Standard Metrics Revisited: #1: Visitors

DSC02875 small1Sometimes it is well worth stepping back from the bleeding edge and taking a fresh look at something very familiar, the non-bleeding edge. In this new series I hope to revisit some extremely well established and accepted metrics with the goal of providing fresh insights.

The first one is the metric that is the bedrock of all web analytics: “Visitors”. Almost every metric and report in any web analytics tools either has this metric or is sourced from this metric. It shows up as a raw count or in the shape of percentages or in the numerator or denominator or when we torture it by doing global averages and on and on.

Yet it turns out that there is no standardization and often this metric (count of visitors) masquerades under different names. There is often a misunderstanding as to what it exactly is and how it is measured.

For example the real inspiration for this post is the new tool I am trying out StatCounter. The Summary that greets you in StatCounter shows a graph of “Unique Visitors”. When I probed the support staff at StatCounter as to what this metric was here is how it went: SC: It is based on cookies -> AK: how do you define it? -> SC: It is count of sessions as identified by cookies -> AK: But that is not unique visitors. : )

google analytics visits1Most commonly prevalent names for visitor metrics are: Visitors, Visits, Total Visitors, Unique Visitors, Sessions, Cookies. (Please add the variations you have heard via comments below.)

It is a disservice to the world that so many names exist for the same metric. When we mention visits and unique visitors depending on the tool it is either the same thing or not, it confuses decision makers and sometimes it means that if we rip out Google Analytics and replace it with CoreMetrics we have to unlearn old definitions for the same thing. Quite sub optimal.

Here is a suggested standard, let’s eliminate all complexity and settle on two visitors metrics that anyone can understand on the nine (make that eight : )) planets in our solar system: Total Visitors & Unique Visitors.

Assumptions:

  • assumption2If your web analytics platform is still using web logs and your website platform does not use cookies this does not apply to you, specially the Unique Visitor bit. (Though for many reasons you might want to switch to a tag based solution and now with cheap and free ones out there you could consider adding a tag based standard solution.)
  • You are using javascript tag based solutions and you set both persistent (to identify a unique browser, as in IE or firefox etc, with an anonymous id) and transient cookies (to identify as session and its browsing behavior).
  • It is worth diving deep into understanding for your web platform what action starts and terminates the session. For example the start is always the same, someone comes to the site, but session end is not standard. ClickTracks will end your session if you leave the website for a search engine, even if you come back to the same website ten seconds later. Other tools would count it as the same session because you came back in less than 29 minutes.

Metric Definitions: (Updated, thanks Steven)

Total Visitors: Count of all the Sessions during a given time period. (sessions as identified by transient cookie values)

Unique Visitors: Count of all the Unique cookie_id’s during a given time period. (cookie_id’s from persistent cookies)

Most typically Visits, Visitors, Sessions refer to Total Visitors. Cookies is a toss up, refers to both depending on your tool. In StatCounter the value that is called Unique Visitors in reality, per their Support Staff definition, is Total Visitors (sum of sessions).

Benefits:

  • From my humble experience working in a few different companies: People instantly relate to these two metrics without any explanation. Total and Unique Visitors. No vagueness. All other versions if you don’t explain the definition it is impossible to understand what they really are.
  • The benefits of standardization is that independent of the package you use if you call something an apple it will be apple in another package.
  • It will be less confusion for our decision makers when they sit around compare notes with their peers around poker tables. Faster and more confident decisions will result.
  • Above bullet is not just applicable to the Visitor metrics but think about Conversion Rate. It is Outcomes divided by Unique Visitors but for the same site, this blog, if I compute the number using ClickTracks I’ll get one number (in CT unique visitors really is unique visitors) but if I do it using StatCounter I’ll get another number (because in SC unique visitors really is total visitors).
  • There will be world peace (I can't guarantee this of course : )).

Implications:

  • Vendors: Our dear vendors will have to agree that world peace is important and in some cases relabel the metrics in their reports (for example Google Analytics will change it from Visits to Total Visitors, StatCounter will change the metric from Unique Visitors to Total Visitors etc).
  • Users (You & I): We will dig into our web analytics package today, or call our vendor support rep, and ask them exactly how they define the Visitor metrics. One simple reason, if you get this metric wrong, every thing else you report out is suspect.

