Why Google Ads, GA4, and CRM Data Never Match

Attribution models, reporting windows, and customer journey nuances create gaps between data from advertising platforms, analytics systems, and CRMs. If you plan PPC channel budgets by comparing data from Google Ads, Meta Ads, GA4, and CRM/CMS, you have probably already noticed: the numbers almost never match. In this case, a logical question arises — which data should you rely on in reporting and how should you optimize marketing considering the real impact of channels on business results. Often, marketers assume that the problem lies in tracking: insufficiently accurate UTM tags, analytics setup errors, or the need for a more complex data collection system. However, in most cases, the root cause lies much deeper — in the very nature of attribution.

The Attribution Trap: Why the “Data-Driven” Approach Doesn’t Always Work

For many years, the digital marketing market formed around the concept of data-driven decisions. It was assumed that correctly configured analytics systems could accurately show which channels work effectively and which do not. The marketer only had to “follow the data.” However, attribution quickly becomes a source of distorted conclusions. Without proper data interpretation, companies begin to reallocate budgets based on an incomplete picture, which can lead to flawed business decisions. It is important to understand the key difference: attribution distributes conversion value among channels, but it does not determine which specific channel actually caused that conversion. At first glance, this may sound like a theoretical detail, but it is precisely this difference that lies at the core of most problems with measuring advertising effectiveness.

Why Ad Platform, Analytics, and CRM Data Cannot Be Identical

Before attempting to “align” metrics between Google Ads, GA4, and CRM, it is necessary to accept a basic fact: achieving a complete match between these systems is impossible. The reason is that each platform was created for different purposes, uses different methodologies, and captures different stages of the customer journey.

Customer Journey as the Basis for Understanding Discrepancies

Imagine a typical situation: A user saw an ad in Meta Ads, then interacted with remarketing on YouTube, and later performed a branded search query in Google and made a conversion — all within seven days. According to standard attribution windows:

• Meta Ads will record one conversion.
• Google Ads will also record one conversion.
• GA4 and CRM will show only one conversion and, most likely, attribute it to Google Ads Paid Search.

Does this mean Meta Ads “invented” a duplicate conversion? No. The Meta platform has no access to user interactions inside Google Ads and cannot determine that the conversion has already been credited by another system. At the same time, GA4 and CRM can almost completely ignore Meta Ads’ contribution to the customer journey. If you rely solely on this data, you might come to the wrong conclusion about the need to cut the Meta Ads budget in favor of branded search in Google Ads.

Structural Causes of Data Discrepancies

The problem is not limited to different attribution models alone. There is a range of systemic factors due to which metrics between platforms will always differ.

Differences in Attribution Dates

Advertising platforms typically tie a conversion to the date of the ad click. In contrast, GA4 and CRM systems most often record a conversion by the date the target action was actually performed. If the customer journey lasts several days or weeks, this automatically creates additional discrepancies between reports.

Cross-Device User Behavior

Another common scenario — a user transitions from Google Ads on a mobile device and later returns via organic search from a desktop and completes a conversion. In this situation:

• Google Ads may credit the conversion to the advertising channel.
• The CRM system may determine the source as SEO or direct traffic.
• GA4 will show its own version of the user journey depending on available signals and identity settings.

The main problem is that a CRM is far from always capable of combining mobile and desktop sessions into a single user.

Privacy Restrictions and Data Loss

Attribution quality is significantly affected by:

• ad blockers;
• browser-level tracking prevention;
• cookie restrictions;
• consent banners and user opt-outs from tracking.

As a result, a significant portion of conversions does not make it into analytics systems at all. Some advertising platforms partially compensate for these losses through modeled conversions, meaning conversions modeled based on machine learning algorithms. However, CRM systems usually do not have access to such mechanisms and cannot restore the full picture of traffic sources.

What Can Be Fixed and What Cannot

Part of the problems can indeed be minimized using technical improvements:

• server-side tagging;
• offline conversion imports;
• unified UTM tags;
• correct consent mode configuration;
• enhanced user identification.

However, even with an ideal system configuration, structural differences between advertising platforms, analytics, and CRM will remain. That is why expecting 100% correlation between Google Ads, GA4, and CRM data is incorrect from a methodological standpoint.

