Facebook Ads Not Converting? Work the Checks in Order, Not All at Once

2026-09-18

Search this problem and you get lists. Ten reasons. Nine steps. Five main causes. They are not wrong, exactly — broken tracking, a mismatched objective, weak creative, and a slow landing page can each derail performance under the right conditions. A list can even be ordered, by how often the author believes each cause bites or by which scenario you recognise. What an ordered list still does not tell you is which unresolved prerequisite makes the later items unreadable. If your conversion event is not firing, then every conclusion you draw about creative fatigue, audience quality, or page copy rests on a number that was never counting the thing you think it counts. You can work the list from top to bottom and still end up further from the answer than when you started.

So this article does something different. It gives you a sequence ordered by dependency, it tells you where to stop, and it tells you when each check does not apply to you. A check you have already answered is wasted effort. Skip a prerequisite too early and every later diagnosis rests on evidence you never established.

First: say which number is broken

"Not converting" is not a diagnosis. It is a complaint that could describe at least four different situations, and until you name yours you cannot pick a starting point.

Write down exactly what behaviour counts as a conversion and what population it is divided by, before you use any conversion rate in a decision. A purchase? A form submission? A booked call that the prospect actually attended? And divided by what — impressions, clicks, landing page views, sessions? Two people in the same meeting can say "conversion rate" and mean numbers that differ by a factor of ten. If nobody in the conversation can state the definition and the denominator, the number is not ready to act on yet.

Then get specific about the shape of the failure:

• Clicks are healthy and the recorded conversions are near zero.
• Conversions exist but the economics do not work at any volume.
• It worked, then it stopped.
• It works at low spend and falls apart when you increase it.

These are different problems and they enter the sequence below at different points. Naming yours is the first thing that saves you money.

The order, and why it is an order

When no single constraint is obvious, work these stages in sequence and stop at the first one that fails:

1. Viability — does a qualified, reachable, willing market exist at economics this business can support?
2. Measurement readiness — is the data real enough to make decisions on?
3. Offer, audience, and message — does the right person understand what this is and why it is for them?
4. Traffic — is the right person arriving, in enough volume to learn from?
5. Page and funnel — does the path from arrival to purchase hold together?
6. Email and lifecycle — is the value of people who did not buy immediately being captured?
7. Economics and scale — do the unit economics support spending more, and can fulfilment absorb it?

Each stage is a precursor to the ones after it. That is the reason the order matters and the reason a flat list of causes fails you: a defect at stage 2 makes any finding at stages 4 through 7 unreliable, so diagnosing your creative before verifying your events is work you will have to redo.

Stage 1 — Viability: is there a buyer at a price you can afford to reach

Count how many people meet every condition required to buy, and check whether reaching them costs less than they are worth. This is the check to run when acquisition metrics have been optimised repeatedly without revenue following, when nobody has ever produced that number, or when the audience is described by category — "women over 55," "small business owners" — rather than by the conditions that qualify someone to buy.

Skip this if you already have repeat customers at an acquisition cost your margin supports. Viability is answered; your constraint is downstream.

Stage 2 — Measurement readiness: establish the data is real before you interpret it

This is the stage a flat list can put anywhere. It belongs second, because everything after it depends on it. Each check below carries its own boundary.

Run the setup check before you diagnose anything

Before interpreting any performance data, verify the date range, timezone, filters, campaign status, destination URL, which variant is actually live, the event configuration, and the sample size. Run this when you are about to diagnose from a dashboard, when something looks suddenly and dramatically wrong, or when a campaign appears inactive. Skip it if you already ran it on this dataset and nothing has changed since.

Fire the event yourself

Complete the action on a real device and confirm the event arrives — before concluding anything from its absence. If it never lands, the number you have been optimising against was never measuring that action. Do this when an expected event is not appearing, when conversions dropped to zero, or when a live variant is reporting nothing. Skip it if the event has been verified since the last change to the page, the funnel, or the tracking.