While we are on the topic of visitors…….

Two Special Unique Visitor Metric Caveats:

  • badge 2D12I am sure you have heard of all the hype around people loving their cookies and eating them up (or blowing them away if they are on a diet). The Unique Visitor metric is affected by the cookie deletion issue, total visitors is not. The numbers for first party cookies are a 2 – 5% and for third party cookies there is a lot of hot air around what the number is.
    All you need to know” (a nod to Stephen Colbert): 1) Get on first party cookies. 2) Trend it over time and you’ll be fine.
  • Check how your vendor computes Unique Visitors. Your vendor should have the ability to compute uniqueness over the time period you want (atleast for a month). So if you want to know Unique Visitors in August it is the sum of all unique cookie_id’s from Aug first to Aug thirty first. If they compute Unique Visitors as the sum of unique cookie_id’s on Aug first plus Aug second plus Aug third …… plus Aug thirty first, that is not really Unique Visitors.
    All you need to know”: In that case simply use Total Visitor numbers, ignore the Unique Visitor number.

Was the process of revisiting the grand old metric helpful? Are you more confused than ever? Do you agree with the two proposed standard names? Will the vendors switch their naming conventions? Please share your feedback and alternative points of view via comments.

[Like this post? For more posts like this please click here.]

Comments

  1. 1
    Robbin Steif says:

    So I guess you *do* get to take Damini to kindergarten today, or maybe by tomorrow. Congratulations to you both and remember: when they go to college, they think they don't need you any more. And at some level, beyond the financial one, they are getting to be sort of right about that one. So you are smart to treasure all these moments with her, the tough ones and the first-day-of-kindergarten ones.

    You really think ClickTracks measures uniques? What about all those people who check out your website at home and buy your product at work? What about those 15 reasons why unique visitors aren't unique, as written by Matt Belkin?

    Why is conversion rate outcomes/unique visitors? I think conversion rate is better measured by outcomes/visits and if you love unique visitors, you can measure unique visitors who took action divided by unique visitors. You will thereby match metrics in the numerator and denominator and learn two things: of all the visits to my site, how many ended up in an outcome? Of all the unique visitors to my site for the period, how many took action at least once? But for the second one, you have to believe in the uniqueness of your visitors, and I really don't. Long ago, I just decided that web analytics were imperfect and that I would get over it, so I started using visits because they I didn't have to obsess about how bad the unique measurement is.

    Robbin

  2. 2

    Robbin: Sadly Damini still has several weeks of recovery and staying at home ahead of her. Blogging turns out is a great distraction for me.

    For the reason you mention no tool really measures "unique" and hence in the post above I stress that unique cookie id's measure "browsers" and mention that that means IE and FireFox. But some websites cope with it better than others. Think of my.yahoo or amazon or wells fargo, all of whom recognize me without me logging in and for all of them in ClickTracks or WebTrends or … I'll be Unique (ahhh I feel special : )).

    Measuring unique visitors and using it for conversion rate reporting is also beneficial for many business reasons and could be critical based on your website visitor behaviour. This is one of the few things on which Matt and I'll disagree.

    There is a section called Why Use Unique Visitors in my post on Excellent Analytics Tip#5: Conversion Rate Basics & Best Practices. Please check it out for more context.

    Each business is unique and the answer of which metric to use will come out different based on the business, its strategy and goals, website purpose and customer behaviour. In turn choosing to use Total or Unique will drive company behaviour around "convince me to buy" strategies for its customers. I would simply encourage a open mind and picking what is best for your business.

    In the end what is important is not what metric you use but rather that you use a tool and you use a metric and that you are able to find actionable insights . That is the ultimate validation. No actionable insights means all definitions are useless. : )

    Thanks so much for taking the time to share your thoughts.

  3. 3
    Angie says:

    Hi Avinash,

    I'm glad Damini is back home, and wish her a speedy recovery!

