For those using Google Analytics, there has been a considerable amount of talk in the marketing world over the last 12 months, on the value of reporting 'Visit' numbers over 'Unique Visitors.'
Many marketers will already have been doing so, and many readers will be aware of the case for changing metrics, but in case you are not, I'll start by doing my best to spell it out in simple terms here.
The first thing to note is what a unique visitor (UV) means to Google. Effectively, what does it mean to be unique? The simplest answer is that Google stores a cookie on a users PC for 2 years and any time they come back to your site, it recognises that you've been there before. If there isn't a valid cookie on the users machine, Google Analytics sees the user as unique and the number of unique visitors increases by 1.
Except we all know the dangers of cookies. Many users (myself included) clear their cache and cookies on a daily basis (or more) and some have browser plug-ins which clear them every single time the browser is closed. Also, using the likes of Google Chrome’s Incognito mode (which many more savvy users now default to) will not store this session cookie once the tab is closed.
With my cookie deletion habits, I can typically count as 10-15 UVs a week on the sites I regularly use, two or three times that number if you consider my mobile and tablet usage and my regular cookie deletions on those devices too.
Now — despite my best efforts to use my own DNA to clone an army of enhanced supermutants — there are sadly not 30 or 40 of me, but Google doesn't know that if I clear out my cookies. If I appear to be 30 unique people, that's how I will be recorded.
You could just ignore this fact completely, and some marketers and SEOs — often those with too much political investment in their UV statistics — have taken exactly that approach, as it works to their advantage. For some it's a case of "who cares if the data is right, at least I can infer a trend" or worse, “at least it looks good to the stakeholders!” But many more objective digital marketers who were previously reporting UVs, have now switched to reporting visit numbers.
Hats off to them, right? Well done for not being too set in their ways to embrace change.
“Surely if Cormac visits my site 30 times, he will clock up 30 visits?”
“If I can't reliably say whether Cormac is one person or 30, I can at least reliably say he made 30 visits, right?”
Sadly, like so many things in life, it's not nearly as simple as it seems on paper.
It is true that under the right circumstances visits will report reliably, just as unique visitors would under their own preferred scenarios, but there are dangers around recording and reporting visits too.
The caveats are a bit more 'quirky' than those which affect visitor numbers, this sadly makes them less intuitive and ultimately less well understood. Once again, I'll do my best to explain the issue in simple terms.
As with UVs, the key to understanding how visits are recorded, is understanding exactly what a visit means in the eyes of Google.
In Google's own words "each visit to your site from a different campaign — organic or paid — triggers a new session."
More specifically, when measuring organic traffic visits (those which comes through a regular search), there needs to be a change in either the medium, source or search term for Google to log the visit.
For those not so familiar with this terminology, traffic sources are the places your traffic comes from (Facebook, Bing search, a Google Ad, etc), medium is a way of grouping similar sources (Bing search and Google search are grouped as "organic," links from Twitter, Facebook and other sites are grouped as "referral," etc) and the search term is what a user typed into the search engine to find your site.
Based on that description, you may have already jumped ahead and realised that if you do a Google search for "Cormac Scanlan" and find this website, and then you leave and do another Google search for "Cormac Scanlan" and again click through to this site, despite having made 2 clearly separate visits, you will only have logged as 1 visit in my Google Analytics account. This is because on both occasions your source is "google," your medium is "organic," and your keyword term is "Cormac Scanlan." Because there is no change in any of the 3 key parameters related to organic search, Google has not logged your 2nd visit as new.
At this point, you may be thinking that's a little annoying, but that it's a bit of an edge case and probably won't account for much of a skew in the data. From what I've said so far you would be quite right, but there are scenarios which make the design of Google's tracking of visits much more problematic.
Things get marginally more complicated from here on in, but stay with me as it's well worth understanding why visit numbers are less reliable than most people realise, and why the problem is actually getting worse.
At this point things require a brief segue to explain an issue which has been bugging SEOs and marketers since 2011.
In 2011 Google introduced changes which began to limit the keyword terms returned from secure searches. This meant that anyone using secure services (such as encrypted.google.com) started appearing under the keyword "(not provided)" in Google Analytics data.
Shortly after, Google defaulted to secure search for all users logged in to their services.
If (for example) you log in to Gmail and run a Google search you will see the URL starts https (which is designated to secure search, as opposed to http for non-secure). The same is true if you search using the Google Chrome’s address bar, even if you are logged out.
In simple terms, this now means that regardless of what the user might choose to type into Google as a search term to find your site — be it “perms,” “worms,” “sperms,” or “other search terms” — when they click through to your site from one of the many forms of secure search, Google Analytics will record the keyword data around the visit as “(not provided).”
When this first took effect, webmasters, online digital marketers and SEO specialists started freaking out over losing 10-15% of their keyword data in Google Analytics; this pretty much happened overnight.