Reconcile the whole count chain

Line up impressions → clicks → page views → sessions → actions → leads → purchases for the same period, and investigate the largest unexplained gap. This is the check for the "clicks but no conversions" version of the problem, and it is the one that tells you whether people are even arriving. A stage showing a loss far larger than the stage before it is where to look. Skip it if the chain has been reconciled for this period and the gaps are explained.

Trust the system that receives the outcome

For every outcome, identify which system is authoritative — the CRM, the calendar, the checkout, the payment processor — and reconcile Meta's reporting against it. Do this when the ad platform and your own records disagree, when you are about to act on platform-reported conversions, or when someone says there are no leads. Skip it if platform numbers and first-party records already agree within a margin you understand. If your records are themselves the doubt, how to tell whether your pipeline numbers are real covers that reconciliation in more depth.

Check whether clicks are becoming page views

Measure page load on a real mobile device and check redirects and script placement before attributing weak engagement to the content. This is the check when clicks are not becoming page views, when a behaviour tool shows far fewer sessions than the ad platform shows clicks, or when the page is heavy, image-rich, or sits behind redirects. Skip it if your click-to-page-view ratio is already healthy and mobile load times are known to be fine.

Do not mix populations in one number

Report outcomes separately by source, period, destination version, and offer version. Blend cold and warm traffic into one rate and that rate cannot tell you which one moved. Keep the combined total for operations, never as a test result. Skip this if there genuinely was only one source, one page version, and one offer in the period.

Attribute revenue to the enquiry that produced it

When buying takes time, link each sale back to the enquiry that started it, and report immature cohorts separately from matured ones. Here is the specific error to look for: if your sales cycle runs longer than your reporting window and you are comparing recent enquiries against same-period sales, you are judging cohorts that have not finished converting — a half-finished result read as a verdict. Skip this if purchase is effectively immediate; then same-period comparison is valid.

And before you declare anything: count the events, not the percentages

Before calling a variant dead or a winner proven, count the actual conversion events and ask what one more event would do to the number. Thin evidence is where an early percentage looks exceptional and a decision gets made anyway. The narrow exception: if you are only confirming that an event fires, that needs one successful test, not a sample.

Stage 3 — Offer and message: can a stranger say what they are buying

Write the offer as one sentence a stranger can parse — what is paid for, what is included, what is a bonus, what it costs, and what the button does. If you have to explain which part is paid, that ambiguity can leave the next step unclear. Skip it if an outsider with no context can already state correctly what is being purchased.

Then check continuity. Mirror the ad's core language and promise above the fold, so the visitor immediately recognises the page as the continuation of what they clicked. This is the check when click-through is healthy and page conversion is poor, when a new creative angle is running to an unchanged page, or when the page's audience qualifier is broader than the ad's. Skip it if the ad and the page already carry the same promise to the same person.

Stage 4 — Traffic: is the right person arriving

Here is the distinction a list of causes collapses. When engagement is strong and downstream conversion is weak, investigate who is arriving before rewriting anything. Healthy click rate with flat opt-ins, a conversion drop that followed an audience or geography change, cheap traffic bought because it was cheap — those are signals to go look at the arriving population. They tell you where to look, not what to conclude; the creative can still turn out to be the thing that needs work. Skip this if engagement and downstream conversion moved together — then the ad and the traffic are telling you the same story.

If what is arriving looks like the wrong person, testing which stage produced the unqualified leads is how to examine the traffic stage before you rewrite the ad.

Two more traffic-stage checks, each with its own boundary:

• Check exposure before comparing creative. Before treating one asset as beating another, confirm they actually received comparable delivery, duration, and audience — and that neither was stopped early or rejected. An early leader that received increasing allocation can look dominant on numbers that were never comparable. Skip if exposure was comparable and both ran to completion.
• Point optimisation at the deepest event with usable signal. Choose the optimisation event by trading outcome depth against signal density: the deepest real business event that receives enough volume to learn from. The situations to check for are optimisation pointed at a cheap shallow event while lead quality is poor, pointed at an event so sparse that delivery has little to work with, or pointed at an event that is not firing at all. Skip if the event is deep, firing, and receiving enough volume.

