Last week we covered how to audit your UTM tracking -- the campaign tagging layer that determines whether traffic arrives in GA4 with proper source, medium, and campaign attribution. If you haven't read that yet, start there. Clean UTMs are the foundation.
But here's the thing: even perfect UTMs can't save you if GA4 itself isn't configured to share data with Google Ads correctly.
There are 6 settings in GA4 Admin that directly determine whether Google Ads receives accurate conversion data, proper cross-device attribution, and usable audience signals. Most paid media managers never check them. And when we audit properties for our agency partners and enterprise clients, at least 2 or 3 of these are misconfigured on almost every property we see.
This post walks through each setting: where to find it, what it does, why it matters specifically for paid media, and what "wrong" looks like. Then we'll go a layer deeper into something we've been seeing a lot in our audits — key events that look healthy on the surface but are actually lying to Google Ads about your conversion volume.
Same 90-day lens as Part 1. Same goal: trust the numbers before you scale spend.
---
The 6 GA4 Admin Settings to Update for Google Ads
1. Reporting Identity
Where to find it:
GA4 Admin → Property Settings → Reporting Identity
What it does:
This setting controls how GA4 stitches users across devices and sessions.
There are three options:
**Blended** (recommended) — Uses Google Signals, device ID, and modeled data to stitch users across devices
**Observed** — Uses Google Signals and device ID, but no modeling
**Device-based** — Each device is a separate user. No cross-device stitching at all
Why it matters for paid media: If you're on Device-based, a user who clicks your Google Ads campaign on their phone during lunch and converts on their laptop that evening shows up as *two separate users*. Your mobile campaigns look weaker than they are. Your user counts are inflated. And your attribution is fragmented across devices instead of telling you the real story.
In a recent audit, we found a property with Google Signals actively collecting data — but Reporting Identity set to Device-based. The signals were being gathered and then completely ignored for identity resolution. Cross-device journeys were invisible.
What to do:
Switch to Blended. This requires Google Signals to be enabled (see Setting 3). If you're concerned about data thresholding in small-volume reports, Blended is still the right choice — the cross-device accuracy gains far outweigh occasional thresholding on niche segments.
Google support: Reporting Identity in GA4
Note: switching to Blended does not change historical data retroactively, but all future reporting will use the new identity model
---
2.. Google Ads Linking
Where to find it: GA4 Admin → Product Links → Google Ads Links
What it does: Links your GA4 property to one or more Google Ads accounts so conversion data, audiences, and attribution information can flow between the two platforms.
Why it matters for paid media: This is the bridge. Without it, Google Ads and GA4 are two separate systems looking at the same traffic and reaching different conclusions. Specifically:
- GA4 key events *cannot* be imported as Google Ads conversion actions
- GA4 audiences *cannot* be exported to Google Ads for remarketing
- Google Ads cost data won't appear in GA4 reports
- Cross-platform attribution is impossible to reconcile
We flag this as high severity in every audit. No link means Google Ads is optimizing in a vacuum.
But linking alone isn't enough. This is where many teams stop. Linking creates the *connection*, but you also need to confirm:
1. Which GA4 key events are actually imported as Google Ads conversion actions
2. Whether Google Ads is using GA4 key events or its own tag-based conversions for bidding
3. Whether GA4 audiences are being exported and used in campaigns
These are questions that require checking inside Google Ads — GA4 can only confirm the link exists. In our audits, we've seen properties where GA4 was linked to Google Ads, but the agency was bidding against a completely different set of conversion actions than what GA4 was tracking. The link was there, but the data wasn't flowing the way anyone assumed.
What to do: Verify the link exists. Then open Google Ads and confirm which conversion actions are active for bidding — and whether they're sourced from GA4.
