Customer storyA large enterpriseNovember 16, 2026

# How a Large Enterprise Found a Mobile-Only Conversion Bug

A lead form was sending far more conversion events than GA4 was showing. Manual tests never reproduced the problem. The clue was already in the events: one property, with the same value on every duplicate.

**A large enterprise** Plan: Enterprise

5 min read · 1171 words

## A conversion event that didn't add up

One of this company's landing pages asks visitors a simple question: do you want us to call you? The visitor leaves a phone number in a form, and the site sends a callback event. That event counts as a conversion. It feeds the reports that say which campaigns bring leads, and the numbers budgets are set on.

On that landing, two tools disagreed. Trackingplan was recording far more callback events than GA4 was showing for the same page.

Two analytics tools never match exactly, and small gaps are normal. This one wasn't small. It was too big to leave alone.

A gap like that can be read two ways. Either one tool is missing events, or the site is sending events it shouldn't. Before deciding which number to trust, the team had to know which of the two it was.

## Duplicates, not leads

The answer came from looking at the events one by one instead of as a total.

The extra events were not new requests. They were copies. The same callback event arrived twice, or three times, in a single session. In the worst session it arrived up to 750 times, every copy identical, in less than a minute.

Nobody fills in a form 750 times in a minute. Something in the page was sending the same event over and over.

That changed the question. It was no longer which tool was right. It was what on the page was firing a conversion event without a new conversion behind it.

## Tests that found nothing

The team did what any analytics team does first. They opened the landing and tried to reproduce the problem by hand.

The form worked. Every manual test ended the same way: no duplicates.

That is the hardest kind of tracking bug, the one that never happens in front of you. The data says something is wrong, and every test you run says it isn't. From there, a team has two bad options. It can keep testing at random and hope to hit the right combination, or it can decide the data is wrong and move on.

## Why manual tests miss bugs like this

A manual test follows the path the tester has in mind. Open the page, fill in the form, submit, check that the event fired. That is the path the page was built for, so it is usually the one that works.

Real visitors take other paths. They type a number wrong, close a window halfway, come back, try again, and do all of it on whatever device they have in their hand. Each of those is a different path through the page, and each one can send events that no test plan checks.

On a single-page app there are more of those paths, and they last longer. The page doesn't reload between steps, so what a component did a moment ago can still be running when the next one opens. A listener that should run once can run again.

Multiply that by phones and desktops, and the number of possible paths is larger than any test plan can cover.

## The clue was in the events

The answer was already in the events the site had sent. It just wasn't where anyone had looked.

Trackingplan's agent pointed to one property on the duplicated events: `error_form`. Every duplicate carried the same value, `invalid phone number`.

The manual investigation had walked past that property. On its own, it didn't look broken. What mattered was the pattern: the same value on every duplicate.

That gave the investigation a direction. The duplicates did not come from just any session that used the form. They came from sessions where the form had flagged the phone number as invalid. Instead of testing at random, the team now knew which state to look for.

## A loop that only lived on mobile

With that clue, the team found the affected sessions and looked at what those visitors had done.

The landing is a single-page app. On mobile, closing a modal triggered a bug in the app that sent the callback event into a loop. That loop is where the copies came from, up to 750 of them in one session.

It never happened on desktop.

That is why the manual tests found nothing. To see the bug, a tester had to be on a phone and close that modal. A test that doesn't go down that exact path never gets there.

## What it means for the numbers

Callback requests are leads. When the event that counts them misbehaves, every number built on it does too: leads per campaign, conversion rate per landing, cost per lead.

The company didn't quantify how many leads were affected, and this story doesn't guess. What the case does show is how a bug like this can sit in a conversion event while every usual check passes. The form works, the tests pass, and the only sign is a gap between two tools that is easy to explain away.

## What the case shows

The bug was in a conversion event, so it went straight to the numbers people make decisions on.

The usual checks didn't see it. What broke was one path, a modal closed on a phone. Paths like that are easy to miss in a test plan. They show up in real traffic.

And the clue that solved it was not new data. It was a property value the site was already sending, read across every duplicate at once.

## Where Trackingplan fits

Trackingplan sees the events your site actually sends, from real visitors, on every device. Its monitoring compares them against what normally happens, per event, per property, per platform, and flags implementation issues like missing properties and duplicate fires as they show up on real traffic.

