· 15 min read
Facebook Ads Not Converting: A Triage Order That Works
When Facebook ads stop converting, the fix order matters: falsify tracking first, read the click, cart and purchase ratios, then touch your creative last.
When a Meta account stops producing sales, the instinct is to touch the most visible thing first: the creative. New images, new hooks, a new angle — refresh the ads and hope. Sometimes that is exactly right. More often it burns a week of spend on a problem that was never in the ad, and it destroys the one thing a diagnosis needs: a stable set of numbers to read.
This guide is a triage order, not a list of tips. The order exists because every stage’s reading depends on the stages before it. You cannot interpret an add-to-cart rate if tracking is unreliable. You cannot judge creative if the click itself is being measured wrong. You cannot blame the landing page until the ad has been ruled out. So the work runs through four gates in sequence: falsify the measurement, read the three stage ratios, only then touch the creative, and take the landing page seriously last — because it is last in the chain, not because it rarely matters.
One boundary note before the order: this is a diagnosis of the ad account, not of the store. If the store has never sold consistently, start with the store-side triage in the guide to a Shopify store that is not making sales, because no amount of ad work rescues a page that cannot close a willing buyer. What follows assumes the store can convert, and that the paid traffic has stopped producing.
Gate 1: Falsify the Measurement Before Anything Else
“Conversions” in a Meta report are not sales. They are events the pixel or the Conversions API reported, matched to ad views inside an attribution window. Every diagnosis starts by separating the two, because a broken measurement produces two opposite-looking failures that get treated identically: purchases that exist but are never reported, and purchases that are reported but never happened.
The cheapest falsification test costs nothing and takes five minutes: compare orders in your own store backend against purchase events in Meta’s reporting for the same seven days. They will never match exactly — attribution windows, refunds, and matching limits guarantee that — but the gap has a shape.
| Pattern | What it usually means | What it does not mean |
|---|---|---|
| Backend orders far above Meta purchases | Tracking is undercounting; the funnel may be fine | The ads failed |
| Meta purchases above backend orders | Double counting, or view-through inflation | The ads are working |
| Both roughly agree and both are low | Measurement is probably honest; the funnel itself is the suspect | Nothing — keep checking the next gate |
Undercounting has a short list of usual causes, all checkable with free tools. The purchase event may simply not exist on the thank-you page, or fires on a domain the pixel is not attached to. If you run both a browser pixel and the Conversions API, the two can fire the same purchase twice unless they share an event ID for deduplication — which shows up as Meta reporting more purchases than your backend has orders. Browser tracking also degrades quietly: an increasing share of buyers blocks or limits cookies, so an account that once matched its backend closely will drift apart over months without anything changing in the setup.
The overcounting direction is more dangerous for diagnosis, because it makes a healthy-looking account rot invisibly. View-through conversions — people who saw an ad, did not click, and bought later — get credited inside some attribution settings, and they can be a large share of reported purchases for retargeting campaigns. If most “converting” traffic is view-through, the account reports sales while clicks produce nothing, and any click-based metric you read downstream is fiction.
Two free checks settle most of it. Meta’s Events Manager includes a test-events view where you can browse your own site, trigger a test purchase, and watch each event arrive with its parameters — event name, value, currency, deduplication ID. Second, compare the event count over a long window against your backend over the same window; a persistent 30–50% gap in the same direction is a measurement finding, not a funnel finding.
Only when backend orders and reported events agree in direction and roughly in magnitude is the measurement honest enough to support the next gate. If it is not, fix that first and skip everything else for a week — every ratio below inherits the error.
Gate 2: Read the Three Stage Ratios in Order
With measurement trusted, the funnel reduces to three ratios, each answering one question:
- Click-through rate (CTR) — does the ad earn the click?
- Landing-to-cart rate — does the page turn a click into add-to-cart intent?
- Cart-to-purchase rate — does checkout close the sale?
The discipline is to find the first broken stage and fix only that one. An upstream fix changes every downstream number, so fixing two stages at once means you never learn which change worked, and the new numbers cannot be compared to the old ones.
No universal thresholds follow — “a good CTR” has no honest cross-industry number, and any table of them you find is either invented or harvested from accounts that have nothing to do with your price point. The baseline that matters is your own account’s history: what this campaign, this audience shape, and this product delivered during its best four weeks. Read each ratio against that.
| Stage ratio | Read against | A real drop usually means | It never means |
|---|---|---|---|
| CTR | Your own last 90 days on the same audience shape | The opening promise stopped matching what the audience wants, or the audience got tired | The landing page is bad |
| Clicks to add-to-cart | Your own history for the same product page | The page broke the promise the ad made, or added friction | The ad is weak |
| Cart to purchase | Your own checkout history | Checkout friction: surprise costs, forced accounts, missing payment methods | The ad or the page is bad |
Two refinements make the readings sharper. First, watch CTR together with CPM: a falling CTR on a stable CPM points at the creative or the audience; a stable CTR on a rising CPM points at the auction — more bidders chasing the same people — and no creative refresh fixes that. Second, segment the ratios by placement where your reporting allows it. An account can hold a healthy blended CTR while one placement quietly converts nothing; the blend hides the leak, and turning that placement off is a cheaper fix than any new campaign.