In retrospect, many of those would now look back to those days and gladly accept a 15% dip in their provided organic keyword data as the current figures for “(not provided)” keywords seen by most websites are typically 40-45%, and on a few large sites I work on, the amount of “(not provided)” data has now tipped over the 50% mark, representing an overall majority. In other words, these sites are getting more Google organic traffic coming through as “(not provided)” than every other traffic source put together.
Returning to our number of visits metric, if you are now wondering how all this is connected, take a step back and look at the scenario we went through before using organic searches for my name.
In terms of the way the stored data is handled by Google Analytics, there is absolutely no difference between the keywords "Cormac Scanlan" and "(not provided)." This means that if you use a secure service to find my website multiple times through organic search, even if you use different keywords each time, you will only log as 1 visit in my Google Analytics data.
Let me give a simple example which illustrates the issue more clearly:
If I use standard (non-secure) Google and I search for the term "dictionary apples" and click onto the dictionary.reference.com page about apples, Google will log my visit as 'google / organic / dictionary apples.' If I then close dictionary.reference.com and search for "dictionary bananas" and click back to dictionary.reference.com, Google will log my 2nd visit as 'google / organic / dictionary bananas.'
If I then use secure search and do exactly the same thing, when I Google "dictionary apples" and click onto the dictionary.reference.com page, Google will log my visit as 'google / organic / (not provided).' Similarly, if I then close the site and repeat my search for "dictionary bananas" using secure search, and again click onto dictionary.reference.com, Google will again log my visit as 'google / organic / (not provided).'
Because the keyword is not provided in secure search, the search term becomes the same for both visits. As a result of this, Google does not recognise the 2nd visit as new, because it can see no difference between the first 'campaign' (to use Google's terminology) and the second.
In short, no matter how many separate visits a user makes via organic search, if they are part of a series of searches done (within a 30 minute window) using secure search, they only count as a single, solitary visit.
Michael M. Holt has got some very interesting data around this and published a great article on (not provided) which I would strongly recommend reading. It also goes into the technical side of how to examine the Google Analytics cookies, replicate his tests, and show conclusively that this is the case, details I have left out of this article to keep it easy to follow.
For sites dealing with 40, 50 or even 60% of visits appearing as not provided, it is fair to say that (not provided) poses a big problem.
The reality is that the Google Analytics method of tracking of visits in this way was set up before the change to stop providing keyword data from secure searches was implemented. Simply put, when the system of logging visits was put in place, it was not designed to cope with scenarios where the bulk of the organic keyword data logged under a single (not provided) keyword.
The sad reality is that in Google Analytics, neither Unique Visitors nor Visits are reliable metrics these days, particularly if you are running a site with a significant percentage of organic traffic.
It is worth noting that businesses who get a lot of traffic through PPC and AdWords campaigns fair much better as keyword data is provided from paid campaigns — even when secure search is used — meaning that each visit is usually reported and tracked correctly.
Quite what the implications are to your business or website ultimately depends on how much of your traffic is organic and how much importance is placed on getting accurate data around UVs or visitors.
If you have a site where your traffic is almost entirely paid, the effects of these quirks are fortunately negligible.
Right now I don’t attest to have any answers (and I really would love to hear from anyone who does - in the comments section below), but I would say that if you work in a business which relies on reporting these metrics and you are only using Google Analytics to do so, this poses a problem, particularly if you sell advertising based on either figure, or if your staff are incentivised on organic traffic numbers.
The difficulty is really based on the fact that very little can be inferred from a lack of data.
When it comes to reporting UVs, it is almost impossible to take a guess at how many people clear out their cookies (and how often), or how many find your site using services like Firefox Private Browsing or Chrome Incognito. Moreover, taking a guess at these things flies completely in the face of data-driven business management and is generally a bad idea. Similarly, there is no way to make any sort of fair guess as to how many visits you lose through the keyword “(not provided)” quirk.
In some cases, trends in visitor data can still be drawn over short periods of days or weeks, but it is important to realise that this is set against a steep upward trend of “(not provided)” keyword data, which has doubled year on year for many businesses.
As for making effective year-on-year comparisons, the problems are only exacerbated when you start trying to factor in the rise of Apple and Android devices and set it against variables such as iOS6 devices defaulting to secure search and showing as not provided, but iOS5 not.
If the trends shown at notprovidedcount.com are to be followed and believed, by the summer of 2015 we’ll be looking at a scenario where all keyword data for organic searches is not provided.
This obviously won’t happen (unless of course Google move everyone to secure search), but the reality is that the "(not provided)" keyword has been on a steady upward rise for two years, and shows little sign of slowing down right now. In short, this is a problem that seems only likely to get worse over time.
No reporting system is 100% accurate, and in time I suspect it is inevitable that Google will have to reassess their definition of a ‘visit’ and how it is tracked and recorded. When they do, the problems above will hopefully subside.
For now, my key takeaways would be to simply understand the falseness of the statement “Visits are reliable, but Unique Visitors are not,” and try to avoid basing major business decisions around either metric, without a proper understanding of the caveats.
comments powered by Disqus