To check your conversion tracking is working, make a conversion yourself and follow it. Watch it arrive live in Google Analytics 4 and in Meta, confirm it was counted once and only at the moment of success, then repeat the test on a phone and again with cookies declined. Finally, compare a week of reported conversions with the enquiries and sales you actually received. The live tests fit into an hour or so; the comparison needs a week of ordinary trading.
This post covers the test, not the setup. It assumes the tags are installed and asks whether to believe them, which matters most before you raise an ad budget or after a website change.
What a broken conversion costs you
Google Ads and Meta both choose who sees your ads partly from the conversions you send back. If the conversion fires whenever someone loads a page, automated bidding goes looking for more people who load pages. If it fires twice, every campaign looks twice as good as it is, and budget drifts towards whichever one double counts most.
The opposite failure is quieter. A form plugin update or a new cookie banner can stop conversions recording altogether. The campaigns then appear to produce nothing, and the natural reaction is to switch off ads that were bringing in work.
What to gather first
- The actions that matter. Usually an enquiry form, a booking and a phone call, or a completed order for a shop. Beside each, note what the customer sees when it succeeds and its name in each platform: the GA4 key event, the Google Ads conversion action and the Meta event (Lead, Purchase or similar).
- Access. GA4, Google Ads, Meta Events Manager and, if the site uses it, Google Tag Manager.
- Your own record of what came in. The form’s inbox, your CRM, booking system or order list. It is the only figure made of real customers, so every platform is measured against it.
- Two free browser tools. Google’s Tag Assistant (at tagassistant.google.com) and Meta’s Pixel Helper extension for Chrome. Both show which tags fired on a page without changing the site.
Warn whoever handles enquiries that a test is coming, and put “TEST” in the name field so nobody rings a customer who does not exist.
Step 1: Open the live views before you test
Open Tag Assistant, enter your website address and connect. Your site opens in a new tab with debugging switched on, and Tag Assistant lists each tag as it fires. On a site using Google Tag Manager, start Preview from your workspace instead, which also shows the trigger behind each tag.
In another tab, open GA4, go to Admin and find DebugView under Data display. Debug sessions appear there within seconds, events in order. Keep Pixel Helper active in the debugging tab so Meta shows up at the same moment.
If your GA4 property uses an internal traffic filter to exclude your office’s IP address, your test is excluded on purpose and will look like a fault. Test on mobile data or another connection if that applies.
Step 2: Convert the way a customer would
In the debugging tab, go to the page your ads send people to and fill in the form with realistic details, keeping “TEST” in the name field. Do not jump straight to the thank-you page; follow the customer’s route. Note the time and the email address you used.
Then read what the tools show. The conversion event should appear once, at the moment the form succeeded, under the name on your list. If you pass a value, it should be the right amount in GBP. In Pixel Helper, the matching Meta event should appear on the same step.
For an online shop, buy something inexpensive, or use your platform’s test payment mode if it still ends on the normal order confirmation page. The purchase event should carry the order number and basket total. Unless your setup sends refunds to the platforms, a refund will not remove the test from their reports, so note the order number.
Step 3: Prove nothing fires without a real conversion
A conversion that fires when nobody converted is worse than one that never fires, because it looks healthy. Stay in the debugging tab and try three things:
- Type the thank-you page address straight into the browser. If the conversion fires, anyone who refreshes that page, returns to it from their history or finds it in Google counts as a lead.
- Submit the form with a required field left blank. The form should refuse it and no conversion should fire. A trigger based on clicks of the submit button counts this as a lead.
- Browse a few pages and leave. The conversion should not appear.
Look out, too, for double firing: the same event listed twice for one action. On a shop, refresh the order confirmation page and see whether a second purchase event appears. Google Ads can discard a repeat of an order it has already counted, but only when the tag passes the order number as the transaction ID.
Step 4: Check Google Ads and Meta received it
The browser tools prove a tag fired, not that the platform recorded anything useful.
Google Ads
Google Ads only counts a conversion that follows an interaction with one of your ads, so a test made by typing in your address will not appear in its conversion columns. That is expected. Instead, confirm in Tag Assistant that the Google Ads conversion tag fired with the conversion ID and label of the matching action. Then open Goals, Conversions, Summary, and look into any action whose status is not a healthy, recording one.
Check which actions are set as a primary conversion, because only those feed bidding and the main Conversions column. A setup I come across often imports the GA4 key event and also installs the Google Ads tag for the same enquiry, both as primary, so every enquiry counts twice. Make one of them secondary.
