Editorial Team You can’t take ad platform data at face value. When the numbers from your ad dashboard don’t line up with your own analytics, it creates a fog of mistrust that makes strategic planning almost impossible. Our latest campaign, “Project Horizon,” was a sharp lesson in why a proper data audit isn’t some side-quest for the tech team. It’s absolutely fundamental to getting to anything resembling true organic accuracy. So how can you, as a marketer, get in there and systematically check what the platforms are feeding you against your own ground-truth data?
Key Takeaways
- Check your ad platform spend and conversions against your internal analytics every single day to spot problems before they get out of hand.
- Make sure the impression and click attribution models are consistent between your ad platforms and your analytics setup, otherwise you’re comparing apples and oranges.
- Do a deep-dive audit every quarter where you compare the platform’s cost per conversion to your own calculated organic cost per acquisition to find where data might be leaking or misattributed.
- Write down clear rules for how your team uses tags and UTM parameters because messy data hygiene will absolutely destroy your reporting accuracy across campaigns.
- You have to understand the weird reporting quirks of each platform, like how they count view-through conversions, if you want to have any hope of interpreting performance correctly.
Project Horizon: A Case Study in Data Discrepancy
Back in Q1 2026, we launched “Project Horizon.” It was a full-scale digital campaign for a new SaaS product where we needed to drive sign-ups at a competitive cost per lead (CPL) and also give our organic search visibility a shot in the arm for some key features. We put a $150,000 budget behind it for a six-week duration, with the bulk of the money going to Google Ads and Meta Ads.
The plan had a few moving parts. We ran Google Search campaigns aimed at high-intent keywords, used the Google Display Network (GDN) for general brand awareness and retargeting, and leveraged Meta Ads to hit a broad audience with interest-based targeting and lookalikes. We had a mix of creative, including short video testimonials and static images that spelled out the product’s benefits, plus some interactive carousel ads. All of it pushed clear calls to action like “Start Your Free Trial” and “Get a Demo.”
Initial Performance Metrics (Platform-Reported)
Three weeks in, the platform dashboards looked great. Google Ads was showing a healthy Click-Through Rate (CTR) of 4.8% on search and 0.7% on GDN. Meta Ads was clocking an average CTR of 1.2%. We’d served over 15 million impressions across the board. The platform-reported Cost Per Lead (CPL) was the best part: just $28.50 on Google and $32.10 on Meta. Based on our early lead-to-customer math, we were looking at a Return on Ad Spend (ROAS) of 1.8x.
Table 1: Platform-Reported Performance (First 3 Weeks)
| Metric | Google Ads | Meta Ads | Combined |
|---|---|---|---|
| Total Spend | $55,000 | $40,000 | $95,000 |
| Impressions | 9,200,000 | 6,000,000 | 15,200,000 |
| Clicks | 225,000 | 72,000 | 297,000 |
| Conversions (Leads) | 1,930 | 1,245 | 3,175 |
| CPL | $28.50 | $32.10 | $29.92 |
The Inevitable Data Audit: Uncovering Discrepancies
Even with those strong platform numbers, our internal source of truth, Google Analytics 4 (GA4), was starting to raise red flags. We saw a small but consistent gap between the conversions the platforms were claiming and what GA4 was attributing to our paid channels. A little delta is normal, but this one was growing, so we had to go digging. Our audit hit three main areas: attribution models, tagging consistency, and conversion event definitions.
Attribution Model Mismatch
This was the first and most obvious culprit. Google Ads defaults to a “last click” model for many reports but also gives itself credit for “view-through conversions” (VTCs) on the Display Network. Meta Ads uses a different logic, typically a 7-day click and 1-day view window. We had configured our GA4 for a data-driven attribution model that spreads credit across multiple touchpoints along the customer’s path, creating a fundamental disconnect. For example, a user sees a GDN ad, doesn’t click, searches organically a day later, and then converts. In that scenario, Google Ads might claim a VTC while GA4 would rightly give credit to both organic search and the ad impression.
Our initial pass showed that about 15% of the conversions Google Ads was reporting were VTCs that GA4 attributed very differently. At the same time, Meta Ads was claiming 10% more conversions than GA4 which made sense given its generous 1-day view window that scoops up credit for users who just saw an ad before converting through another channel.
