Upcoming LunaMetrics Seminars
New York City, Aug 4-8 San Francisco, Aug 11-15 Los Angeles - Anaheim, Sep 8-12 Washington DC, Sep 22-26

“Absolute Unique” and “Unique” Visitors in Google Analytics





Recently several people asked me about confusing numbers involving unique visitors (one was a customer and the other used the great little blog idea submission form at the right).

Visitors represent the number of distinct people who come to your site (rather than Visits, which is the number of times that they come). Seems pretty straightforward, right? Well…

One person wrote:

I’ve been puzzled about the difference between “Absolute Unique Visitors” and “Unique Visitors”. In some cases, # of Unique Visitors is double of Absolute Unique Visitors! How it could be? I have no idea how Unique Visitors is calculated and its definition.

And the other asked:

How much variability is there in google analytics?  When I run a custom report to get unique visitors over a month it can range from 195k to 214k.  Sometimes I get a daily count of 0 and other times I get a count of 261 for the same day.

There are potentially several issues going on here.

Two kinds of “unique” visitors

First of all, as the first question points out, there are actually two metrics for unique visitors in Google Analytics. There’s Absolute Unique Visitors and Unique Visitors.

Absolute Unique Visitors is the metric you get in the report of the same name in the Visitors section. It used to be the only place you could get visitor metrics in GA. It’s also fairly limited: notice that you can’t use Advanced Segments on this report (one of the few where you can’t).

Unique Visitors is another metric that was introduced for use in Custom Reports (which you can use Advanced Segments on as well). It was an update to the Absolute Unique Visitors that was more compatible with these new features, basically.

The metrics themselves actually differ slightly in the way they measure visitors (how they tabulate the individual cookies). So the numbers won’t match up exactly. In fact, Unique Visitors is actually more accurate than Absolute Unique Visitors (surely a nominee for the Worst Title Ever in Google Analytics). Unique Visitors will usually show a slightly greater number (on the order of 5-20%, based on a few sample accounts I looked at). Since Unique Visitors is the more accurate and flexible metric, Absolute Unique Visitors is being phased out over time and someday we will have the One True Visitor Metric.

Uniques depend on the time period

One really important thing to recognize is that, “unique” depends on what time period we are referring to.

Take this Absolute Unique Visitors report.

For the 30 days that I am looking at here, there were 12,926 Absolute Unique Visitors (listed at the top there). But if I went through and added up the numbers for each of those 30 days (in the bar chart), it wouldn’t equal 12,926. Why?

Well, suppose I visit the website on Monday. Maybe I visit once or twice or 67 times, it doesn’t matter. I am 1 unique visitor for Monday, and the bar for Monday goes up by 1.

Now I visit again on Tuesday (once or twice or 67 times again). Tuesday goes up by 1.

BUT, for the whole month, I am still just 1 unique visitor out of those 12,926. It doesn’t matter whether I only came once on one particular day during that period, or once every day, or 67 times every day.

So it’s really important that you recognize this when you talk about unique visitors. Are these daily unique visitors? Monthly? It’s really important what the time frame is. You can’t, for example, take the number of monthly unique visitors for January, February, …, December and just add them up to get the annual unique visitors.

The dreaded sampling

The data in the standard Google Analytics reports is all pre-computed, and it’s pretty exact. If a visit got recorded, you’re going to see it there.

However, when you start combining things that weren’t pre-computed — for example, with an Advanced Segment or Custom Report — Google Analytics does the computations on the fly. If you have a really busy site, this would potentially take a looooooong time.

So instead, if you have more than 500,000 visits in the time period you are looking at and you request data that needs to be computed on the fly, Google Analytics takes a statistical sample of the data and gives you an estimate with a range of error, rather than tabulating every single visit. You’ll get a little notice that says “this report is based on sampled data” and a +/- error margin along with your metrics. (There’s more info about how this works in the Google Analytics documentation.)

This is something you should also be aware of when you’re looking at reports, because it can drastically affect accuracy, especially for segments or dimensions where the total number of visits is small compared to the total visits. If you need exact numbers, try to keep to time periods with fewer than 500,000 visits.