    Regarding your definitions, I would argue that the "Sum of all the Sessions during a given time period" is actually "Total Visits" or "Total Sessions," not "Total Visitors." When we talk about "visitors" we're trying to get a feel for the number of people who came to a site. We all recognize it's an imperfect metric for many reasons (cookie deletion, cookie blocking, multiple browsers, multiple computers, shared logins, etc.), but we do the best we can with the available technology. Visits (a/k/a sessions), on the other hand, describe the activity being performed by these people. It's a very important distinction… I wouldn't recommend adding up visits and calling the sum visitors: IMHO that makes things much more confusing then they need to be.

    You'll be happy (I think!) to know that the WAA Standards Committee has almost completed definitions for what we call the "Big 3" terms: page views, visits, and unique visitors. I will send you a link when they have been finalized, which will hopefully be within the next week or two. We are including an "ask your vendor how they measure this" clause in each because many aspects of the big 3 are configurable depending on the web analytics tool you use. So obviously I am 100% behind your advice for users to dig into their tool's documentation or call their vendor rep!

    With respect to Robbin's comment, either visits or visitors could be appropriate for a conversion rate calculation depending on what type of conversion rate you are trying to measure. If you are measuring a conversion that can be performed multiple times by each visitor (think cost-avoidance conversions like reading a company standard on an intranet or downloading a product specification sheet from a corporate site), it may even be appropriate to measure both ways.

    angie

  4. 4

    Avinash,

    Great post as usual :) Please do write mediocre posts every now and then though, you're spoiling us rotten!!

    Anyhow, in the spirit of Stephen Colbert (great minds think alike), I am embarking in a daily series:

    Better Know a Web Metric!
    :)

    First installment tomorrow at http://juline.coquet.free.fr

  5. 5

    Auuuuugh! Avinash please reconsider what you wrote above. You said:

    Total Visitors: Sum of all the Sessions during a given time period. (sessions as identified by transient cookie values)

    Unique Visitors: Sum of all the Unique cookie_id’s during a given time period. (cookie_id’s from persistent cookies)

    Why would you define "total visitors" as the "sum of all sessions"?! The "sum of all sessions" is SESSIONS. Your definition will dramatically overcount "visitors" in situations where individuals/browsers return multiple times during the timeframe under examination!

    Angie from the WAA is spot on. Your definition makes the problem worse, not better.

    Regarding the term "unique visitors" … I suppose we all agree there are problems with the words and their interpretation. Some technologies struggle to count "unique visitors" and force them into oddly shaped containers (your rant on "Daily Unique Visitors" comes to mind.) Others do not and are able to provide de-duplicated visitor counts over whatever timeframe the analyst is interested in. Your advice to "ask your vendor" is good indeed!

    Personally, I envision a beautiful world someday where there is a single definition of "page view", a single definition of "visit/session" and a single definition of "visitor" that we can all agree upon. This pollyanna vision is of course clouoded by a number of external forces, not the least of which is the emergence of "Web 2.0" technology.

    So please consider revisiting your definition of "Total Visitors". It is obvious that many people are listening closely to you and if your definition became widely used we'd all have to ask "when you say 'Total Visitors' do you really mean visits or sessions? Or are you really talking about unique browser applications?"

    Your loyal reader,

    Eric T. Peterson

  6. 6

    Angie: Excellent point. Like me I am sure it will make others as well really think.

    You are right that the sum of all sessions is Visits. But visits are by "people" right? That is how I was thinking of it in terms of this post: "We have "visitors" to our websites, how many did we have in total and of those how many were unique?".

    From "Total Visitors" there is a logical drill down to "Unique Visitors" and no need to understand the "complexity" : ) of sessions and cookies etc. If I dig down deep into my psyche I suppose that is what it is.

    While using cookies and sessions and javascript tags the intention is to get the focus on "people" and not browsers and to shift, just the thinking, to a more customer centric mindset.

    Think of each "visit" as one by a "visitors" who has a "primary purpose" that you, website owner, needs to understand and then solve for. Forcing a understanding down to a session level in a way that Visits does not but Visitors does. Absolutely positively IMHO.

    You have presented a excellent view on why Visits should Visits and not move to Total Visitors, great food for thought for me and our readers. Thank you.

