This is the traffic sources graph for a post I created a few hours ago:
It's not just this post. Every single one of my posts is dominated by direct/unknown sources. Is it like that for everyone else in here? Or is it just me? It doesn't make any sense at all.
Top comments (39)
@jess Brought this issue up on Forem awhile ago, but seems to be fix (question mark?)
Describe the bug
Not seeing a breakdown of external referrers for some posts. In a team discussion, it was suggested that these views may have been from dev.to but miscategorized. Either way, it is unusual for there to be no other referrers.
Expected behavior
A clear breakdown of the different referral sources, such as:
Screenshots
I would raise up in the issues on Forem again since sometimes it is fixed, but later it comes back.
In the issue you mentioned, it's showing "All other external referrers":
But in my case, it's showing "Direct / Unknown":
Is it still the same issue? I read the current issues, but couldn't find a similar one. Maybe there are that many Direct / Unknown visits? From bots perhaps?
Probably similar? Not sure what Direct / Unknown is referring to, but I would bring this up on the Issues on GitHub.
But I am unsure what to say about the issue. We do not have the actual number of Direct / Unknown views, right?
Yes. Try your best to describe what your experiencing. Otherwise, if others have questions, they will let you know!
Yeah. Then I'll open an issue about my original problem - too many Direct / Unknown views. Is that ok?
That's fine. No need to ask permission to create issues on GitHub!
Haha. I am always afraid of creating issues π
Quick question: How long do you spend on this platform daily? I just wanted to know.
No worries! Though it is good experience to document things :)
I come and go throughout the day whenever I have time, though I am known to be very punctual on responses in most cases since I get notified. Back then, I believe a few people called me a bot because I am just everywhere lol.
The first post I ever read on this platform was yours, but I still couldn't find it. Yeah, you and @unitbuilds are EVERYWHERE.
I'm everywhere? π€
Yeah whatever post I read you are there in the comments
πΆβπ«οΈ No I'm not
@francistrdev I opened the issue. Still unsure whether it's a valid one.
github.com/forem/forem/issues/23876
Any info is good info! That's a perfect Issue and it's exactly the kind of analytical info that the team needs to figure it out. If you have any theories, you can also list them, but if you're genuinely confused at why it's happening, then just a notification is perfect.
Thanks π
I also opened another issue recently, and somebody fixed it!
github.com/forem/forem/issues/2388...
Yeah, I saw that a few months ago. It gets even weirder when you look at followers, read times, etc. ALOT of Dev.to traffic is bots. For what purpose? Probably data mining for AI training? So dont pay much mind to the stats, it's the same reason why you wake up 1 day and suddenly get 200+ new followers, but when you look at their profiles, most are just generic shells, no comments, no posts, not even a bio.
Update: It's quite different for this post:
Maybe the bots aren't crawling because it's about them? π
My weekly analytics. 100+ readers today, but average read time is 30s, it takes more than 30s to read, so chances are it's scale of 10, 10x longer read times average, but 10x fewer readers, the rest are bots.
What?? And they haven't found a solution to this?
Well it's a bit more complicated unfortunately, Forem has limited people who review and implement the PRs, so while I made a solution for them months ago, they just dont have the manpower to go through 5k+ lines of code to verify whether it's safe or not to implement. Unfortunately, it's not something that can just be 'hot-patched', because they currently dont expose enough to actually do a good job at it. Hence DevGuard uses browser identity to determine if it's a playwright browser, etc. Not the kinda stuff they'd have any real knowledge about auditing, even though it's essentially how DataDome does their WAF. Further than that, a few bots are preferred over losing a single real person, which is the other end of the stick. They've been trying a system that ended up flagging anyone posting links as bots, which they seem to have fixed now, but it's a mess to figure it all out and I dont blame them for taking their time, cuz it's not an easy task staying ahead of bots.
Yeah. That's also correct. It flagged me as a bot for including a link in a comment. I guess it no longer does that.
That's why it's so important to report bugs, it's a hassle... But it allows them to address it directly, because the more evidence you can give them (assumptions are great too!), they have more to work with to uncover what's breaking behind the scenes and how to fix it.
Actually 1 of the reasons why I made DevGuard, they havent implemented it yet, but it's essentially an anti-bot system (I had some experience bypassing WAF with MCP-Lite, so I used that as the basis of how to flag a bot), that counts for profiles, readers, comments, everything. Cuz it's traffic that they pay for, people think that hosting a server is all it takes, but you pay for network use too and the bots likely account for a serious amount of traffic...
So, that's probably the reason behind the many followers on my profile who haven't read a single one of my posts? I have also got a few purely promotional comments that have nothing to do with my post. Fortunately, they were automatically removed. But I still got the notification.
Yip. It is unfortunate, 99% of them woulda been filtered out if they used cloudflare proxying, but it is what it is. That's why readers and readtimes cant be trusted, so best option is run on engagement to judge (comments and likes). The bots have no reason to like or comment and the bot comments are (mostly) filtered out.
Yeah. It's quite good at filtering comments. But it has many false positives. But that's ok, it works.
You mean devguard.org? Or something else?
Nope, this: github.com/UnitBuilds-CC/DEV.to-De...
Oh. I always search on Google before naming my projects. So I'll get it when I search for it (e.g., ReptClip). Have you tried that?
Most of the time yeah, except for DevGuard, I let the AI pick the name, cuz it's just meant as an internal tool (Like Sloan is a bot).
I suspect ad blockers are to blame. It also depends on how many people actually viewed your article. In my case, I had 842 views from "other," over 2,000 from Google, and another 1,500 from some other source. I think that's normal.
WOW! First time seeing such high numbers! So we can see it in that much detail? I didn't know that so many people on DEV use ad blockers.
Maybe Google's crawler
Is it similar in your posts as well?
yes
For the first time the graph is flipped!
Some comments may only be visible to logged-in visitors. Sign in to view all comments.