Google support: Link GA4 to Google Ads
Additional Notes:
- Open Google Ads → Goals → Conversions → check which conversion actions are set to "Primary" for bidding
- Verify that primary conversion actions are sourced from GA4 (not from a separate Google Ads tag)
- If GA4 key events are not imported as conversion actions, import them and set the right ones as Primary
---
3. Google Signals
Where to find it: GA4 Admin → Data Collection → Google Signals data collection
What it does: Enables GA4 to collect cross-device data from users who are signed into Google and have Ads Personalization turned on. This powers cross-device reporting, demographics, and interest data.
Why it matters for paid media: Google Signals is the data source that makes Blended Reporting Identity work (see Setting 1). Without it:
- Cross-device attribution is weaker — you can't see that the same user interacted with your ads on multiple devices
- Demographic and interest data in GA4 is limited
- Remarketing audiences are narrower because they can't leverage cross-device signals
- Blended Reporting Identity can't use its full cross-device stitching capability
What to do: Enable Google Signals. Be aware that enabling it can cause *data thresholding* — GA4 may withhold rows in reports where user counts are too small to protect anonymity. For most properties with reasonable traffic volume, this is a non-issue. For very low-traffic properties, you may see some rows hidden in detailed reports, but the cross-device accuracy gains are worth it.
Google support: Google Signals data collection
---
4. Attribution Settings
Where to find it: GA4 Admin → Data Display -> Events -> Attribution Settings
What it does: Controls which attribution model GA4 uses to distribute conversion credit across marketing touchpoints. The options include Data-Driven Attribution (DDA), Last Click, and others.
Why it matters for paid media: Your attribution model determines how conversion credit is divided between channels. If you're still on Last Click:
- Branded search gets over-credited (it's usually the last click before conversion)
- Upper-funnel paid campaigns — display, video, paid social — get under-credited because they assist conversions rather than closing them
- You may cut spend on campaigns that are actually driving awareness and consideration, because Last Click doesn't see their contribution
Data-Driven Attribution uses machine learning to analyze your specific conversion paths and distribute credit based on actual contribution. Google recommends it for all properties with sufficient conversion volume.
What to do: Switch to Data-Driven Attribution. If your property has very low conversion volume (fewer than ~300 conversions/month across all key events), DDA may not have enough data to model effectively — but for most paid media properties, DDA is the right choice.
Google support: Attribution settings in GA4
Note: changing the attribution model affects *future* reports; historical data will recalculate based on the new model
---
5. Audience Configuration
Where to find it: GA4 Admin → Audiences (or GA4 left nav → Admin → Data Display → Audiences)
What it does: GA4 Audiences are user segments built from behavioral data — pages visited, events triggered, conversions completed, time on site, and more. Audiences can be exported to Google Ads for remarketing and exclusion targeting.
Why it matters for paid media: Custom audiences are one of the most powerful — and most underused — bridges between GA4 and Google Ads. Without them:
- Remarketing campaigns target overly broad segments ("All Users") instead of high-intent visitors
- You can't exclude past converters from acquisition campaigns (wasting spend on people who already converted)
- You miss recovery opportunities like targeting form abandoners or users who visited a key page but didn't convert
- Lookalike/similar audiences in Google Ads have weaker seed lists
In a recent audit, a property had two GA4 audiences exported to Google Ads: "All Users" and "Converting Visitors." That's a start — but the high-value segments were missing entirely. There was no audience for estimate-page visitors who didn't convert, no form abandoners, no DMA-specific segments, and no past-converter exclusion list.
What to do: Build at least 3–5 custom audiences in GA4 and export them to Google Ads. Start with these:
1. **High-intent visitors** — users who visited your conversion page (pricing, contact, estimate request) but didn't convert
2. **Past converters** (for exclusion) — users who completed a key event in the last 30/60/90 days
3. **Engaged visitors** — users with 2+ sessions or engagement time above your property's average
4. **Form/funnel abandoners** — users who started but didn't complete your primary conversion flow
This is a strategic improvement rather than an urgent fix, but it's often the highest-ROI configuration change a paid media team can make.
