Analytics
2.2K members online now
Understand information in your reports and troubleshoot reporting issues such as self-referrals, (not set) data, and inaccurate information
 
Guide Me
star_border
Reply

Users - Google Analytics vs Cloudflare discrepancy

[ Edited ]
Visitor ✭ ✭ ✭
# 1
Visitor ✭ ✭ ✭

Hello,

I noticed the huge difference between data on users/unique users between Google Analytics and what is reported in my Cloudflare Analytics. I know that users are counted differently and with different tracking techniques by the two. GA counts users through a javascript and can filter out bots and crawlers, while CF counts them (almost) all and measures users through server access tracking. Yet, this should lead to a difference of 30, 40, even 100% logically, while in my case the difference is 1000%. GA gives me a users number (formerly unique users) of about 25,000 per months on average, while GF reports over 250,000 of them. All bots and crawlers? How can it be?
Can someone provide me a better thought on this point?

Any help is very appreciated!

1 ACCEPTED SOLUTION

Accepted Solutions
Marked as Best Answer.
Solution
Accepted by topic author Riccardo B
February 2016

Re: Users - Google Analytics vs Cloudflare discrepancy

[ Edited ]
Explorer ✭ ☆ ☆
# 7
Explorer ✭ ☆ ☆

Well, 'reliable' is a relative term. With web data, this is especially true due the nature of web tracking. At it's most basic level, web data is only available because, in general, we are 'allowing' ourselves to be recorded. Every tiny little action we take on the web is recorded by an entity that we have no relation to. The only reason we, as a population of free individuals, allows this recording to occur is the notion that we are in fact 'allowing' it, that we can turn this off when we want, that all recording preserves our anonymity, and that tracking data is silo-ed within each domain. Without the maintenance of control and anonymity it wouldn't be tracking, it would be surveillance and that's just not something the free world would allow...especially by corporations.


Because all web tracking applications are forced to play on this same field, the are all equally reliable within an acceptable range.GA, Omniture, whatever. All the same really. What gives the impression of unreliability is a few factors:

  1. Slightly different definitions of similarly named outputs
  2. Different methods of capturing and storing data
  3. Almost complete flexibility in implementation
  4. User error in producing outputs


That said, you should also consider who holds this perception of 'unreliability'. Before you started delving into this field you probably assumed that web data was inherently precise. Now you know it's not and you're worried about providing inaccurate figures to clients. In actuality you're giving them the most accurate view that users of their site are allowing them to have. They should be thrilled, and they will be as long as you don't put any asteriks all over the data. Web data isn't totally precise because it can't be, but it's massive enough to provide a very, very representative sample, which is the key. The sheer volume provides us with a high level of confidence that what we see is representative of the whole. As for reliability, GA is highly reliable relative to itself. As are all of the other web tracking platforms. To get good at web analytics you need to follow few guidelines:

  • Pick a platform, learn it, and stick with it
  • Accept the inherent lack of precision
  • Maintain confidence in your results, they're the best you're gonna get and they are a highly accurate representation of the total population
  • Never, ever, ever present results to clients with any type of definitions regarding web analytics. A user is a user. A session is a session. End of story. It's data they can work with and it will ebb and flow relative to itself. That's all they need.If they do ask about definitions, be educated enough about it so that when you're done explaining it to them, you've given them confidence in the numbers. If they find the numbers unreliable, that feeling is going to transfer directly to you.


As for your cloudfare 'gap', your problem is that Cloudfare isn't operating in the previously explained, client-controlled web tracking environment. All web data is generated by the tracking pixels when they are 'read' by client machines. Cloudfare data is generated on the server side making it inherently different. Your comparing apples and dandelions. Sure they're both in the plant kingdom, but that's about it.


In short, if that's possible now, let cloudfare do what it does and let GA do what it does.

View solution in original post

Re: Users - Google Analytics vs Cloudflare discrepancy

A T
Participant ✭ ✭ ✭
# 2
A T
Participant ✭ ✭ ✭
It very well can be.

Try to see if any filters are applied for GA views.

Re: Users - Google Analytics vs Cloudflare discrepancy

Visitor ✭ ✭ ✭
# 3
Visitor ✭ ✭ ✭

This can hardly be the problem, I have no filters currently applied in GA, even the bots and spiders filter is not enabled, while I guess that at least some bots/crawlers are excluded in CF due to the firewall settings. Anyway, considering that we are talking about unique users and not pageviews, it sounds strange to me that my website is accessed by some 225,000 bots' and spiders' unique IPs a month. Maybe I don't fully understand the correlation between users and unique IPs.

Re: Users - Google Analytics vs Cloudflare discrepancy

Explorer ✭ ☆ ☆
# 4
Explorer ✭ ☆ ☆
I can't answer this specifically, but I can tell you that this exercise isn't going to benefit you in anyway. I've spent copious amounts of time in my career begrudgingly trying to reconcile variances between two systems. In every single instance, understanding the difference led to the same thing: Pick one and stop using the other. In every single instance, both systems were 'correct' based on it's definition of the metric.

At the heart of the issue is the fact that they calculate differently and, in your case, they probably aren't really even the same thing. Your going to spend a lot of time and energy trying to understand something that isn't going given any actionable information.

My advice, stick with GA (or Omniture or any other web tracking product). It's designed specifically to give the best idea of what is happening on your site. That's the entirety of it's purpose. Cloudflare, on the other hand, is primarily concerned with site performance and security. Analytics is a peripheral component of what they do so you'll probably never get the depth of understanding you need. So why drive yourself crazy trying to figure it out?

That said, you do have an extremely wide variance there. The first question you should ask yourself is which one of those figures 'feels' more right to you. If it's the GA number, then just move on with your day. If it's the cloudflare number, then you need to either dig deeper into your GA code implementation and/or reporting or you should stop using GA until you've committed to validating it's implemented properly.

