Each log only sees the requests that reach it. Here is what each source misses and how to read them together.
Every log sees a different slice
A request can be answered by the CDN, the web server or WordPress. Each layer records only what reaches it, so the counts differ even when nothing is wrong.
What the CDN sees
The CDN sees nearly every request, including pages it serves from cache and requests it blocks. It is the only place a blocked crawler shows up at all.
What the server and WordPress see
The server’s access log sees requests that pass the CDN, including images and files. A WordPress plugin sees only requests that run WordPress. Pages served from the CDN cache reach neither, so a crawler can look absent when it was in fact served.
Impostors inflate raw counts
Raw logs count anything that calls itself Googlebot. A source that verifies crawler IP addresses counts fewer. Compare verified counts with verified counts.
Read absence carefully
If a crawler is missing from one source, it was not observed in that source; it may still have visited. Crawl Analytics shows pages not observed as upper bounds when a cache can serve them.
Use the right source for the question
Was a crawler blocked? Check the CDN or firewall. Did it reach the page? Check the server log. How fast did WordPress answer? Check the plugin. Ndovu takes the WordPress crawl-log plugin and Cloudflare as sources and lets you filter by source. See also why a crawler gets a 403.