Single Source of Truth: The Attribution Trap

After teams accept the fact that figures between Google Ads, GA4, and CRM will always differ, the next decision usually follows — choosing a single source of truth. Most often, this system becomes GA4 or CRM. It is at this exact moment that marketers fall into the so-called attribution trap. Each tool operates on the basis of a specific attribution model. And regardless of whether it is first-click, last-click, linear, time decay, or data-driven attribution, every model has fundamental limitations.

Every Attribution Model Has “Blind Spots”

Last-click attribution

One of the simplest models to understand — and at the same time, one of the easiest to manipulate. Last-click attribution awards all conversion value to the final point of contact before the conversion. In practice, this often means that the main value is received by branded search or direct traffic. As a result, the model systematically underestimates the channels that generate demand, work with brand awareness, and create the initial interest in a product.

First-click attribution

First-click attribution works on the opposite principle. The focus is placed on the user’s first interaction with the brand, while all subsequent stages of the customer journey are practically ignored. Such a model demonstrates the sources of primary audience acquisition well, but it fails to account for the channels that actually helped guide the user to the conversion.

Linear and time-decay attribution

At first glance, these models look more balanced. Linear attribution distributes value evenly among all touchpoints, while time-decay attribution gives more weight to interactions that occurred closer to the time of conversion. However, even such approaches remain largely arbitrary. A logical question arises: why should all touchpoints receive identical value? And why exactly should proximity to conversion determine the importance of a channel? Real customer journeys rarely follow rigid scenarios or a mathematically uniform distribution of impact.

Data-driven attribution

Data-driven attribution is often positioned as the most technologically advanced model. The idea is that the advertising platform or analytics system independently determines which distribution of value most accurately reflects reality. However, in practice, this model remains a “black box.” Platforms almost never disclose the details of how exactly the algorithms for distributing value between channels work. Therefore, marketers are effectively forced to trust the systems without a full understanding of the logic behind their decisions.

What Happens When a Company Chooses a Single Source of Truth

Attribution answers only one question: if a conversion has already occurred, which touchpoints should receive credit for it. The problem arises when the entire decision-making system is reduced to a single tool. In this case, the company automatically inherits all the limitations and “blind spots” of the attribution model on which that system is based.

Risks of focusing exclusively on CRM

If a company relies only on CRM data, it effectively operates within the logic of last-click attribution. As a result, the marketing strategy gradually concentrates mainly on branded search and channels that are located as close as possible to the conversion. The problem is that demand generation and upper-funnel activities begin to look ineffective, even though they are the very ones that create future demand. After a few years, the business may face a situation where demand begins to shrink, despite the CRM continuing to demonstrate “good results.”

Risks of focusing only on advertising platforms

The opposite extreme is total dependence on data from advertising systems. In this case, companies often encounter significantly inflated performance indicators. Marketing platforms can report revenue that exceeds the actual financial results of the company by 2 to 4 times. This creates a conflict between marketing and the finance department:

• marketing argues for the necessity of increasing budgets;
• finance sees a completely different picture of the real business income.

Why even GA4 is not an absolute source of truth

Against the background of CRM and advertising platforms, GA4 is often perceived as the most “neutral” solution. However, this system also has fundamental limitations. GA4 analyzes only that part of the customer journey that takes place directly on the website or in the app. At the same time, a significant part of the marketing impact can occur outside the boundaries of web visits. For example:

• awareness campaigns;
• video campaigns;
• brand lift activities;
• campaigns for reach and ad memorability.

Such interactions do not always lead to an immediate transition to the website, but they can still significantly influence the user’s future behavior and the probability of a conversion.

Why Business Begins to Talk About Incrementality

When it becomes obvious that all attribution systems have structural limitations and “blind spots,” the next question naturally arises — incrementality. In other words: did the advertising campaign truly create additional conversions that would not have occurred without it?

Incrementality Tests: The Ideal Solution or a New Limitation?

Incrementality is an approach to measuring marketing effectiveness that evaluates the result created specifically due to an advertising campaign. In other words, it refers to conversions that would not have occurred without the ad being shown. The concept of incrementality is often explained through the model of “two parallel realities”:

• in the first one, the advertising campaign was launched;
• in the second one, it was not.

The difference between the results in these two scenarios is the incremental impact of the advertising. Everything else is the demand or conversions that the business would have received regardless of the campaign’s activity.

