If you are looking at GA4’s user counts to evaluate Google Ads performance and you have not checked your Reporting Identity setting, the numbers you are working with may be systematically wrong — not wrong in a way that is obvious, but wrong in a way that makes your paid traffic look either better or worse than it is.
Reporting Identity is a property-level setting that controls how GA4 identifies users across devices and sessions. It determines whether GA4 stitches together a logged-in user’s mobile and desktop sessions as a single user journey, and whether it fills in gaps from cookie consent refusals using behavioral modelling. The setting has three options and they produce materially different user counts.
Most GA4 properties have never touched this setting. Most are on the wrong option for their use case.
The Three Reporting Identity Options
Find this setting at: Admin > Data display > Reporting Identity.
Blended (Default)
GA4 uses all available identity signals to count users:
- User-ID: If your site passes a user ID (from login state) to GA4, this is the primary identifier. Two sessions with the same User-ID are the same user, regardless of device or browser.
- Google signals: For users who are logged into a Google account and have opted into ads personalization, GA4 uses Google’s cross-device identity graph to stitch sessions across devices.
- Device ID: Falls back to the cookie-based device identifier when neither User-ID nor Google signals are available.
- Modelled data: Where consent is not granted and signal data is sparse, GA4 fills in estimated user counts using behavioral modelling from consented users with similar patterns.
Blended gives you the most complete and deduplicated view of your users. If someone browses on mobile, converts on desktop while logged in, Blended counts that as one user journey rather than two.
The risk: Modelling introduces estimated data. The user counts you see in Blended mode are partly real (observed signal data) and partly modelled (statistical inference). GA4 labels modelled data with a small indicator, but it is not always obvious which numbers are observed vs. estimated.
Observed
GA4 uses User-ID and Google signals but excludes modelled data. It only counts users for whom it has actual observed signal data.
This option is appropriate when you need to report on confirmed, observed data without statistical estimation — for example, in regulatory contexts where reporting on modelled user data would be misleading or problematic.
The trade-off: User counts will be lower than Blended because gaps from consent refusals are not filled. A high consent refusal rate (common in EU markets) can make Observed user counts significantly lower than the actual number of users on your site.
Device-based
GA4 uses only the device ID (cookie-based) and ignores User-ID and Google signals entirely. No cross-device stitching, no modelling. Each device is counted as a separate user.
This is Universal Analytics behavior. It is the most transparent and most limited approach — you see exactly what the cookies report and nothing else.
When to use it: If your property has User-ID implemented incorrectly (assigning the same User-ID to multiple users, or firing User-ID on non-authenticated sessions), Device-based will be more accurate than Blended, which would be stitching incorrect journeys together. It is also useful as a reference point for comparing with historical data from Universal Analytics properties.
How Each Setting Affects Your Numbers
The impact is not abstract. Here is what changes depending on which setting you use:
User counts: Blended produces the highest user counts (cross-device stitching reduces duplicates from the same user on multiple devices). Device-based produces the highest raw count by treating each device independently. Observed sits in between.
Wait — that seems counterintuitive. If Blended stitches sessions together, should it not produce lower counts than Device-based?
Yes: within the set of users who are identifiable across devices, Blended reduces duplicates. But Blended also adds modelled users for those who opted out of consent — users who Device-based mode simply does not count at all. The net effect on your total user count depends on the balance between these two forces and is property-specific.
Session and engagement metrics: Cross-device stitching in Blended can change engagement time, session counts, and pages per session by combining sessions that Device-based treats as separate.
Conversion paths: In Attribution and the user journey reports, Blended provides more complete cross-device conversion paths. A user who clicked a Google Ad on mobile but converted on desktop will show the complete two-step path in Blended, but two separate incomplete sessions in Device-based.
The Consent Mode Connection
This is where Reporting Identity becomes most important for Google Ads advertisers.
When a user refuses cookie consent (common in EU markets), GA4 cannot set a first-party cookie. The user visits your site, but GA4 has no device ID to associate with their session.
Without any Reporting Identity modelling, that user’s sessions are invisible. Every page they viewed, every event they triggered, every conversion they completed — gone from your data.
Google Consent Mode v2, combined with Blended Reporting Identity, partially recovers this data through behavioural modelling. GA4 observes the consenting users’ patterns and uses those patterns to estimate what the non-consenting users did.