Google support: Create, edit, and archive audiences
Confirm high-value audiences are exported to the linked Google Ads account
And in Google Ads, verify the exported audiences appear and are being used in campaigns
---
6. Enhanced Measurement
Where to find it: GA4 Admin → Data Streams → [your web stream] → Enhanced Measurement
What it does: Enhanced Measurement automatically tracks 8 types of user interactions without requiring any code or GTM changes:
1. Page views (always on)
2. Scrolls (90% scroll depth)
3. Outbound clicks
4. Site search
5. Video engagement (YouTube embeds)
6. File downloads
7. Form interactions
8. Page changes (for SPAs)
Why it matters for paid media: These signals feed your understanding of what paid traffic actually *does* on your site — beyond just converting or bouncing. Without them:
- You can't tell if paid traffic is actually reading your content (no scroll data)
- You don't know if users are searching your site for something they didn't find in your ad or landing page (no site search data)
- You miss engagement signals that happen *between* landing and converting — outbound clicks, file downloads, video watches
- Landing page analysis is weaker because you're only seeing page views and conversions with nothing in between
What to do: Enable all 8 toggles unless you have a specific reason not to. The most common reason to disable one is if you have a custom GTM implementation that tracks the same interaction more precisely (e.g., a custom scroll depth trigger at 25/50/75/100% instead of Enhanced Measurement's 90%-only scroll tracking). If you're not sure, turn them all on — you can always refine later.
Google support: Enhanced Measurement
Note: If you see "Form interactions" turned off, pay special attention — this is often disabled accidentally and it affects how GA4 reports on form engagement
---
Bonus: Are Your Key Events Actually Trustworthy?
Even if all 6 admin settings are configured correctly, your paid media optimization can still be built on bad data if your key events (formerly called "conversions") are misfiring.
Key events are the single most important signal you send to Google Ads. They're what Smart Bidding, tCPA, and Max Conversions algorithms learn from. If your key events are inflated, duplicated, or inactive, those algorithms are learning from bad data — and making bad decisions with your budget.
Here are three red flags we check in every audit.
Red Flag 1: Key Event Rate Over 100% Per User
If a key event fires more than once per user on average, something is wrong. A `generate_lead` event should fire once when someone submits a form — not on every page view.
**How to check:** In GA4, go to Reports → Engagement → Events. Look at your key events. Divide the "Event count" column by the "Total users" column. If any key event shows a rate significantly above 100%, investigate.
**Real example:** In a recent audit for a national home services company, we found a call tracking event with a *1,094,700% conversion rate per user* — 10,947 conversions attributed to a single user. The tracking tag was firing on every page view instead of on actual phone calls. Two form submission events also exceeded 110% per user.
If Google Ads is using these events for bidding, the algorithm thinks conversions are abundant and cheap. It's learning from dramatically overstated data.
**What to do:** Audit every key event. If the rate per user exceeds 100%, check the GTM trigger — it's likely firing on page load or a page view instead of the actual user action (form submission, button click, call initiation).
Red Flag 2: Configured But Inactive Key Events
GA4 lets you mark any event as a "key event." But marking it doesn't mean it's firing. In the same audit, the property had 18 configured key events — but only 5 were actively recording conversions in the last 90 days. The other 13 were ghost signals: configured in GA4, possibly imported into Google Ads, but producing zero conversion data.
*How to check:** Go to GA4 Admin → Events. Look at which events are marked as key events. Then check each one in Reports → Engagement → Events to see if it has any conversion count in the last 90 days.
**What to do:** Remove the key event designation from any event that hasn't fired in 90 days. If Google Ads has imported these as conversion actions, check whether they're set to "Primary" (used for bidding) or "Secondary" (observation only). Inactive primary conversions confuse Smart Bidding.
Red Flag 3: Too Many Key Events
If you have 10, 15, or 18 key events, Google Ads doesn't know which ones actually matter. More signals isn't better — *clearer* signals are better.