The cart-to-purchase stage deserves one extra caution. A low closing rate on a warm cart is often not a marketing problem at all — it is shipping cost, payment method coverage, or a checkout that demands an account. Those belong to the store-side diagnosis, not to the ad account, and the fastest way to confirm it is to place one test order yourself on a phone.
Gate 3: Only Now Touch the Creative
If the first broken stage is CTR, creative is finally on the table — but as a read, not a repaint. Three checks come before any new asset is made.
Frequency first. Rising frequency against falling CTR is fatigue, and the fix is a new opening promise, not a new campaign. The sequence and the observation window are laid out in the guide to ad fatigue signs; the short version is that frequency moves first, CTR responds to repetition second, and cost per result only reacts downstream. Diagnosing fatigue from the CPA line alone means acting a week or two late.
The offer, not the picture. Most CTR problems are specificity problems. An ad that says the category (“comfortable running shoes”) competes with every other ad in the auction saying the same thing; an ad that names the claim (“the pair our testers returned least”) gives the scroll a reason to stop. The hook guide treats the opening line as an observable asset — which competitors repeat across ads, how long they keep it live — and the copy guide handles how the promise is budgeted across the three fields Meta actually renders.
The external prior. Before paying to test a new angle, read what the niche has already voted on. A competitor’s longest-running live ad is the cheapest research available: it survived auctions you did not have to pay for. Scanning live ad counts, the oldest live creatives, and how long promises stay up turns creative work from guessing into updating. That read is exactly what a competitive ads database is for — TrendTrack indexes 10M+ Meta ads, and the free methods for reading them are covered in the guide to spying on competitor ads, with the weekly version in the competitor tracking workflow.
And if the first broken stage was not CTR, changing creative is not just useless — it is expensive, because a new ad resets the numbers you were reading and the diagnosis starts over.
One honest note on cost. The free extension and the free account cover every check in this guide, and the paid tiers only earn their keep once reading competitors becomes a weekly job rather than a one-time look. If that is where you land, the TrendTrack promo code is WINNING - 20% off all plans for the first 3 months, after which billing returns to the standard rate.
Verified 2026 · 20% off Starter, Pro and Business for your first 3 months
Gate 4: The Landing Page Comes Last but Often Wins
When clicks are healthy and add-to-cart has collapsed, the ad is exonerated and the page is the suspect. Four checks, all free, in the order they kill conversions:
Message match. Read the ad’s promise and the page’s first screen as a stranger would. If the ad sells “free shipping on your first order” and the page opens with a seasonal collection banner, the visitor has to do the matching work themselves — most do not. The promise the click was bought with has to be visible without scrolling.
The price, early. Pages that hide the price behind a size selector, a quiz, or a shipping calculator convert visibly worse on paid traffic than pages that state it, because paid visitors arrive with less patience than organic ones. If you would not put the price in the ad, at least put it on the first screen.
Speed on a phone. Most paid clicks are mobile. Load the page on a mid-range phone on cellular, not on your desktop on fiber, and watch what renders before a reasonable person gives up. Slow hero images and blocking scripts cost you the top of the page exactly where the promise should be.
Friction after the click on “add to cart”. Forced account creation, a coupon field that invites a search elsewhere, or a shipping cost revealed only at the last step — each one leaks a cart that the ad already paid for.
One caution: a page change is a funnel change. Make one at a time, and give each one the same honesty you would give an ad test — the creative testing guide is about ads, but the sampling arithmetic applies unchanged to page variants, because a verdict needs trials no matter which element you are judging.
The 30-Minute Triage Order
| Minutes | Check | If it fails |
|---|---|---|
| 0–5 | Backend orders vs reported purchases, same 7 days | Stop. Measurement gate. Do not read any ratio yet |
| 5–10 | Test purchase in Events Manager; event fires once, with value | Fix dedup or the missing event before anything else |
| 10–15 | Read CTR, click-to-cart, cart-to-purchase against your own history | Identify the first broken stage; ignore the rest for now |
| 15–20 | If CTR is the break: frequency trend, then promise specificity | New opening promise; check fatigue before rebuilding |
| 20–25 | If click-to-cart is the break: message match, price visibility, mobile speed | Page fix, one change at a time |
| 25–30 | If cart-to-purchase is the break: place a test order yourself on a phone | Checkout friction or unit economics — see below |
The order is the point. Run backward — new creative first, tracking never — and you will keep paying for refreshes while the actual leak drains the account from a gate you never opened.