    (PS: I absolutely look forward to reading the new standards and welcome the great efforts that the WAA Standards Committee is putting forth to better our world. Committee members have our eternal gratitude.)

    (PPS: Eric I was writing this reply to Angie when your comment came through. The logic behind the thinking is in this comment. Thanks.)

  7. 7
    Steven says:

    Total Visitors: Sum of all the Sessions during a given time period. (sessions as identified by transient cookie values)

    Unique Visitors: Sum of all the Unique cookie_id’s during a given time period. (cookie_id’s from persistent cookies)

    I would rephrase "sum" to "count". This minor difference reminds us that these particular metrics are not additive.

  8. 8
    benry says:

    Great post Avinash.

    When I first got into web analytics, I found this confusing as well — in particular because the many professionals pontificating about KPIs were inconsistent in their use of visitor definitions. Thinking of this count in the context of people has always helped in my communication with others. Interestingly the WAA (Guy Creese and Jason Burby) put together their thoughts on the issue in a paper. In it they leaned towards this definition:

    Unique Visitors — IAB definition of “Unique Visitors” determined by the
    persistent cookie method and affiliated adjustments.

    Does their definition still fly?

  9. 9

    Benry (Scott): The WAA definition, which is based on the IAB definition, has wings and very much "flies". : )

    "Unique Visitors as determined by persistent cookie method": is very much the standard in most/all tools now (and lines up with what is in the post above).

    "and affiliated adjustments": here the IAB refers to filtering out robots, this is not a issue with the javascript tag based solutions but it is very sound advice as we compute numbers (some tools like Gomez site monitoring "robot" will execute tags and accept cookies).

    Thanks so much for the comment and providing the reference to the IAB / WAA definition.

  10. 10
    Angie says:

    Benry, Avinash, we have made some changes to that definition you're discussing. The new one will be posted soon (hopefully after this week's meeting).

    Avinash, thanks for your reply to my comment…

    …But visits are by “people” right? That is how I was thinking of it in terms of this post: “We have “visitors” to our websites, how many did we have in total and of those how many were unique?“.

    Think of your visitors as trees, and their sessions as apples. If you have ten apples, you can't make any assumptions about how many different trees those apples came from. If you bought them at a large grocery store, they *might* have been harvested from ten different trees. But if you bought them at the local farmer's market, or I picked them from my back yard, maybe they all came from one tree (or two, or seven, or ten). The only assumption we can make about the number of trees is that it was somewhere between one and ten. If I assume I have a total tree count of ten, I could easily be off by as much as 900% so I'd just stick with saying I have ten apples.

  11. 11
    Jaimie Scott says:

    Hello Avinash,

    Thanks for sharing Damini’s story with us. I’m glad to hear she came through the third surgery well and that she is recovering. She sounds like a very courageous little girl.

    I like the idea of revisiting "standard" metrics definitions and from the responses you evoked it looks like they are not as "standard" as they should be. I'm responding to your request for variations used:

    Most commonly prevalent names for visitor metrics are: Visitors, Visits, Total Visitors, Unique Visitors, Sessions, Cookies. (Please add the variations you have heard via comments below.)

    While my experience is mostly limited to Omniture, I’ve discussed this with the folks at Omniture and read their whitepapers on the topic. Some of the SiteCatalyst reports contain the metric “Visitors” as well as “Visits.” I am told that if it is not otherwise specified, Visitors refers to Daily Unique Visitors in SiteCatalyst reports.

    The numbers reported for Visitors in my particular implementation consistently exceed those reported for Visits by about 10%. I am told that this can be explained by the fact that Unique Visitors can be determined using other methods such as IP Address and User Agent String in cases where cookies are disabled (for which Visit data can not be collected) as you suggest in your post.

    Jaimie Scott

  12. 12

    Angie: Hooray for analogies! I love analogies because to a simple minded person like myself, and I am absolutely that, they are a very effective communication medium. After all we learn from stories.

    I imagine websites as interaction centers. They exist to sell stuff, help you and I find information, or just a massively waste time by reading the latest gossip. But in all cases we consciously choose to go to these websites by initiating an action from our end and then subsequently interact with them just until we are either satisfied or upset and we leave. The constant is us, the difference is the site: how many we visit and how often.

