Your conversion tracking is probably lying to you.

A reliable tracking setup records a genuine enquiry once, captures calls as well as forms, and connects marketing activity to what the sales team actually received. Platform reports are for optimising campaigns. Your own customer records establish what happened.

When those two things get confused, advertising can look profitable for months while the business quietly struggles to find customers.

This is why analytics and tracking is one of our six seats, and why nothing launches until that seat has signed it off. Signing it off means following a real enquiry from the ad click to the record in the CRM. It does not mean confirming that a tag exists.

Why the platforms disagree with your own records

Every advertising platform measures results using its own attribution rules: the conditions under which it is willing to take credit.

Someone finds you through a Meta ad on Monday. On Thursday they search for your name, click a Google ad and submit the form. Depending on settings and available signals, both platforms may claim that enquiry. You received one lead. Adding the dashboards together produces two.

Neither tag is broken. The platforms are answering different questions about the same person.

The same gap opens up whenever two reports differ on any of these:

• What counts as a conversion — a button click, or a completed enquiry that reached somebody.
• The attribution window — how long after an interaction credit remains available.
• Which date the conversion lands on — the date of the ad click, or the date of the enquiry.
• How gaps are handled — cross-device journeys, missing signals and modelled conversions.

It runs the other way too, which is the half most articles leave out. Platforms undercount as well. Missing tags, browser restrictions and consent choices all create holes, and a platform that cannot see a conversion will not claim it.

Five setup faults that inflate the numbers

Attribution differences are expected and largely unavoidable. The faults below are neither, and we find at least one of them in most accounts we take over.

Duplicate tags counting the same action twice

A site ends up with the Google tag installed through the theme, again through a plugin, and again through Tag Manager. One submission fires three events. A related trap is counting the same enquiry through both a native Google Ads conversion and an imported GA4 conversion, with both set as bidding goals.

The fix. Map every conversion action to its trigger and its destination. Submit one test enquiry and count how many events each destination receives. Keep exactly one bidding signal per outcome. Where you run both browser and server-side tracking, confirm deduplication is actually working rather than assuming it.

Form attempts counted as successful enquiries

Clicking submit does not prove a form worked. A required field can be empty, the server can reject the request, and the tracking event can fire regardless. Nobody receives a lead, and the dashboard records one.

The fix. Fire the lead event only after a confirmed successful submission. Then test three times: an empty form, an invalid form, and a valid one. Only the third should count. GA4's automatic form interaction events are useful for understanding behaviour, but Google's own documentation distinguishes a generic form submission from the specific event you want to treat as a business outcome. Verify before promoting one to a goal.

Thank-you pages generating conversions without enquiries

A thank-you page is the easiest thing in the world to track, and it can be bookmarked, refreshed, revisited and opened directly. If every pageview counts as a lead, the number climbs on its own.

The fix. Tie the conversion to a confirmed submission, ideally carrying a unique reference so the same enquiry cannot be counted twice. Test direct access and a refresh. Hiding the page from search does not stop repeat counting.

Phone enquiries missing entirely

If customers routinely call before buying, form-only tracking leaves a hole you cannot reason around. And a tap on a telephone link is not a call: people tap, change their mind, or never connect.

The fix. Use call tracking that records actual connected calls and ties them to a source where that is possible. Separate answered enquiries from missed calls, spam and existing customers ringing about something else. Call duration helps as a filter, but a long call is not automatically a qualified lead. What happened next is the better evidence.

Test traffic becoming campaign performance

The developer tests the form. Your team tests it again. The agency checks it after editing the landing page. All three arrive as new leads, and all three go into the optimisation signal.

The fix. Test on a staging environment where you can. Where you cannot, label test submissions clearly and exclude them from both reporting and bidding. Internal traffic filters will not catch a remote colleague, an agency on a different network, or an automated check.

Why Meta attribution looks generous

Meta advertising genuinely influences people who never click straight through. Someone sees an ad, remembers the name, and searches a week later. Attribution that counts eligible views and engagements will recognise that journey, and produce a higher number than a click-only report.

Meta has been narrowing this. Its March 2026 measurement update redefined click-through attribution to cover link clicks only, and moved shares, saves and similar interactions into a separate engage-through metric. Check which settings your campaigns are actually running on before comparing anything.