Attribution vs. Incrementality: What is the Key Difference

This distinction holds much more significance than it might appear at first glance. A large portion of conversions recorded by advertising systems — especially in retargeting and branded search campaigns — often comes from users who were already prepared to make a purchase even without the ad. Such users:

• were already at the decision-making stage;
• already knew the brand;
• already demonstrated a high intent to convert.

Attribution credits the conversion to the advertising channel if the user interacted with the ad before purchasing. However, incrementality attempts to answer a different question: did the advertisement truly cause this conversion. For budget decision-making, this difference is critically important.

Why High ROAS Does Not Always Mean Real Effectiveness

A retargeting campaign can demonstrate a very high ROAS within attribution models but generate minimal incremental effect at the same time. In practice, this means:

• the campaign receives credit for conversions that would have occurred even without the advertisement;
• the marketing team sees strong performance metrics;
• the business continues to invest budget into a channel that actually generates almost no additional demand.

In some cases, after turning off such a campaign, the total volume of conversions remains virtually unchanged. Thus, the company is effectively paying for an illusion of effectiveness formed by attribution models and a “single source of truth.”

How Incrementality Tests Are Conducted

Incrementality testing is based on an experimental approach. To conduct a test, it is necessary to create two groups:

• test group — the audience that sees the ad;
• control group — the audience that does not see the ad.

After this, the difference in results between the groups is compared.

Geo holdout testing

One of the most common approaches is geo holdout testing. Within this method, the market is divided into several geographically similar regions:

• in some of the regions, the campaign is launched;
• in others, the advertising is temporarily turned off.

After the test concludes, the difference in conversions between the regions is analyzed. Advantages of the approach:

• relative simplicity of implementation;
• practicality;
• sufficiently high reliability of results.

That is why geo holdout testing is often considered one of the most realistic ways to evaluate incrementality.

Audience holdout testing

Google and Meta platforms also allow for the creation of holdout groups directly inside the advertising systems. In this case, a certain percentage of the target audience is intentionally excluded from seeing the ads. Next, the analysis principle remains analogous to geo holdout testing: the company compares the behavior of users who saw the ad and those who did not. However, this approach has an important limitation. Since the testing is based on the data of a specific advertising platform, comparing incrementality between different ad networks is practically impossible. For example:

• the results of a Google Ads holdout test can only be correctly compared with other campaigns inside Google Ads;
• the same applies to Meta Ads.

Comparison between different platforms loses methodological value.

Time-based testing

Another approach is time-based testing. In this case, the campaign is completely stopped for a certain period of time, after which the change in the total volume of conversions is analyzed. If results barely change after the pause, this may indicate a low incremental impact of the campaign. However, this method is considered one of the riskiest. The results can be significantly influenced by external factors:

• seasonality;
• competitor actions;
• market changes;
• news environment;
• economic situation.

Furthermore, if the campaign truly created an incremental effect, turning it off during the test could lead to a real drop in business results during the testing period.

Is Incrementality Right for Your Specific Business?

If a company operates with large advertising budgets — conditionally from €1 million per month and more — the topic of incrementality testing is likely already a part of its marketing strategy. However, for businesses with smaller budgets, the situation looks different. In most cases, incrementality tests at low spending volumes prove to be of little use for practical application. The reason is simple: to obtain statistically reliable results, a significant difference between the test group and the control group is required, and this demands a large volume of data. In turn, generating a sufficient volume of data almost always requires substantial advertising investment.

Where Incrementality Can Be Useful Even with Limited Budgets

Despite this, there are specific scenarios where a business can use simplified approaches to evaluate potentially problematic areas. The most typical example is branded search campaigns. In this case, it is worth analyzing Auction Insights in Google Ads and assessing the level of competition for your own brand.

If competitors are actively bidding on your brand

In such a situation, branded search campaigns often remain necessary. The reason is that advertising helps intercept demand that the brand itself created through other marketing activities. Without branded search, a portion of this demand may shift to competitors.

If competition for the brand is minimal

If competitors practically do not target the company’s brand, the situation changes. In this case, the business can:

• temporarily suspend branded search campaigns;
• allow SEO and organic search results to intercept branded demand;
• reduce advertising expenditures without a significant loss of conversions.