Think about it from the algorithm's perspective: if you're telling it that a page scroll, a video play, a form submission, and a phone call are all equally important "key events," it will optimize for whatever is cheapest and easiest to get — which is probably the scroll, not the form submission.
**Our recommendation:** Consolidate to 3–5 key events per property, each representing a distinct business outcome. For lead-gen, that's typically:
- Primary form submission (lead, estimate request, demo)
- Phone call (validated, not just page-level tracking)
- Secondary form (contact, newsletter)
For ecommerce, it's usually `purchase`, `add_to_cart`, and `begin_checkout`.
Checklist:
- [ ] Count how many key events are configured in GA4 Admin → Events
- [ ] For each key event, check the conversion rate per user (Event count ÷ Total users) over 90 days
- [ ] Flag any key event with a rate over 100% per user — investigate the trigger
- [ ] Flag any key event with zero conversions in 90 days — remove the key event designation
- [ ] If you have more than 5 key events, evaluate whether each one represents a truly distinct business outcome
- [ ] In Google Ads, verify which imported GA4 key events are set to "Primary" vs "Secondary"
---
What The Helm Automates
We built these exact checks into The Helm. When you connect a GA4 property, The Helm runs all 6 admin configuration checks via the GA4 Admin API and a full key event validation against 90 days of GA4 data — automatically.
Each check produces:
- A severity rating (info, low, medium, high, critical)
- A confidence score based on evidence quality
- A recommended fix with an owner category (GA4 Admin, GTM team, Google Ads platform, etc.)
- A decision impact assessment: what measurement risk exists, what the optimization consequence is, and what to do about it
The results feed into a weighted Measurement Integrity Score (0–100) that gives you a single number for how trustworthy your paid media measurement is.
If you want to start with something lighter, our free [AI Measurement Integrity Scan](https://the-helm.co) evaluates your website and GA4 setup without requiring full GA4 access — just your website URL and email.
---
The Full Pre-Spend Checklist
Before you scale paid media spend, run through this list. It takes 15 minutes and can save you from optimizing against bad data.
**GA4 Admin Settings:**
- [ ] Reporting Identity is set to **Blended**
- [ ] Google Ads is **linked** and key events are **imported** as conversion actions
- [ ] Google Signals is **enabled**
- [ ] Attribution model is **Data-Driven Attribution**
- [ ] At least **3 custom audiences** are built and exported to Google Ads
- [ ] All relevant **Enhanced Measurement** toggles are on
**Key Event Integrity:**
- [ ] No key event has a conversion rate **over 100%** per user
- [ ] No configured key events are **inactive** (zero conversions in 90 days)
- [ ] Total key events are **5 or fewer**, each representing a distinct business outcome
- [ ] In Google Ads, primary conversion actions are **sourced from GA4** (not separate tags)
**From Part 1 (UTM Audit):**
- [ ] No **(not set)** campaign names in Traffic Acquisition
- [ ] No **inconsistent casing** in source/medium values
- [ ] **Direct/(none)** traffic is below 50% of total sessions
- [ ] **Google Ads is linked** (appears here and in the admin settings above — it's that important)
---
What's Next
This is the second post in our paid media measurement audit series. We've covered the campaign tagging layer (UTMs) and the property configuration layer (admin settings + key events).
Next, we'll go into the data quality layer — the problems that corrupt your numbers even when your settings and tagging are correct:
- **Duplicate event detection** — similar event names that inflate counts
- **Bot and spam traffic** — sessions from non-human sources that distort your metrics
- **Session logic validation** — landing pages showing (not set) and key event rates that don't add up
- **Self-referral and direct traffic analysis** — when your own domain shows up as a traffic source
Same 90-day lens. Same goal: trust the numbers before you increase media budgets.
---
*Related posts:*
- How to Audit Your Paid Media Measurement in 2026: Start with UTM Tracking
*This is part of how we're revamping our Google Analytics auditing methods in 2026. These checks — and many more — are automated inside The Helm, our AI agent for Google Analytics 4 - https://app.analyticsmates.com *
See Article Images
