Tracking November 20, 2026 · 7 min read

Run a Conversion Tracking Audit Before You Scale Ad Spend

Bidding algorithms optimise toward whatever you tell them is a conversion. If that signal is wrong, more budget only buys more of the wrong thing, faster.

Run a Conversion Tracking Audit Before You Scale Ad Spend
Photo: Unsplash

Key takeaways

  • A conversion tracking audit checks that every event the ad platforms optimise on is real, counted once, and matches your backend before you add budget.
  • Meta deduplicates browser pixel and Conversions API events only when both send the same event_name and event_id, so a missing or mismatched ID inflates conversions.
  • GA4 and ad platforms will never match exactly because they use different attribution models and dates, so reconcile both against backend orders, not against each other.
  • A good audit ends with a ranked fix list ordered by impact on bidding, not a long report of every warning in the account.

Every bidding system we work with, Meta Advantage+, Google Smart Bidding, TikTok's optimisation, learns from one thing: the conversions you send back. When those conversions are duplicated, mislabelled or missing, the algorithm does not know. It simply optimises toward the wrong people with more confidence. That is why we treat a conversion tracking audit as step zero before any account is scaled.

This guide covers why the audit matters, what we check, and what a useful output looks like.

Why audit tracking before scaling ads?

At low spend, bad tracking is an annoyance. At scale, it is expensive. Three things happen when data is wrong:

  • Inflated conversions make CPA look healthy. A purchase counted twice halves your reported CPA. You scale a campaign that was never profitable.
  • The wrong event becomes the target. If an "add payment info" or "begin checkout" step fires a Purchase event, the algorithm learns to find people who reach checkout, not people who pay.
  • Missing signal starves the model. If iOS users, consented-out users or offline sales never get reported, the algorithm under-values the audiences that actually convert and shifts budget away from them.

A tracking audit before scaling ads catches these before the extra budget amplifies them. It is cheaper to fix one event than to spend a quarter re-training campaigns that learned on bad data.

The conversion tracking audit checklist

We run the same sequence on every account. Adapt it to your stack, but do not skip steps.

“Scaling on broken tracking just teaches the algorithm the wrong lesson faster.”
FOXA Digital

1. Event inventory

List every conversion event across every platform: Meta, Google Ads, GA4, TikTok, any MMP. For each one, record:

  • Where it fires (browser, server, app SDK, CRM import)
  • The exact trigger (URL, button click, dataLayer push, backend webhook)
  • Whether it is used for bidding (primary) or reporting only (secondary)
  • The value and currency it sends

Most accounts we open have at least one event nobody can explain. Then trigger each one yourself, using GTM preview, Meta's Test Events tool and GA4 DebugView, and confirm it fires once, at the right moment.

2. Pixel and Conversions API deduplication

Meta CAPI deduplication is the most common failure we see. When you send the same purchase from the browser pixel and from the Conversions API, Meta only merges them if both carry the same event_name and the same event_id. If the server sends an order ID and the browser sends a random string, or one side sends "Purchase" and the other "purchase", you get two conversions.

Check in Events Manager that deduplication is being applied, and compare browser versus server counts. TikTok's Events API works on the same principle, so check it the same way. While you are there, review Event Match Quality: sending hashed email, phone and click IDs server-side improves how many conversions Meta can match to people.

3. GA4 vs ad platform vs backend reconciliation

The GA4 vs Meta discrepancy question comes up in every audit. They will not match, by design:

  • GA4 uses its own attribution model across all channels; Meta only credits conversions it can tie to its own ads.
  • Ad platforms usually report a conversion on the date of the click or impression; GA4 reports it on the day it happened.
  • Meta includes view-through and some modelled conversions; GA4 sees only tracked sessions.

So do not reconcile GA4 against Meta. Reconcile both against your backend: orders, paid bookings, qualified leads in the CRM. Pull a week of data, compare totals and revenue values, and track the ratio over time. A stable gap is fine. A gap that jumps after a site release is a broken tag.

4. Attribution windows

Note the attribution setting on every campaign and conversion action. Meta's long-standing default for most objectives is 7-day click and 1-day view; Google Ads lets you set a window per conversion action. Mixed windows across campaigns make performance comparisons meaningless, and a long view-through window on a retargeting campaign can make it look far better than it is. Pick windows that match your real purchase cycle and keep them consistent.

5. Consent mode and privacy

If you serve users in regions that require consent, Google Consent Mode v2 must pass the correct ad_storage, analytics_storage, ad_user_data and ad_personalization states, and the banner must actually block or release tags. Test both paths: accept and decline. In Indonesia, the Personal Data Protection Law (UU PDP) is now in force, so even local brands should know what they collect and on what basis. Misconfigured consent can quietly remove a large share of conversions from Google's data.