Three Hypothetical Diagnoses
All three are invented but arithmetically consistent, to show how the order changes the outcome.
Case A: the numbers were fiction
An account “stops converting”: purchases drop from the low twenties a week to four. Backend orders for the same week: 38. Reported purchases: 4. The store is selling; Meta is nearly blind. In Events Manager, the test purchase shows the browser pixel firing page views but no purchase event — the checkout was migrated to a new domain two weeks earlier and the pixel was never reattached. Nothing about the ads, the audience, or the offer changed. Fix the event, and the funnel reads exactly as it did before. Total creative work required: none. Anyone who started at Gate 3 would have spent the week rebuilding ads for a tracking bug.
Case B: the click was fine, the page broke the promise
Link CTR holds at 2.1%, inside its own 90-day range, on 12,000 clicks over two weeks. Click-to-cart, normally about 11%, reads 4%: roughly 480 carts where 1,320 used to arrive. Cart-to-purchase is unchanged at 3%. The first broken stage is clearly the page, not the ad. The ad promised “restock of the sold-out pack, shipping included”; the page opens with a new-season banner and no mention of the pack. The fix is one line in the hero section. Two weeks later click-to-cart is back near 10% and purchases recover proportionally — with the same ads nobody touched.
Case C: the audience got tired, not the offer
Over five weeks, frequency climbs 1.4 → 1.9 → 2.6, CTR slides 1.9% → 1.7% → 1.3%, and cost per purchase drifts up behind them — the classic sequence, in order. Click-to-cart and cart-to-purchase never move, so the page is fine. The promise has been shown to the same audience eleven times a month. The fix is a new opening promise for the same offer — cheaper than a new campaign, faster than a new product, and it works because the diagnosis ran in order instead of starting at the most visible lever.
What Triage Cannot Fix
Sometimes every gate passes and the account still does not convert, because the failure is in the arithmetic, not the funnel. If the product’s margin only supports a break-even cost per purchase that the funnel would need an implausible conversion rate to hit, no amount of triage saves the campaign. That check is one division, and the break-even ROAS calculator walks through it — margin structure in, minimum conversion rate out. If the required number is far above anything this funnel ever delivered in its best month, the honest options are repricing the offer, cutting the cost line, or turning the campaign off. Triage finds where the funnel leaks; it cannot make a losing offer win.
It also cannot see the auction. Two accounts can run identical ads at identical CTRs and pay very different CPMs, because what surrounds your ad in the feed is other people’s budgets, not yours. The closest free proxy for competitive pressure is again the ad libraries: how many advertisers are live on a promise, and how long they have held it.
FAQ
Why do my Facebook ads get clicks but no sales?
Clicks without sales mean the first broken stage is after the click, so the ad has already done its job. Read click-to-cart and cart-to-purchase against your own history: if carts are scarce, the landing page broke the promise the ad made; if carts are plentiful and purchases are scarce, the problem is checkout friction or unit economics rather than anything the ad account controls.
How do I tell if my pixel is underreporting purchases?
Compare orders in your own store backend against purchase events in Meta’s reporting for the same week. A persistent, one-directional gap that grows over months means tracking is decaying, not that the funnel collapsed. Confirm with a test purchase in Events Manager and check that browser and server events share a deduplication ID so the same order is not counted twice.
Should I fix the creative or the landing page first?
Whichever the first broken stage points to, and never both at once. If CTR has dropped against your own history, the ad is the suspect. If CTR is stable and click-to-cart fell, the page is the suspect. Fixing both together means you cannot tell which change moved the numbers, and the next diagnosis starts from a baseline you no longer understand.
What CTR means a Facebook ad is failing?
There is no honest universal number, because CTR moves with price point, audience temperature, placement mix and format. The usable baseline is the same campaign on the same audience shape during its best four weeks. A meaningful fall against that baseline, especially while frequency climbs, says the opening promise has stopped earning attention for this group.
Can an ad have a good click-through rate and still lose money?
Yes, in two common ways. The page or checkout can leak the traffic the ad paid for, which shows up as healthy CTR with collapsed cart or purchase rates. And the account can be profitable on paper through view-through credit while the clicks themselves produce nothing. Both are why diagnosis starts at measurement and the stage ratios instead of the CTR line alone.
How long should I wait before judging a fix?
Long enough for the sample to support the verdict, which depends on the baseline rate and the daily volume rather than on a fixed number of days. The sampling arithmetic in the creative testing guide gives the trials a decision needs. Changing several elements at once and judging in two days is the fastest way to make a fixable funnel look permanently broken.