That's my two cents anyway.

Re: Users - Google Analytics vs Cloudflare discrepancy

[ Edited ]
Visitor ✭ ✭ ✭
# 5
Visitor ✭ ✭ ✭
Hi VariantArray,
 
Thank you so much for your support!
I think yours is a good point.
The reason why I am digging into this matter is to provide reliable traffic data to my advertisers and subscribers. I have always used data from GA for such a purpose.
I want to provide them REAL information indeed, I would much prefer to give my clients lower, yet reliable, data, than to claim inflated and unrealistic numbers (like, sadly, many big names in the IT business do).
Yet, I can't simply dismiss a 1 to 10 ratio in unique users, I need to know which among the two is reliable and which is not.
I am pretty sure that data from GA are correct, and I guess I will stick with them unless I can prove them wrong, perhaps because I badly implemented the GA code.
Anyway, since I can access my website(s) from different networks and IPs at the same time, I can do a little testing on that and verify if all accesses (which I know are real, since it's me to do them) are registered by GA and correctly counted as such.
Though, it will still be a puzzle for me to guess where those incredible amount of users reported by CF could come from....Smiley Wink

 

Re: Users - Google Analytics vs Cloudflare discrepancy

[ Edited ]
Visitor ✭ ✭ ✭
# 6
Visitor ✭ ✭ ✭

I guess I should add another element to this discussion.
I cast a glimpse on additional third-party information, such as Facebook Insight statistics, and they confirm the GA data consistency. Indeed, when I compare FB clicks on link posts to the GA traffic from FB numbers, everything matches.
I will test my GA implementation, as I said earlier; yet, I am pretty sure that, when taking unique visitors into account, the GA numbers are much more reliable than the CF ones, at least when you rely on the most accepted definition of what a unique visitor is.
That said, and giving that my opinion is correct, I am just wondering how reliable a market - where a website's unique visitors number could range from 25,000 to 250,000, depending on what data you dare to present to your clients - could be.

Marked as Best Answer.
Solution
Accepted by topic author Riccardo B
February 2016

Re: Users - Google Analytics vs Cloudflare discrepancy

[ Edited ]
Explorer ✭ ☆ ☆
# 7
Explorer ✭ ☆ ☆

Well, 'reliable' is a relative term. With web data, this is especially true due the nature of web tracking. At it's most basic level, web data is only available because, in general, we are 'allowing' ourselves to be recorded. Every tiny little action we take on the web is recorded by an entity that we have no relation to. The only reason we, as a population of free individuals, allows this recording to occur is the notion that we are in fact 'allowing' it, that we can turn this off when we want, that all recording preserves our anonymity, and that tracking data is silo-ed within each domain. Without the maintenance of control and anonymity it wouldn't be tracking, it would be surveillance and that's just not something the free world would allow...especially by corporations.


Because all web tracking applications are forced to play on this same field, the are all equally reliable within an acceptable range.GA, Omniture, whatever. All the same really. What gives the impression of unreliability is a few factors:

  1. Slightly different definitions of similarly named outputs
  2. Different methods of capturing and storing data
  3. Almost complete flexibility in implementation
  4. User error in producing outputs


That said, you should also consider who holds this perception of 'unreliability'. Before you started delving into this field you probably assumed that web data was inherently precise. Now you know it's not and you're worried about providing inaccurate figures to clients. In actuality you're giving them the most accurate view that users of their site are allowing them to have. They should be thrilled, and they will be as long as you don't put any asteriks all over the data. Web data isn't totally precise because it can't be, but it's massive enough to provide a very, very representative sample, which is the key. The sheer volume provides us with a high level of confidence that what we see is representative of the whole. As for reliability, GA is highly reliable relative to itself. As are all of the other web tracking platforms. To get good at web analytics you need to follow few guidelines:

  • Pick a platform, learn it, and stick with it
  • Accept the inherent lack of precision
  • Maintain confidence in your results, they're the best you're gonna get and they are a highly accurate representation of the total population
  • Never, ever, ever present results to clients with any type of definitions regarding web analytics. A user is a user. A session is a session. End of story. It's data they can work with and it will ebb and flow relative to itself. That's all they need.If they do ask about definitions, be educated enough about it so that when you're done explaining it to them, you've given them confidence in the numbers. If they find the numbers unreliable, that feeling is going to transfer directly to you.


As for your cloudfare 'gap', your problem is that Cloudfare isn't operating in the previously explained, client-controlled web tracking environment. All web data is generated by the tracking pixels when they are 'read' by client machines. Cloudfare data is generated on the server side making it inherently different. Your comparing apples and dandelions. Sure they're both in the plant kingdom, but that's about it.


In short, if that's possible now, let cloudfare do what it does and let GA do what it does.

Re: Users - Google Analytics vs Cloudflare discrepancy

Visitor ✭ ✭ ✭
# 8
Visitor ✭ ✭ ✭
Dear VariantArray,
I did some little investigation on all this matter and, though I am absolutely not an expert in this particular field of IT and my guess could be largely defective, nevertheless, I think I now fully understand what you are suggesting me. GA, CF, Omniture... they are "reliable", depending on what they were intended for, knowing that ""A rose by any other name would smell as sweet", and that I will always have a limited knowledge on how such statistics are actually built. Nevertheless, I guess I have now a reasonable idea on where the discrepancy comes from. It is something involving, more than bots, crawlers, or prefetch techniques in themselves, the way they are taken into consideration and gauged; In my case, the large amount of images in my website probably plays a considerable role on that 1 to 10 ratio.
Now, better I stop guessing and, as you said, pick a platform and do not puzzle myself too much any more...Smiley Wink

All the Best,
Riccardo