Stage 5 — Page and funnel: does the path hold together

Walk the page and the whole path on real devices, looking for outright breakage, and fix what you find before drawing any conclusion from the data. Weak conversion where nobody has actually used the page, a display or layout fault, a funnel that has not been walked end to end since the last change — start there. Skip it if the path has been walked on real devices since the last change and nothing is broken.

Then check the page's job. Define the traffic source, what the visitor just saw, and the single action this page must sell — then remove everything that does not serve it. This is the check when paid traffic is landing on a general company page, when one page serves several sources, or when the page explains the whole business to someone who needs to make one decision. Skip it if the page is already purpose-built for this source and this action.

If you want the numbers side of this stage, whether your landing page conversion rate is decision-ready goes deeper on the denominator problem.

Stage 6 — Email and lifecycle: a sequence is not a rescue

A reflex worth resisting: conversion is weak, so an email sequence gets inserted between the ad and the offer. Before adding that step, name the reason the buyer needs time — and check that the step is not simply adding another place to lose people. The situation to watch for is nurture being used as the default answer to any funnel problem. Skip this caution if the purchase genuinely requires consideration and the sequence serves that.

Stage 7 — Economics and scale: what can you afford to pay

Some campaigns are not broken. They are just being judged against a number nobody derived. Model the whole customer — price, margin, fulfilment cost, order bump, upsell, later offers, retention, refunds — and set your acquisition targets from that model rather than from a benchmark you read somewhere. This is the check when you do not know what a customer is worth, when acquisition cost is being judged against a figure from elsewhere, or when a front-end loss is being justified by backend revenue nobody has measured. Skip it if you have a current model and actual cohort data is tracking against it.

And make scale decisions on the right metric. Decide to scale, taper, or stop based on contribution and trend against your allowable acquisition cost — not on the cost of clicks or leads. Increasing spend on strong upstream metrics without downstream proof is exactly the move this check exists to prevent. Skip this if the campaign is inside its economics and no scaling decision is live.

One caution on the "works small, falls apart at scale" version of the problem, because it splits in two. If recorded conversion still holds and acquisition simply becomes uneconomic as spend rises, this stage is where the decision lives: contribution against allowable cost. If recorded conversion itself collapses, do not treat economics as the explanation — you have not established that yet. Go back to the earliest stage above you have not resolved, because an unresolved measurement, traffic, exposure, page, or fulfilment question can still account for the collapse.

Earn the right to optimise, then the right to scale

One discipline holds the whole sequence together: confirm the current stage has enough evidence before moving to the next. Setup verified, then a usable sample, then diagnosis, then controlled fixes, then a confirmed outcome, then economics, then scale. Detailed optimisation attempted on thin data, or spend increased on unproven conversion, is effort spent refining something upstream that was never verified. The flip side is real too — if the current stage's evidence requirement is met, move on. Over-verifying is its own kind of waste.

If none of this isolates the cause

These checks will not exhaust the possibilities, and finishing them proves nothing by elimination. Insufficient volume, execution problems, timing, and interactions between several factors at once can all survive every check above. If you work the sequence and no single stage fails clearly, the honest read is that the evidence you currently hold is not sufficient to diagnose it, and the next step is more first-party investigation — not another round of creative.

What the sequence does buy you is this: you stop guessing which of ten reasons is yours, and you stop drawing conclusions at stage five from data that failed at stage two.

If you would rather have a second pair of eyes work the sequence against your actual account, GrowthHackr's Campaign Leak Audit is a free diagnostic that looks for the earliest stage where your evidence stops holding up.