Meta
Events Manager records events from every visitor who allowed tracking, not only those from ads, so your test should appear in the dataset’s overview. It is not instant; give it a while.
If you also send events through the Conversions API, each enquiry reaches Meta twice, from the browser and from your server. Meta merges the pair only when both carry the same event name and event ID, and Events Manager reports deduplication for each event that arrives both ways. Last, confirm the ad set optimises for that same event, not for landing page views or link clicks.
Step 5: Repeat the test on a phone
The phone version of a site is not always the same as the desktop one. Menus collapse, forms move into pop-ups, and the main call to action is often a tap-to-call button the desktop layout never shows, so tracking can break on mobile alone.
Run the enquiry again on a phone and check it reaches DebugView and Events Manager. Test on an iPhone in Safari, which shortens the life of some tracking cookies. Then tap a link inside the Facebook or Instagram app: it opens your site in the app’s built-in browser, which is where a visit from an ad in those apps begins, and some forms and payment pages behave differently there. If you record taps on your phone number, tap it and check the event fires.
Step 6: Decline cookies, then test again
UK sites need consent before setting analytics or advertising cookies. Under PECR those cookies are not strictly necessary, and the ICO expects rejecting to be as easy as accepting, with nothing non-essential loading before the visitor decides. So test both answers.
Open a private window, load the site, choose Reject all and submit the form. Then check:
- No Google Analytics or Meta cookies were set. In Chrome, open the developer tools, then Application, then Cookies, and look for names beginning _ga or _fbp.
- Pixel Helper shows no Meta Pixel activity.
- If the site uses Google’s Consent Mode in its advanced form, Google’s tags may still send signals without cookies. Tag Assistant shows the consent state on each event, and analytics and ad storage should both read as denied.
Your test will not appear as a normal visit in GA4 or as a lead in Meta. That is the correct result, not a fault. People who decline are missing from platform figures, or estimated by Google’s modelling rather than counted, so the platforms will always report fewer conversions than you received. Firing tags before consent to close the gap swaps a measurement problem for a compliance one.
Now open a fresh private window, accept cookies and test again. Everything from Step 2 should fire. Watch for tags that only start on the page after consent: the first page of a visit from an ad carries the click identifier in its address, and if the tags miss it, the enquiry may be recorded but not credited to the ad.
Step 7: Compare a week of numbers with your own records
The live tests show the tracking can work; a week of real figures shows whether it does. For each day, write down four numbers: real enquiries or orders from your own records, GA4 key events, Google Ads conversions and Meta results.
They will not match, and they should not: visitors who declined cookies are missing, and each platform credits conversions by its own rules and time windows. The reasons are covered in why GA4 and Facebook conversion numbers do not match. Here you are reading the pattern.
| What you see | What it usually means |
|---|---|
| A platform reports more conversions than you received | Counting twice, firing on a page view, or counting spam submissions |
| A platform reports none while enquiries keep arriving | A missing or broken tag, or a banner that never releases the tags after consent |
| The figures dropped sharply from one date | Something changed that day: a site update, a new form or a new banner |
| Each platform sits a fairly steady amount below your own total | Normal: declined consent, ad blockers and differing attribution rules |
Check for spam too. Every bot submission counts as a conversion in every platform, so if the platforms match your inbox but much of it is junk, the tracking is fine and the form needs better spam protection.
Faults this test turns up most often
The thank-you page counts anyone who reaches it
Fire the conversion from the form’s own success event where you can. If you keep the page trigger, add a noindex tag and keep the page out of your menus.
The same tag installed twice
Often a pixel added by a plugin and again by hand, or GA4 in the site’s code and in Tag Manager. Remove one and re-run Step 2.
A form that changed underneath the tracking
Many forms submit without loading a new page, so a trigger waiting for one never fires. Trigger on the success message, or on an event the form plugin sends to the data layer.
A cookie banner that does not control the tags
Some banners record the choice and change nothing; others block tags even after acceptance. Each tag should wait for consent in its category, then fire.
The wrong action set as the goal
Newsletter sign-ups or contact page visits set as primary alongside real enquiries teach bidding to chase the cheapest. Make them secondary.
Keeping it working
Tracking breaks when something around it changes. Run the test again after a website update, a new form, plugin or cookie banner, and before you raise an ad budget. Add an annotation in GA4 on the days the site changed so a sudden drop has an obvious explanation.
If you would rather not fix what the test finds yourself, putting the tracking right is the first thing I do when I take on Google Ads management or Facebook and Instagram ads, before any budget changes. I follow every conversion from the form to the platform and back to your records, so you make decisions on figures you can trust.