Tagging Consistency and UTM Parameters
Inconsistent UTMs are a self-inflicted wound. We thought we had a standard naming convention, but the audit found spots where manual ad creation had introduced tiny, report-breaking variations. Someone used utm_source=googleads in one ad set and utm_source=google_ads in another. That tiny difference was enough to fragment our reporting in GA4, which made it impossible to get a clean performance rollup. Because of these sloppy UTMs, about 8% of our paid traffic was getting misfiled in GA4, usually making our direct or organic traffic look better than it was.
Even worse, we found a couple of landing pages where a recent website update had broken the GA4 tracking code. It just wasn’t firing. This meant we were completely blind to a slice of our paid traffic, which threw off both the platform and our internal numbers. That one issue cost us an estimated 50 lost conversions in the first three weeks and made our CPL look higher than it was.
Conversion Event Definition Alignment
We also realized our “sign-up” event wasn’t defined the same way everywhere. Inside Google Ads and Meta Ads, a conversion fired right after a successful form submission. Simple. But in GA4, our main “sign_up” event was tied to the user seeing a unique confirmation page, which in some cases only happened after an extra email verification step. This delay created a small but real discrepancy, with about 3% of platform-reported conversions not showing up right away in GA4 because of that extra step.
What Worked, What Didn’t, and Optimization Steps
The creative itself was solid. The video testimonials on Meta Ads did surprisingly well, with a video completion rate of 75% for the first 15 seconds. And our Google Search campaigns aimed at specific product feature keywords brought in high-quality leads that had a much lower bounce rate than our broader campaigns. The GDN retargeting also converted well, which told us our audience segmentation was on the right track.
What bombed was our broad targeting on GDN. It got us a ton of impressions, but the CPL was too high and the leads were junk. The inconsistent tagging we found was a major operational failure that directly hurt our data’s reliability. And while the different attribution models are a known issue, we should have had a plan to reconcile them from day one instead of waiting for the numbers to look weird.
Optimization Steps Taken:
- Standardized UTM Protocol: We immediately documented and enforced a strict UTM naming convention for every new campaign and went back to fix the old ones. Every single ad now gets a URL built with the official campaign URL builder. No exceptions.
- GA4 Tagging Audit & Fix: The web dev team jumped on the broken GA4 tracking code and fixed the pages. We also started a weekly tag health check using Google Tag Assistant to prevent this from happening again.
- Unified Conversion Definitions: We re-engineered our conversion events to be as close to identical as possible across Google Ads, Meta Ads, and GA4. This meant changing GA4 to track the initial form submission to match the ad platforms, and then creating a second, separate event for the email verification part.
- Adjusted Attribution Reporting: For all our internal reports, we now use GA4’s data-driven attribution model as the single source of truth for paid channels. When we have to show platform-specific data, we explicitly state the platform’s attribution model and then show the reconciled GA4 view right next to it for context.
- Refined GDN Targeting: We killed the underperforming broad GDN campaigns. The budget was moved to more specific custom intent audiences and affinity segments that we built from what we learned in the audit.
- Daily Data Reconciliation: We have a new process now: every morning, we reconcile platform spend and conversions against GA4. If the variance is over 5%, it triggers an immediate investigation.
Post-Optimization Performance & Recalibrated Metrics
After we made these fixes, the data started to make sense. We ran the campaign for another three weeks, and the final audit gave us a much more honest picture of what happened. That initial ROAS of 1.8x turned out to be a more realistic 1.65x once we looked at it through the lens of GA4’s data-driven attribution. Our true, all-in CPL, after finding the lost conversions and cleaning up the data, was $31.20, a little higher than what the platforms first told us, but it’s a number we can actually trust.
Table 2: Reconciled Performance Metrics (Full 6 Weeks)
| Metric | Google Ads (GA4 Attributed) | Meta Ads (GA4 Attributed) | Combined (GA4 Attributed) |
|---|---|---|---|
| Total Spend | $88,000 | $62,000 | $150,000 |
| Conversions (Leads) | 2,700 | 1,900 | 4,600 |
| CPL | $32.59 | $32.63 | $32.61 |
| Cost Per Conversion | $32.59 | $32.63 | $32.61 |
| Total Conversions (Platform Reported) | 3,200 | 2,300 | 5,500 |