When something looks wrong, you can ask the agent about it. It investigates before it answers, works from your own hits, and shows the evidence behind its conclusions. Here, that meant pointing to the one property that explained the duplicates. It turned a bug nobody could reproduce into one the team could find.

## What to check on your own site

If a conversion event looks higher in one tool than in another, run a few checks before you decide which tool is wrong.

Look for identical events in the same session, close together in time. A person doesn't convert twice in a few seconds.

Compare the property values on the extra events with the normal ones. A value that every extra event shares, like an error state, tells you where to look.

Test on a phone, not only on a desktop browser, and test the paths around the form: errors, modals, closing and coming back.

On a single-page app, check what happens to the event when a component closes and opens again without a page reload.

This customer story has been extracted from one or more real customer interactions. Names and other related data have been anonymized for privacy reasons.

Company

A large enterprise

Plan

Enterprise

On this page

1.  [*01*A conversion event that didn't add up](#a-conversion-event-that-didn-t-add-up)
2.  [*02*Duplicates, not leads](#duplicates-not-leads)
3.  [*03*Tests that found nothing](#tests-that-found-nothing)
4.  [*04*Why manual tests miss bugs like this](#why-manual-tests-miss-bugs-like-this)
5.  [*05*The clue was in the events](#the-clue-was-in-the-events)
6.  [*06*A loop that only lived on mobile](#a-loop-that-only-lived-on-mobile)
7.  [*07*What it means for the numbers](#what-it-means-for-the-numbers)
8.  [*08*What the case shows](#what-the-case-shows)
9.  [*09*Where Trackingplan fits](#where-trackingplan-fits)
10.  [*10*What to check on your own site](#what-to-check-on-your-own-site)

**Stop babysitting tags**

Trackingplan audits your real user data around the clock and flags issues before they cost you.

[Start free trial →](https://app.trackingplan.com/signup)

A large enterprise

**A large enterprise**

Share Copy link [LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.trackingplan.com%2Fcustomer-stories%2Fhow-a-large-enterprise-found-a-mobile-only-conversion-bug) [X](https://twitter.com/intent/tweet?url=https%3A%2F%2Fwww.trackingplan.com%2Fcustomer-stories%2Fhow-a-large-enterprise-found-a-mobile-only-conversion-bug&text=How%20a%20Large%20Enterprise%20Found%20a%20Mobile-Only%20Conversion%20Bug)

More stories

[All Customer stories →](/customer-stories)

-   [A global enterpriseEnterpriseJuly 13, 2026
    
    ### How a Global Enterprise Discovered That Paid Traffic Was Invisible to Its Analytics
    
    “You can have a perfectly implemented tag, in the right place, firing the right event — and still lose most of your data. If the timing is wrong relative to how real users behave on your specific traffic mix, it is operationally invisible until something makes you look at it differently.”
    
    ](/customer-stories/how-a-global-enterprise-discovered-that-paid-traffic-was-invisible-to-its-analytics)
-   [![CSA (Havas Media Network)](/images/695e4088ac2a68690388b766_havas-dark.svg) Advertising ServicesMay 11, 2026
    
    ### How CSA (Havas Media Network) Regained Client Trust with Out-of-the-Box Data Quality Monitoring
    
    “We came across exactly the solution we were looking for – and this is rare; it’s not every day you find something so perfectly suited to your problem.”
    
    **David Rascón** · Data & MarTech Consultant at CSA (Havas Media Network)](/customer-stories/how-csa-havas-media-network-regained-client-trust-with-out-of-the-box-data-quality-monitoring)
-   [![On Top Media](/images/67bc9fd4eaaeeffbf46df158_ontop-logo.avif) Advertising ServicesApril 4, 2026
    
    ### How On Top Media Prevented Tracking Breakdowns with Trackingplan
    
    “Before Trackingplan, we spent around five to six hours per project every month just debugging and ensuring data accuracy.”
    
    **Álvaro Salmerón** · Digital Analytics Consultant at On Top Media](/customer-stories/how-on-top-media-prevented-tracking-breakdowns-with-trackingplan-automated-validation)

Trackingplan

## See everything. Miss nothing.

Your implementations audited around the clock with real-time, real-user data. Real-time alerts about errors or changes in your data, campaigns, pixels, privacy and consent. Let AI flag issues before they cost you.

[Start free trial →](https://app.trackingplan.com/signup) [Book a demo](/demo)