Jonathan Weber

About Jonathan Weber

Jonathan Weber is the Data Evangelist at LunaMetrics. He spreads the principles of analytics through our training seminars all over the East coast. The next seminar he'll be leading will be a Google Analytics training in Boston. Before he caught the analytics bug, he worked in information architecture. He holds a Master’s degree from the University of Pittsburgh School of Information Sciences. Jonathan’s breadth of knowledge – from statistics to analysis to library science – is somewhat overwhelming.

http://www.lunametrics.com/blog/2010/07/30/absolute-unique-unique-visitors-google-analytics/

37 Responses to ““Absolute Unique” and “Unique” Visitors in Google Analytics”

VaBeachKevin says:

In my opinion a Visitor can only be one of two things, New or Returning. Why do all the WA vendors try to confuse everything? WA is already a challenge, why make it harder?

George says:

So, I gather that absolute unique visitors is useless as a metric.

Arek says:

What is the difference between “Visitors” and “Unique visitors” in custom reports?

My understanding was the GA said “Absolute Unique Visitors” because it could give that number for even custom time period (say from April 12 to 24) whereas other vendors do not have the computational power to do it, and will give correct uniques for only the regular time periods of day, week, month, quarter and year.

Jonathan Weber Jonathan says:

George — I prefer Unique Visitors where available, yes. And as I said, the plan is to eventually replace Absolute Unique Visitors with Visitors in the future in all places.

Arek — “Visitors” in Custom Reports is Absolute Unique Visitors; “Unique Visitors” is Unique Visitors. Yet another labeling flub.

Jacques — You are correct (about the original naming purpose for Absolute Unique Visitors). That still remains true about Unique Visitors, however. It’s a distinction of Google Analytics, even though it’s lost in the current labeling confusion.

Jo says:

Good stuff.
There are two humongous WA thought leaders out there who are arguing (have been) that Uniques is not as valuable as most have been led to believe.
What do you think?

Eric says:

Really interesting article guys – thanks! However, I’m seeing some strange things in my reports that have me asking two questions:

1 – I created a Custom report with just Unique Visitors and Visitors as metrics (dimension is Day). They data is exactly the same at the Day level (and matches up with Absolute Unique Visitors report too), however the totals above the table show there is a difference in the way they are calculated in aggregate. Neither, however matches with the total for the Absolute Unique Visitors report. I’ve even focused it down to just 2 days worth of data, and even then the data doesn’t match. Any thoughts?

2 – I also have this data going to two separate profiles. One has literally zero filters, the other filters out my company IP address – this way I’m not affecting the data. However, the Visitor count (pick any of the three metrics – Abs Unique Visitors, Visitors, Unique Visitors) for the Filtered Out Profile is actually HIGHER than the completely Unfiltered Profile! I know, it sounds crazy to me too, but it’s happening. Any thoughts as to why or how this could be happening?

Thanks! Keep up the great work!!

Craig Eric says:

From what I understand, GA identifies visitors through first-party cookies stored on the visitor’s hard drive.

I don’t think it was mentioned above, but the number of unique visitors is dependent on whether or not the visitors frequently delete their cookies.

If a visitor deletes their cookies daily- over the course of a month, the number of unique visitors may be skewed upwards.

Just a thought.

Jonathan Weber Jonathan says:

Eric — not sure about either of your questions. For #1 — for some metrics, Custom Reports don’t always tally in the expected way (totals instead of averages, etc.) so you may be facing an effect like that here with certain combinations of dimensions with the visitors metrics. Unfortunately, I don’t have any guidance on what those are because this isn’t called out in the official documentation.

For #2, that just seems bizarre! I spot-checked a few accounts where I have the same thing (an unfiltered profile, plus one filtered for internal traffic) and I didn’t see anything like this.

phar says:

a great article to help me understand better.

I am interested in the sentence below.

“The metrics themselves actually differ slightly in the way they measure visitors (how they tabulate the individual cookies). “

Are there more specific descriptions about the way to tabulate individual cookies?

Jonathan Weber Jonathan says:

Phar — The technical difference works like this:

Google Analytics stores an identifier in each visitor’s __utma cookie that identifies them as unique. It’s created from a hash value, which means that different visitors can occasionally get the same value — this is known as a “hash collision”, see http://en.wikipedia.org/wiki/Collision_(computer_science).

