Have you ever experienced something like "the numbers slightly differ every time you open the report" or "the figures do not match even when aggregating data from different periods" while using Google Analytics 4 (GA4)? Such fluctuations in numbers are a common concern faced by many GA4 users.

Many of these issues stem from a lack of understanding of the fundamental mechanics of GA4, which are fundamentally different from Universal Analytics (UA). If you do not know how data is collected and processed, the assumptions of your analysis can collapse, potentially leading to misguided decision-making.

This article explains five often-overlooked important settings and analytical tips that are essential for ensuring data accuracy in GA4 and conducting meaningful analysis. By understanding and practicing these points, you can enhance the reliability of your data and maximize the potential of GA4.

1. Why do the numbers fluctuate? The pitfalls of GA4 "data sampling" and how to avoid it

The main reason numbers in reports fluctuate slightly is "data sampling." This is a mechanism that, when the volume of data being analyzed is enormous, samples a portion of the data to infer overall trends instead of processing all data. While this reduces system load and can display reports quickly, it may sacrifice the accuracy of the numbers.

In the standard version of GA4 (free version), once the number of events subject to analysis (queries) in Exploration Reports exceeds 10 million, sampling is automatically applied.

You can check whether sampling is occurring via the "Data Quality Icon" in the upper right of the report screen. If a green checkmark (✔) is displayed, it indicates "unsampled data," meaning all data is used. Conversely, if a red exclamation mark (!) appears, it indicates "data with high sampling rates," meaning that it is aggregated based on only a subset of data. Clicking this icon shows a message stating, "Based on x% of available data," allowing you to confirm the specific sampling rate.

The main methods to avoid sampling and analyze with more accurate data are as follows:

• Shorten the analysis period: This is the easiest and most effective method. For example, if you want to analyze a year's worth of data, you can break it down into monthly or quarterly analyses to keep the number of events per analysis within limits.

• Reduce unnecessary dimensions and metrics: By limiting the items included in the report to the bare minimum, you can reduce the query load and make it easier to avoid sampling.

• Export to