6. Offline and CRM conversions

For lead generation, a form submit is not a sale. Feed qualified leads and closed deals back to the platforms: Google Ads offline conversion import or enhanced conversions for leads, and Meta's Conversions API for CRM events. Check that click identifiers (GCLID, fbclid via fbc) are captured at form submit, stored in the CRM, and returned with the outcome. If they are lost at any step, the platform optimises for form fills, including the junk ones.

7. App MMP and SKAN, where relevant

For apps, the MMP (AppsFlyer, Adjust and similar) is the source of truth, and iOS adds SKAdNetwork. Check that SKAN conversion values map to the events that matter for your business, that postbacks reach both the MMP and the ad networks, and that in-app events from the SDK match backend counts. We cover this in depth in how we lowered iOS app CAC with SKAN and AppsFlyer.

What breaks bidding algorithms when data is wrong?

The failure patterns repeat across accounts:

  1. Double counting from failed dedup or a tag firing on page reload. Reported CPA drops, real CPA does not.
  2. Wrong event as primary. A micro-conversion set as primary in Google Ads, or a mislabelled Purchase in Meta, pulls bidding toward low-intent users.
  3. Missing or zero values. Value-based bidding (tROAS, highest value) cannot work if half the purchases send a value of 0 or the wrong currency.
  4. Signal loss on iOS and consent. The model sees fewer conversions than really happen, learns slower and keeps re-entering learning.
  5. Sudden changes. A redesigned checkout that renames an event resets what the algorithm thinks success looks like.

The fix for each is in the data layer, not in the campaign settings. That is why we audit tracking before touching bids.

What does a good audit output look like?

Not a 60-page PDF. A good conversion tracking audit delivers a ranked fix list: every issue scored by its impact on bidding and the effort to fix it, ordered so the first items change performance soonest.

Each line should state:

  • The issue, in one sentence
  • The evidence (screenshot, event count, reconciliation gap)
  • The impact on bidding or reporting
  • The owner (marketing, developer, CRM team) and the exact fix
  • How to verify it is fixed

Example from our own work: for Bali Bird Park, we found and fixed one mis-firing checkout event, so purchase data could be trusted again. The same engagement also delivered +56% organic website traffic.

How we run it at FOXA

Our audit takes 7 to 10 working days and follows the checklist above. We test every event live, reconcile platform data against your backend, and deliver the ranked fix list with developer-ready tickets. When you continue with a paid media retainer, the audit fee is credited to month one.

For HappyFresh, we rebuilt iOS measurement (SKAN in AppsFlyer) and moved bidding onto MMP data. New-user CAC fell 69% in 3 months.

Clean tracking also feeds more than ads. Accurate events and structured data help your site get read correctly by search engines and AI assistants, which we cover in how to get your brand cited in AI search. See our tracking service for the full scope.

Before you raise the budget

Run the checklist on your top three conversion events this week, starting with dedup and backend reconciliation. If the numbers do not line up, book a call before the budget goes up rather than after.

Questions we get asked

What is a conversion tracking audit?

A conversion tracking audit is a structured check of every event your ad platforms and analytics record, from pixel and SDK to server-side and CRM. It confirms each conversion fires at the right moment, is counted once, carries the right value, and reconciles with your backend data. The output is a prioritised list of fixes so bidding learns from accurate signals.

Why does GA4 show fewer conversions than Meta?

GA4 usually credits conversions using its own attribution model and only sees sessions it can track, while Meta counts conversions it can attribute to an ad click or view within its attribution window, including some it models. Consent, ad blockers and cross-device journeys also hit GA4 harder. A gap is normal; a sudden change in the gap is the warning sign.

How do I check if Meta CAPI deduplication is working?

Open Events Manager, pick the event and check whether it shows both browser and server sources with deduplication applied. Confirm the pixel and the Conversions API send the same event_name and the same event_id for each action. If server event counts roughly double browser counts with no dedup reported, IDs are missing or mismatched.

How long does a tracking audit take?

For a typical web or app account, a focused audit takes about 7 to 10 working days: a few days for inventory and live testing, a few for reconciliation against backend data, and the rest for writing the ranked fix list. Accounts with several apps, markets or a custom CRM pipeline can take longer.

About FOXA DigitalFOXA Digital is a performance marketing agency in Jakarta, Indonesia. Since 2018 we have run paid media, tracking, SEO and content for apps, consumer brands, hospitality and attractions, judged on CAC, ROAS and leads.

Want us to look at your numbers?

30 minutes. No deck. You leave with three things to fix this week.

Book a strategy call