Absolute Unique Visitors just uses the hash values. Unique Visitors uses the hash value plus the user agent (browser identifier) to help correct for hash collisions. As a result, UV should be greater than or equal to AUV. I don’t usually see much difference, on the order of a couple percent, but UV is more accurate because of the hash collision correction.

phar says:

Thanks Jonathan for quick answer.

I wondered how frequently “hash collision” occurs in the GA.
So, I checked the ga.js and it seems it uses “random() & a hash function”, and it also seems the key(seed) of the hash function uses user-agent and other system parameters(language,width,,) as the argument.

I wonder two things:

1. Is it still OK, this is a hash value?

2. If the id was created by “random & hashed” numbers, the collision should be happened much less than 1%.
Because the hashed value id(in the __utma) is 10 digits, which has 10billion slots.

Mark says:

Jonathan,

my Unique Visitors total in my custom report is more than twice as much as in the default Visitors report.

Way more than your 5-20% error margin.

How can there be such a huge difference?

Jonathan Weber Jonathan says:

Phar — even though there are 10 billion slots, the distribution isn’t truly random; it depends on how the hash function assigns the random parameter to the space of 10 billion slots. I haven’t looked at the details of the hashing function GA uses — and I probably wouldn’t understand them even if I did! I am no algorithms expert. :)

Mark — I’m not really sure. One thing to make sure of is that you are taking into account the time period effect mentioned in the blog post — you always have to think about the time period you are referring to for calculating uniques, and you can’t simply add up smaller time period to total a longer one. If that’s not your problem, however, it’s hard for me to say exactly what is going on.

Well there is much difference in unique and absolute unique visitors.

David says:

What does the % stands for in this graph? (at the end of each bar)

That is obvious percentage .

@Mark
This much difference is expected there.
There are many visitors which once come to your site,
But after that they never come i mean new visitor always that is.

Lukas says:

@Arek/Jonathan: I would disagree with Jonathan on “Visitors” in Custom Reports is “Absolute Unique Visitors”. I just compared the following numbers for the same time period:

26,634 Absolute Unique Visitors

26,835 Unique Visitors

23,927 Visitors

My assumption thus is that “Visitors” is what you usually call “Daily Unique Visitors” and not “Absolute Unique Visitors”.

That especially makes sense because when I narrow the timeframe to one day, “Visitors” and “Unique Visitors” are exactly the same, but when widening the timeframe, “Visitors” are always lower than the other ones.

Jonathan Weber Jonathan says:

Lukas — You’re totally correct that the totals for “Visitors” in Custom Reports don’t match “Absolute Unique Visitors”. I was referring to the fact that Visitors corresponds to the “old” definition of visitor (without the hash-collision correction described above).

As for what the total for Visitors means… The idea that the totals for Visitors in Custom Reports is based on total of daily visitors is a good one, but I don’t think the numbers bear it out.

Suppose I have Visitor A and Visitor B visit the site Monday. 2 daily unique visitors.

Visitor A and Visitor C visit the site Tuesday. 2 daily unique visitors.

So, 3 Absolute Unique Visitors. Totaling over the daily unique visitors, 4.

Notice that no matter how users behave, a total of daily unique visitors must be greater than or equal to Absolute Unique Visitors. Never less.

So it’s still not clear what exactly “Visitors” does in Custom Reports. Frankly, I don’t even know why it’s still available — it’s confusing to users, it can’t be combined with many other dimensions and metrics, and moreover it’s not really clear how the total is calculated as we just saw. “Unique Visitors” is a much better choice for use in Custom Reports, because it’s specifically designed to work there. (And, it’s the metric available in the API as ga:visitors.)

Ahmet says:

I think that Google now uses absolute unique visitors and unique visitors as if they are the same thing. At least that’s what I understand from this: http://www.google.com/support/analytics/bin/answer.py?hl=en&answer=33087

Is that correct?

Ulrik says:

Here’s an easy explanation. When someone visit your site any analogs tool will put that visit in data “buckets”. If you visit a website on the first of January 2012, you will be a visitor in the daily bucket of Jan 1st, the weekly bucket of week 1, the monthly bucket of January, the quarterly bucket of Q1 and the yearly bucket of 2012.