Modelling is a different matter and worth separating out. Platforms model to fill measurement gaps, and that is not in itself evidence of inflation. Google models conversions too, for the same reason.

The distinction that actually matters is between an attributed conversion and an additional one. An attributed conversion met the platform's crediting rules. An additional conversion is one that would not have happened without the advertising. Standard attribution reporting cannot tell you the second thing, and no amount of dashboard comparison will make it.

Compare the attribution views available to you, check them against your real enquiry total, and run a controlled test when you need to know whether a channel is actually adding anything. Our guide to Meta ads and Google Ads covers how the two channels do different jobs in the same buying journey.

The one number nobody can inflate

You need one business-level measure that does not depend on anybody dividing up credit.

Define the outcome precisely. If you mean a paying customer rather than an enquiry, divide by new customers instead. The two are not interchangeable and the difference is usually where the money is.

For the number to stay honest:

• Count advertising spend consistently across every channel.
• Strip out spam, tests and duplicate records.
• Count one person's form submission and follow-up call as one enquiry.
• Keep new-business enquiries separate from support requests.
• Use consistent periods, and allow for the lag between enquiry and sale.

Two honest caveats. Poor record-keeping will distort this just as badly as platform attribution does. And it includes the effect of referrals, organic search and existing demand, so treat it as a measure of business-wide efficiency rather than proof that advertising caused every enquiry.

Track qualified enquiries and customers next to it. Cheap leads are expensive when none of them buy.

Ask people how they found you

Add one question to your enquiry form: where did you first hear about us? Offer a few clear options, an "other" box and a "can't remember". Keep it optional if making it mandatory costs you submissions.

This catches journeys no tracking can describe. A recommendation in a private message. A podcast. An ad somebody half-remembers from a month ago.

It is independent of every platform's attribution rules. It is not unbiased. People forget, they confuse ads with organic results, and they tend to name the last place they saw you rather than the first.

So use it as another line of evidence, stored alongside the tracked source rather than replacing it. If a steady stream of customers mentions Instagram while analytics shows mostly branded search, that tells you something real about how people discover you. It is not a reason to reassign every branded search to Instagram.

When the sale happens on the phone

For most service businesses the form submission is the start, not the result. What matters is the qualified conversation, the booked appointment, the accepted quote, the paid invoice.

Offline conversion tracking connects those later outcomes back to the advertising. Google's documentation covers how to report outcomes that happen away from the website, including sales closed over the phone.

A workable process:

1. Give every enquiry a unique record in the CRM.
2. Retain whatever attribution and matching information your setup needs.
3. Record consistent stages: qualified, appointment booked, won.
4. Send eligible outcomes back through the platform's supported integration.
5. Check upload errors, duplicates and match rates rather than assuming it worked.

Handle customer data with the permissions and platform requirements that apply to you.

Then choose bidding goals deliberately. In Google Ads, primary and secondary conversion settings determine which actions are used for bidding rather than only for reporting. Reporting on submissions, qualified leads and sales is useful. Treating all three as the same kind of success is how an account ends up optimising toward the cheapest, worst leads.

What to verify before you trust a dashboard

Run this before launch, and again after any change to forms, tags or integrations.

• A valid enquiry counts once in each destination, and only once.
• Failed submissions do not count, including validation errors and server rejections.
• Direct visits and refreshes of the thank-you page create nothing.
• The enquiry actually reaches your team, with usable contact details.
• Calls are measured separately from taps on a telephone link.
• Tests, spam and duplicates are excluded from genuine lead totals.
• Bidding goals reflect the outcome you actually want.
• Attribution settings, date basis and time zones are written down somewhere.
• Offline outcomes upload cleanly, and failures get investigated.
• The gap between CRM totals and dashboard totals can be explained.

Use tag debugging and the platforms' test event tools, then reconcile against the actual enquiry record. A tag firing is the first half of the test, not the whole of it.

This is the standard behind our Google Ads management and Meta ads management: agree what counts as a result, verify it is recorded correctly, then let that evidence decide the spending.

Before you increase a budget, follow one enquiry the whole way through. If your records cannot account for it, fix the measurement before you spend more against it.