The practical impact: if 40% of your EU visitors refuse consent and you are on Device-based Reporting Identity, your GA4 data is showing you roughly 60% of your actual user base. You are making decisions about campaign performance on incomplete data.
On Blended, that gap is partially filled by modelling. Your user counts and conversion numbers will be higher — not because more people converted, but because GA4 is now including estimated activity from users it previously had no data on.
Important: modelled data is not factual data. It is a statistically informed estimate. For general trend analysis and campaign comparison, it is useful. For exact counts that need to be reconciled with backend sales data, treat modelled numbers with appropriate uncertainty.
How This Affects Google Ads Performance Evaluation
Every time you open GA4 to evaluate a Google Ads campaign — checking which campaign drove more engaged users, comparing conversion rates across traffic sources, analyzing the attribution path — your numbers are a function of your Reporting Identity setting.
Scenario 1: Blended, EU-heavy traffic, high consent refusal rate
Your user counts are inflated by modelling. Your conversion rates may look higher than they would in Observed mode because GA4 is including modelled conversions. Your Google Ads traffic may appear to have a lower cost per user than the actual observed cost.
Scenario 2: Device-based, EU-heavy traffic, high consent refusal rate
Your user counts are understated. Conversion rates from paid traffic are understated because consented users (who are counted) may behave differently from non-consented users (who are not). Your campaign performance data is systematically incomplete.
Scenario 3: User-ID implemented, Blended
If your users log in and you pass User-ID to GA4, Blended correctly stitches cross-device journeys. A user who clicked a Google Ad on mobile and converted on desktop shows as a single user who converted, with the correct attribution. Device-based would show two separate sessions with incomplete paths.
Which Setting Should You Use?
Use Blended if:
- You want the most complete picture of user behavior
- You have Consent Mode v2 implemented and want modelled data to fill consent gaps
- Your users log in and you have User-ID implemented correctly
- You understand that some numbers are modelled estimates
Use Observed if:
- You need to report on confirmed, observed data only (regulatory or audit requirement)
- You want to see what GA4 actually knows vs. what it has estimated
- You are debugging unexpected numbers and want to rule out modelling as a cause
Use Device-based if:
- Your User-ID implementation is broken or inconsistent (fixing the underlying problem is preferable, but Device-based prevents bad stitching in the meantime)
- You need numbers that match how Universal Analytics reported users
- You are in a context where cookie-based counting is the agreed-upon methodology
How to Check and Change the Setting
Admin > Data display > Reporting Identity.
Select your option. The change applies going forward and affects how historical data is displayed in reports (because Reporting Identity affects the reporting layer, not the raw data collection).
Important: changing Reporting Identity does not change what data GA4 has collected. It changes how that data is assembled into user-level metrics. You can switch between options and compare how your numbers change.
Diagnostic exercise: Switch between Blended and Device-based, then look at your user counts in the Acquisition overview report for the past 90 days. If the difference is small, your property has high consent rates and few cross-device users. If the difference is large, you have significant consent gaps or cross-device usage that Blended is stitching together.
Annotating When You Change It
If you switch Reporting Identity, add an annotation in GA4 with the date and what you changed. Report readers in the future need to know that the user count methodology changed on a specific date, or trend analysis across that date will be misleading.
GA4 annotations: Admin > Annotations > Create annotation.
Key Takeaway
Reporting Identity is the setting that determines what “a user” means in every GA4 report. It is not a setup checkbox — it is a data philosophy decision that affects every user-level metric you ever read.
For Google Ads advertisers in markets with active Consent Mode usage (particularly EU), Blended with Consent Mode v2 is the configuration that gives you the most complete view of campaign performance. For properties where completeness is less critical than observational accuracy, Observed is the right choice.
Check the setting. Understand what mode you are in. Then read your user counts knowing what they represent.
Up next in the GA4 Advanced Series: GA4 Cohort Analysis — How to Measure Retention and Campaign Quality Over Time
Related Posts
GA4 Channel Groupings: Why the Default Is Wrong and How to Fix It
12 min read
GA4 Cohort Analysis: How to Measure Retention and Campaign Quality Over Time
11 min read
GA4 Predictive Audiences: How to Use Purchase Probability and Churn Signals in Google Ads
12 min read
Need Help With Your Google Ads?
I help e-commerce brands scale profitably with data-driven PPC strategies.
Get In Touch