An analytics tool has to store visitors this way, otherwise they would have to create a bucket for any combination of dates. You can imagine how much data that would be…

So, imagine you looked at the visitor count for the date range 1. Jan to 10. Jan.
In order to show you a total visitor count the tool must find the largest correct buckets and add them together. So if week 1 is from Jan 1st to Jan 7th, the tool would take the visitor count from the weekly bucket of week 1 and add the visitor number of the daily buckets of the 8th, 9th and 10th of January.

Thus when looking at daily visitors and weekly visitors, for a given report the number would vary greatly.

Marc says:

Thanks for the explanation, but I’m very disappointed to hear that Absolute Unique Visitors will be going away. In fact it looks like it was removed from the “new” GA within the past few weeks, because I know I was looking at the report there.

Absolute Unique Visitors is EXACTLY the type of report I am looking for, and I find it much more useful than “Unique Visitors”. For our customer extranet, I want to know how many different people visited the site during a given time period — maybe a month, a quarter, or a year. If you visit the site every day for a month, I want you to be counted as 1 unique visitor, not 30. With Absolute Unique Visitors, I can easily answer the question “how many people (customers) visited our site at least once this year?”

Also, you haven’t really explained how “Unique Visitors” is calculated. Is it unique visitors on a single day? If so, then Unique and Absolute Unique on a single day should match, and they don’t. (My report shows 533 unique and 504 absolute unique on a randomly-selected day.) So it must be a shorter window than that. Perhaps 6 hours? Without knowing the time frame, the number is useless.

Marc says:

OK, I reread the post and subsequent comments, and I think I misunderstood. Are you saying Unique and Absolute Unique are essentially the same thing, just calculated differently (due to the hash-collision issue)? As long as Unique is still based on a user-selected date range, I’m happy.

@Lukas/Jonathan, it does look like Visitors in custom reports is the same or very, very close to Absolute Unique Visitors. I compared the custom report’s “Visitors” for a single day to “Absolute Unique Visitors” for the same day and they matched exactly. But looking at the 2nd quarter, the numbers vary slightly (6447 “visitors” vs. 6576 “absolute unique visitors”).

Too many options with similar names, doing essentially the same thing but calculated in different ways.

Couponsophy says:

Interesting article. I never realized the difference between “absolute unique visitors” and “unique visitors” was so important. Makes me wonder why I trust GA so much when I look at reports…

i didn’t get you……
now also i m confused cn you explain it with an example…

Move to NZ says:

According to me both are different…

Brendan says:

The most important aspect of this issue has been totally overlooked. We are not actually talking about visitors or people here – what is being measured are browsers. One person can access a website via many browsers.

DerekM says:

I don’t see where you’re explaining the difference between UV and Absolute UV (AUV). You state that UV is more accurate than AUV, but why?

DerekM says:

OK, I read the clarification in the comments–you should update the actual post–but the entire post is just confusingly written.

Len says:

Hi, I have found this post very informative but we have said good buy to Absolute Unique Visitors (I think) but I am still having problems with the Unique Visitor Metric. It does not play well with other metrics, especially visits and it segments poorly. Does anyone else have problems with Unique visitors?

That was a very healthy discussion,with some good and effective points,I really appreciate this thread,thanks for all the information.

technophobe says:

Are you saying then that if you wanted to find out how many unique visitors you had in a period of 7 days you would have more people if you looked every day than looking over the whole week? I AM SO CONFUSED and have set up a client to only pay once their advert has been shown to 10000 unique hits, which until now and looking at my my analytics (fiddling around with the day week and month view of unique visitors) I thought would be pretty straight forward. GAHHH. If I look at my unique visitors I have way more in one month than if I add the 4 weeks, give or take a day in that month.

Is it correct to assume that the number of unique visitors gets less and less accurate as you increase the time, because of cleared cookies? So looking at uniques over a year would likely over count the number?

Vikas Sharma says:

By reading this post, i got more confused about unique visitor? How will be calculate visitor if I remove cookies and cache file each time after visiting website? Am i unique or regular visitor?

Marc says:

If someone visits a site from home PC, from work PC and from smartphone, will he show up in the statistics as 3 ‘unique visitors’? (in other words, isn’t is actually ‘unique IP adresses’?)

If so, are the number than not way off? What is the average ratio that should be taken into account for a blog with many returning visitors for example? Perhaps you need to divide the uniqie visitors by 2?

Abi says:

Thanks for that information. I was really confused about unique visitors and now I am very clear bout it.