    So a different analogy from apples and the trees. Imagine a supermarket. It is there and constant. You and I and EricP and Benry walk into the supermarket and initiate the interaction. I walk in once (because I am not into shopping) and you could walk in twice this week (you forgot to buy apples the first time :) and Eric and Benry like supermarket pre-cooked food so they walk in five times each during the week.

    So in the end we could absolutely count visits to the supermarket. But my stress is that we should not think of it as an "abstract" thing as a "visit". Rather that we had 4 unique visitors who came in (and the supermarket knows because we used the club card) and in the week the supermarket had a count of 1+2+5+5 = 13 total visitors walking in through the doors.

    The analogy could just as well apply to amazon or nytimes or symantec.com/support.

    The difference is in the perspective of thought. The difference, to stretch my analogy to its logical point of death, is standing outside the supermarket (where it does look like a “visit”) vs. standing inside the doors of the supermarket and looking at doors opening and people walking in. (Imagine the scene in the movie Matrix 1 where the camera pans brilliantly around Neo as he goes into his kicking pose!)

    The challenge/opportunity is that there are existing standards and generally accepted terms (Scott pointed to IAB definitions). Visits is one of those and it measures something and means something (just like another great one “clicks” or “impressions”). Depending on

    1) what we are trying to measure
    2) which industry we are interacting with
    3) what outcome we desire

    we will stick with the metric that exists or evolve to new ones that help us think and make decisions better. Either of those is a valid outcome.

    Absolutely great discussion. Through your comments on this post and via email you have absolutely stretched my personal thinking (and those of our blog readers) and I thank you for that.

    The beauty of our world is that we are still babies and like babies we have unlimited potential for what we become and a unlimited set of options to choose from (every single day and during the course of our evolution from babies to single minded dogmatic adults!!). Thanks again.

  13. 13
    Clint says:

    For anyone interested, I believe that you can download the IAB's measurement guideline document here and read their guidelines on 'unique measures' (pp.16) and it goes a bit beyond what has been stated here by making distinctions between unique users, visitors and unique browsers.

    I'm all for having accurate definitions but I'm also not a fan of recreating the wheel and think that the IAB definitions still hold up well.

  14. 14
    Peter Cohen says:

    Excellent post Avinash, thanks to you and Angie for pushing the debate further. I concur with others in the comments that there is not a lot of clarity, or standardization, between the analytics tools as to how even the most basic metrics are reported.

    IAB and other standards are fine to have but if no one is following them at implementation it is hard to see their value.

    At our company we have voted to using the term "visitors" instead of "visits" that our web analytics tool provides, for reasons you and others have already touched on. But we decided against using "total visitors".

    So how do you feel after stepping into a minefield? :)

  15. 15

    Avinash, you say:

    "So in the end we could absolutely count visits to the supermarket. But my stress is that we should not think of it as an “abstract” thing as a “visit”. Rather that we had 4 unique visitors who came in (and the supermarket knows because we used the club card) and in the week the supermarket had a count of 1+2+5+5 = 13 total visitors walking in through the doors."

    Your "Total Visitors" definition works in a grocery store/movie rental/ice cream parlour analogy but falls apart in most other situations.

    Take, for example, Dylan's challenge at TurboTax. Would Dylan really have 13 visitors he could sell software to? Or does he really only have four people he can sell software to?

    I think the latter, especially since 1/4 is a great conversion rate but 1/13 is slightly less good.

    I totally get what you're trying to say. I just think there is another way to approach it. You could have just as easily said (my changes in bold):

    Total Visits: Count of all the Sessions during a given time period. (sessions as identified by transient cookie values and/or page views separated by at least 30 minutes of inactivity)

    Visitors: Count of all the Unique cookie_id’s during a given time period. (cookie_id’s from persistent cookies)

    You could have then cautioned your readership that sometimes vendors say "Visitors" when what they mean is actually "Visits" and encouraged them to use one or the other metric depending on what they were trying to calculate.

    This, incidentally, is why I always encourage retailers to measure both their ORDER (= orders/visits) and BUYER (=customers/visitors) conversion rates. Individually they're interesting but in combination they tell you a great deal about your sales process online.