Triangulation as a Practical Approach to Decision-Making

If attribution has fundamental limitations and incrementality is accessible primarily to large advertisers, a logical question arises: how should marketing decisions be made then? One of the most practical approaches becomes triangulation. This refers to using all available systems simultaneously — with an understanding of their limitations and “blind spots.” The key task of a marketer is not to find a single “correct” number, but to form a stable system of data interpretation. Equally important — explaining this approach to clients, management, and finance teams, so that the company does not fall into the “single source of truth” trap.

Starting with CRM or CMS is key

CRM and CMS systems record real sales, transactions, and business revenue. That is why all other figures should be perceived as an attempt to explain these results, rather than as absolute truth. For example:

• Google Ads and Meta Ads can collectively report $50,000 in revenue;
• Shopify, at the same time, shows only $35,000 of actual income.

In this case, it is Shopify that reflects the business reality. In addition, a CRM or CMS often remains the only system capable of correctly determining whether a customer was new or already existing. Advertising platforms do not provide such accuracy.

Why this matters for marketing effectiveness evaluation

It is precisely the CRM that allows for the analysis of nCAC (new customer acquisition cost) — the cost of acquiring a new customer. This is a critically important metric because it helps build budgeting around customers that the business would not have acquired without marketing activity. This approach is much closer to evaluating real incremental impact than the standard attribution models of advertising platforms.

How to use advertising platform data in triangulation

Following this, the results of advertising systems must be overlaid onto the customer journey. The difference between the revenue in the CRM and the revenue shown by advertising platforms is effectively an interpretation of the channels’ contribution from the perspective of the ad networks. The task of the marketer is to understand:

• which campaigns duplicate each other;
• where overlap occurs between channels;
• which touchpoints actually create additional demand.

For example, if a business simultaneously launches:

• Demand Gen campaigns;
• Meta Ads retargeting;
• branded search campaigns;

an overlap of audiences and attribution will almost guaranteed occur between them. Accordingly, the results in reporting will be duplicated as well. It is exactly in such situations that time-based incrementality tests can help determine which of the channels creates a greater real impact.

How to Improve the Triangulation Approach

Optimization of attribution windows

The longer the customer journey, the more difficult it is to interpret the effectiveness of campaigns. One practical approach is to segment campaigns according to the stages of the customer journey and separately configure:

• attribution windows;
• micro-conversions;
• performance evaluation models.

In many cases, shorter attribution windows provide more precise signals for optimization, provided the system is configured correctly.

Analysis of ratios between platforms

The gap between data from advertising platforms and CRM/CMS usually remains relatively stable. That is why it is advisable to create separate reports that track these ratios over time. For example:

• the ratio of revenue in Google Ads to revenue in the CRM;
• the ratio of attributed conversions to confirmed sales;
• the share of new customers across different channels.

If these proportions remain stable, the measurement framework operates predictably. However, if the ratios begin to change sharply, this may signal:

• tracking issues;
• a shift in customer behavior;
• a change in the incrementality of specific channels;
• the emergence of new sources of overlap between campaigns.

Why triangulation is more valuable than an “ideal figure”

Triangulation does not yield a single universal number that supposedly describes marketing effectiveness with absolute precision. However, this approach allows for the construction of:

• a consistent decision-making system;
• a realistic data interpretation;
• a stable framework for evaluating channels.

In the long term, this is significantly more valuable for a business than the illusion of absolute accuracy promised by separate attribution models or a “single source of truth.”

Discrepancies between data from Google Ads, Meta Ads, GA4, and CRM are neither an anomaly nor necessarily the result of analytics setup errors. They are a natural consequence of each system using its own attribution logic, different methods of user identification, and interpreting the customer journey differently.

That is why attempts to find an “ideal” or single source of truth often lead to distorted management decisions. A CRM may systematically overvalue last-click channels, advertising platforms may inflate their own contribution to revenue, and GA4 may fail to account for a significant portion of upper-funnel impact.

Incrementality testing helps bring businesses closer to understanding the actual contribution of advertising to business results; however, such an approach requires substantial data volumes and budgets, thus remaining accessible primarily to large advertisers.

For most companies, the most practical solution becomes triangulation — a comprehensive approach in which the marketing team analyzes data from multiple systems simultaneously, understanding their respective strengths and limitations.

In modern digital marketing, effectiveness is less and less frequently defined by a single metric or a single platform. Far more important is the ability to correctly interpret data, evaluate channels within the context of the full customer journey, and make decisions based on systemic analysis rather than the illusion of absolute accuracy.

Read this article in Ukrainian.