one ad set took most of a shared budget, compare campaign-level results first. The Meta campaign budget guide shows when to test a different split.
Are clicks turning into page loads and checkouts?
Clicks turn into sales in three steps on your site: the page loads, people start checkout, and they finish it. Meta counts the first step for you. A landing page view is a click on an ad that successfully loaded the destination page, and Meta itself suggests comparing landing page views with link clicks to see how many people clicked but left before the site loaded1.

Screenshot of Meta's Business Help Centre, 11 November 2026. Meta's own definition of landing page views suggests comparing them with link clicks to see who left before the page loaded.
Meta's troubleshooting page puts it the same way: when someone clicks a link but closes the page before it loads, that counts as a link click but not a landing page view8. So the ratio between the two is your first site check.

Where the drop happens names the fault: before the page loads, it's speed or a redirect; between the page and checkout, it's the page or the offer; inside checkout, it's the checkout itself.
Read each step against the one before it, for the same dates as the delivery read:
- Link clicks to landing page views: a widening gap means people leave before the page loads. Check page speed on a phone and any redirect between the ad and the page. That's a job for a developer or the site host.
- Landing page views to checkouts initiated: people see the page and don't start buying. Check that the page matches the ad and that price and shipping are clear. The site or store owner can do this.
- Checkouts initiated to purchases: people start and give up. Test the checkout yourself on a phone, with each payment method.
We don't know of a normal click-to-page ratio from Meta, so we compare an account with itself: the ratio in the weeks that converted against the ratio now.
Meta tracks landing page views through its in-app browser or the Meta pixel, and where data is partial it may use statistical modelling for some of them1, so treat the ratio as a pointer. GA4 landing-page sessions for the same dates are a useful second opinion.
Show GA4 sessions and key events by landing page for the last 14 days, and list the landing pages with the most sessions and the fewest key events.Why did my ads stop converting after a strong first week?
Ads that convert well for a week and then stop usually hit one of four things Meta documents: the creative wore out, an edit reset learning, too few sales landed inside the attribution window, or spend shifted between days. Meta's own statuses tell you which, so read them before you guess.
Meta counts every recent exposure of the image or video for fatigue, including other campaigns from the same Page. So an ad can fatigue faster than its own frequency suggests if the same video runs elsewhere. The fatigue status only shows for ad sets with one creative, not for dynamic creative or Advantage+ catalogue ads. To test an ad in an ad set with several ads, see Facebook ads creative fatigue.
Should you rebuild, keep scaling or wait?
Whether to rebuild, scale or wait depends on which of the three comparisons broke: tracking, delivery or the site. In most cases the answer is to fix that one and leave the campaign alone. Scaling is the risky one after a drop: Meta advises against frequent budget changes because they can send an ad set back into the learning phase2. Change one thing, then give it a week.
Fix tracking, keep the ads if
- Store orders held while Meta's results fell
- Events Manager shows fewer events than in the weeks that converted
- The drop started the day the site or tags changed
Wait if
- The ad set is in learning after a recent edit
- Daily spend swung, but the week stayed on budget
- Results have been down for less than a full week
Change creative or the page if
- The Delivery column shows Creative fatigue or Creative limited
- Landing page views hold but checkouts don't start
- A full week of steady delivery still brings no sales
A full rebuild makes sense only after the reads rule out tracking and the site, and the creative is worn out across the board. Use the Meta ad performance comparison to identify what moved between two periods. The ad preview tool checks the replacement copy and shows where controls can cover a vertical video on Reels.
What if tracking, delivery and the site all look fine?
If tracking, delivery and the site all check out, we'd look at the ad and the offer next. Give it the full week from the delivery read first, so one bad day doesn't decide it.
Then check whether the ad has worn out. In ad sets with one creative, the Delivery column shows Creative limited or Creative fatigue when cost per result climbs above your past ads3. If neither shows, compare the offer with the weeks that sold: price, shipping, stock and any promotion that ended. Those changes don't show in Ads Manager, and they're the ones we'd ask the store owner about first.