    Anyway, I agree, this kind of friendly discourse is so nice! It's funny that it happens here and all we seem to talk about at the Yahoo! group anymore is Google Analytics and job openings.

    Eric Peterson

    P.S. For the record, my omnipresent beer-belly comes from exactly that, the occassional Pacific Northwest microbrew! Pre-cooked supermarket food rarely passes my lips, if ever. ;-)

  16. 16

    Dear Mr. Peterson: It is really hard to take our eyes away from admiring your beautiful pony tail and admire the rest of you!! So no comments on the "belly" part (but I'll try to notice the next time I see you!!).

    For the record, my omnipresent beer-belly comes from exactly that, the occassional Pacific Northwest microbrew! Pre-cooked supermarket food rarely passes my lips, if ever. ;-)

  17. 17

    Avinash, you wierdo. Are you Fred McMurry? ;-)

    I can't tell if your avoiding the question means you're tired of the post or you just give up and Angie and I win.

  18. 18

    Ok, no metrics packages will agree in their numbers for the same site (were they all being used to measure the same exact site at the same time – just as an example) – because of how a visitor is defined. I wrote a post at webmetricsguru.com about it.

  19. 19

    Eric: On the blog Occam's Razor there are only winners, no losers. :)

  20. 20
    Steve says:

    Hi Avinash,

    I'll have to politely disagree. Two principle reasons.

    1. I really don't care how many visitors or visits or sessions or whatcamacallit's. I only care about how many *people*. And trying to define a person by what technology they use (Javascript, Cookies etc) is fundamentally flawed. I can easily provide examples where your definitions will fail to measure real people who use various web sites I manage etc.

    One of your earlier posts even spoke to this: The numbers suck. Get over it. :-)

    2. The IT industry and indeed any industry, is too keen to pervert common language into terms that have special meaning in that industry.
    I dunno about the USA, but here in Oz, if a friend is comes to your house they are a Visitor having a Visit. :-)

    If vendors are perverting these words to mean different things (I came across one very recently which does exactly that… 25,000 "Visitors" in an organisation of 4500? From when I analysed the Intranet. I don't think so.) then those vendors are still stuck in the numbers game – not the helping achieve understanding game. And that is skirting damnably close to fraud in my very personal and highly cynical opinion.

    In my own tool that I wrote to make use of the Apache module mod_usertrack and logging the results (unique cookie) thereof, I stick with the house analogy. A Visitor comes for a Visit.

    If a person leaves and comes back in the same analysed period: week, month, year; then they are still the same person. The same Visitor. But this is a new Visit.

    The principle reason I have stuck with these terms is for the aforementioned one. They are common terms. People already understand their meaning. I *DON'T* have to explain what a session is. Or a Unique Visitor. How can you have a "Unique Visitor" Every visitor/person is unique! ;-)

    Thoughts?

    – Steve

  21. 21

    Steve: If I am understanding you right…..

    I am in agreement with you that we should care about "people" as much as we can and use measurement systems that get us as close to measuring behaviour (not just count) of "people" as possible. We have different technologies (tags, logs, cookies etc) and customer behaviour (switching machines, browsers, logging in and out) that help or hurt this cause.

    But try we can and try we should. Both metrics below have that basic thrust built into them, Think People is the slogan, even if we have slight debate on the names of the metrics.

    Thanks so much for the feedback, it is greatly appreciated.

  22. 22
    Steve says:

    Avinash: Exactly. People are the focus. Regardless of what type of site you run.

    The names used by us do matter, as otherwise we will end up with another "Hits" scenario. What a word means and how it's used become impossibly skewed and meaningless.

    – Steve

  23. 23

    The whole point seems to revolve around the potential ambiguity that "visitors" and "visits" are *almost* the same words. (and folks like StatCounter don't seem to be helping matters.)

    For our own product, we simply call a person a "user" and their visit to a site a "visit". Case closed. Thus we track users, visits, visits per user, etc.

    For any given site, the likelihood that their users reject and/or delete persistent cookies simply leads to an exaggeration of "users" (i.e. overcounts).

