Google Analytics数据差异咨询：3月28日起当日与次日数据不符

Hey there, let’s unpack this data discrepancy you’ve been seeing since March 28—where your Google Analytics (GA) numbers drop between checking them the same evening and the next day (like your 26,649 user count falling by 2000 on April 16-17). The short answer: AMP pages can contribute to this, but they’re not the only possible cause. Let’s break down both the AMP-specific quirks and standard GA behavior that might be at play.

Standard GA Data Processing Explains Most Day-to-Day Shifts #

First, it’s important to remember GA doesn’t lock in real-time data immediately. Overnight, it runs several cleanup processes that can reduce reported numbers:

• Hit deduplication: GA automatically removes duplicate hits (e.g., users refreshing pages repeatedly) that get counted in real-time but filtered out during post-processing.
• Bot filtering: GA’s bot detection runs on a schedule—some bot traffic that shows up in real-time reports gets stripped out once full data is processed.
• Sampling adjustments: If your site has high traffic, real-time reports might use sampled data that gets refined with full, unsampled data the next day, leading to more accurate (and sometimes lower) counts.

AMP-Specific Factors That Could Amplify Discrepancies #

Since your site has AMP pages, there are unique tracking nuances that might be contributing:

• AMP Analytics vs. Standard gtag.js: If you’re using the <amp-analytics> component instead of the standard gtag.js on your AMP pages, hits are sent to GA’s measurement protocol with slightly different parameters. These hits might be processed later or prioritized differently, leading to gaps between real-time and finalized counts.
• Offline tracking behavior: AMP can store hits when users are offline and send them later. If your evening snapshot included pending offline hits that never actually got sent (e.g., the user never came back online), those counts would drop when GA cleans up unprocessed hits overnight.
• Cross-domain/cache tracking gaps: If your AMP pages are hosted on a subdomain (like amp.yoursite.com) or accessed via Google’s AMP Cache, improper cross-domain tracking setup could cause sessions to be counted twice initially. When GA deduplicates these cross-domain sessions overnight, you’ll see a drop in user counts.

How to Test if AMP is the Culprit #

To narrow this down, try these quick checks:

• Segment AMP vs. non-AMP traffic: In GA, create a segment for AMP pageviews (use pagePath or a custom dimension you’ve set for AMP pages) and compare the day-of vs. next-day changes for AMP vs. regular traffic. If only AMP traffic shows the drop, that’s a clear sign to focus on your AMP tracking setup.
• Audit your AMP analytics tag: Double-check that your <amp-analytics> tag uses the same GA property ID as your main site, and includes transport:beacon (this ensures hits are sent reliably even if the user navigates away quickly).
• Compare real-time vs. historical reports: Pull the real-time report from April 16 evening and compare it to the historical Audience > Overview report on April 17. Look at metrics like users, sessions, and pageviews to see if the drop is consistent across all metrics or isolated to AMP-related ones.

A 7-8% drop (like your 2000 out of 26k) isn’t unheard of with GA’s normal processing, but since this started suddenly on March 28, it’s worth checking if you made any changes to your AMP tracking or GA settings around that date.

内容的提问来源于stack exchange，提问作者Robert Thomas