Run the triage in one chat
The triage fits in one chat when the delivery and site reads come back together, instead of in separate Ads Manager and GA4 tabs. AdCtrl compares ad results with website traffic in one answer: a Meta performance report by placement next to a GA4 landing page report, so you can see which pair broke. Every prompt below only reads; nothing changes in the account.
For the last 14 days, show Meta spend, results, CPM and frequency by placement, then GA4 sessions and key events by landing page for the same dates, and tell me which pair moved first.
Made-up example data, not a real account, in AdCtrl's report card. This example account optimises for leads, so leads are its results: one row per placement over two weeks, with spend, leads, CPM and frequency. In this example account, Audience Network is cheap per impression and brings only 2 leads.
Some reads stay manual. AdCtrl doesn't read Events Manager diagnostics such as Event Match Quality or the deduplication rate, it has no connection to Shopify or any other store backend, and it doesn't fix the pixel or the checkout. A Tag Manager review shows how the tags are configured, not whether one fired.
Checklist before you change the campaign
Run this checklist before you edit, rebuild or scale a Meta campaign that stopped converting. Each line is a yes or no, and a single no is where to start.
- Store orders and Meta results are compared for the same dates and time zone.
- You know the attribution setting each ad set uses.
- Events Manager shows purchase events arriving at the usual volume.
- Event Match Quality and the deduplication rate look the same as in the weeks that converted.
- Spend, CPM and frequency are read by day over a full week.
- The same week is split by placement.
- The Delivery column and Last significant edit are checked for every ad set that dropped.
- Link clicks and landing page views are compared for the same dates.
- Checkouts initiated and purchases are compared, and the checkout is tested on a phone.
- Only one change is planned, with a week to read its effect.
Frequently asked questions
How do I tell a real sales drop from a tracking drop if I get leads, not orders?
Use your own record of leads in place of store orders (if those leads look fake, see where fake Facebook leads come from): the enquiries in your CRM or inbox for the same days, and GA4 key events for form submissions if you track them. If your leads come through a form on your site and they kept arriving while Meta's results fell, the reading is the same as for a store: it most likely points to tracking.
Is a high CPM the reason I'm not getting sales?
Not by itself. For ad sets optimised for conversions, CPM may not be a good indicator of performance. Read it next to results and cost per result, by day and by placement.
Can Events Manager show events while tracking is still broken?
Yes. The Overview counts events received before they are deduplicated or discarded6, so a healthy-looking total can still hide duplicates or events with weak match quality. Check the deduplication rate and Event Match Quality as well.
Why don't landing page views match GA4 sessions?
The two tools count differently. Meta counts landing page views through its in-app browser or the pixel, and may model some of them, while GA4 counts sessions with its own tag. Compare the trend over the same dates, not the exact numbers.
Does AdCtrl fix my pixel?
No. It reads Meta reports, GA4 and your Tag Manager configuration so you can see where the break is, but the fix to the pixel, the Conversions API setup or the checkout is done on your site, in Events Manager or in Tag Manager, where you publish the change yourself.
What to do next
Start with the store: put your orders next to Meta's results for the same recent dates. If they match, read delivery by day and placement for a full week, then link clicks against landing page views. Fix the first pair that broke, change one thing and give it a week. If Meta reports more sales than your store took, check both platforms against store revenue with blended ROAS.
To go deeper, why conversions don't match explains the counting differences between Meta and GA4, and Google Ads not converting runs the same triage for Google. To run these reads in Claude or ChatGPT, see how to connect AdCtrl and start the free trial.
Sources
- Website landing page views (Ads Manager metric), Meta. Accessed 12 November 2026.
- About the learning phase, Meta. Accessed 12 November 2026.
- About creative fatigue recommendations in Meta Ads Manager, Meta. Accessed 12 November 2026.
- Understand fluctuations in ad performance, Meta. Accessed 12 November 2026.
- About attribution models and attribution settings, Meta. Accessed 12 November 2026.
- Conversions API: Verifying Your Setup, Meta. Accessed 12 November 2026.
- Insights API: Breakdowns, Meta. Accessed 12 November 2026.
- Troubleshoot conversion optimisation, Meta. Accessed 12 November 2026.