    It's not really that complicated. Not sure why we can't just all agree. ;)

    cheers
    Dave

  24. 24
    Jacques Warren says:

    Ouch! Am I really going to add to this great discussion? OK, disclosure first: I am a WebTrends partner. Now, notwithstanding that fact (and PLEASE, don't see the following as a "product plug" in any way), here is how in that world the visitor question is treated, and I believe it is a clear one.

    A visitor is always unique. Why would you need to calculate total visitors, if the number ends up being the equivalent of visits? Why would you sum the daily numbers in order to get a weekly or monthly one? Why say 3 visitors this week, when in fact it was 1 visitor who came 3 times?WT will not use the term "unique" anymore, since uniqueness is part of what a visitor is; what matters then is the time period you're looking at.

    Then, the challenge is what you will use to count them uniques. Well, it depends on what you have. If you have only the IP addresses and add the User Agent, well, you can count visitors, but you will be wildly off the mark, without being able to know by how far. If you go cookies, you'll get different numbers whether they're session or persistent ones, the latter being the favored mode of calculation in the industry. Finally (btw, there are other methods, but they rarely apply), you can count visitors on logins; only then do you count people.

    Personally, I prefer to calculate conversion on Visits, i.e. the % of desired outcome on the number of opportunities (I follow Jim Novo totally on this). It makes that calculation less dependent on many inconsistent factors (such as definition!).

  25. 25
    Sasi kiran says:

    One major issue that i find with conversation rates using unique visitors is the assumption that only one visitor is involved in the transaction. In our case, where we sell web hosting services this assumption is not true. Let us say, if 3 people from my prospect's office browse the site, i cannot say that i have 3 prospects. Infact, it is just one prospect. Is there a way to solve this?

  26. 26
    Hugh gage says:

    Hi Avinash,

    Thanks for this great post which I've read rather belatedly.

    There is one point you touch on briefly which also vexes me; the issue of measuring total unique visitors (persistent cookie) over periods of greater than a month.

    Some tag based analytics vendors seem only able to do this for a maximum period of a month (mentioning no names). For example if I am asked by a client “how many unique visitors came to our site last year?” I’m unable to give them a straight answer. Being able to count daily, weekly and monthly total unique visitors is useful but I think it would be of considerable interest to be able to assess this over longer periods of time. I think I remember being given an answer by one vendor support along the lines of “if looking at periods greater than a month there is generally too much data to process – especially if the site has very high traffic”.

    Having said all that it is possible to count number of New (unique) Visitors over longer periods of time. So then we get into the issue of returning unique visitors – monthly, weekly and daily or over any set period of time and how these are calculated. Again, on this point with some tag based tools it seems only possible to measure returning unique visitors for periods of up to one month meaning we are stymied when trying to measure total unique visitors over periods of greater than a month.

    On the other hand, if using a decent log analyser with log files configured to collect a 1st party persistent cookie I think this isn’t an issue – although I accept that one wouldn’t be using a log analyser for high traffic sites. In this case is does seem possible to measure repeat visitors for periods of greater than a month since the analyser (after some huffing and puffing) is able to de-dup repeat unique visitors over the whole selected period.

    Am I missing something here (hoping not to be totally shot down in flames :-) ) or is my assessment correct?

    Many thanks and best wishes, Hugh

  27. 27

    Hugh: Great question, here's an attempt at a cogent answer.

    Almost all vendors have a different way to deal with Unique Visitors and it is very important that you spend time understanding exactly how your web analytics tool is measuring Unique Visitors.

    Most vendors will only do daily, weekly, monthly because beyond that it costs them money and time to compute Real Unique Visitors (having to look back for a long duration of time to check if the new cookie id is unique or not). And it does not matter if they are log file based or javascript tag based.

    But this is not hopeless, let me share some examples:

      I understand from Matt Belkin that it is possible for a Omniture client to request Omniture to compute Real Unique Visitors for any time period you want. It is not a out of the box functionality but Omniture will enable it for you (please check with your Omniture representative if there is a cost to this and if the computation is going forward or can be applied retroactively). But it is possible.

      In Google Analytics there is a metric called Absolute Unique Visitors that should give you exactly what you are looking for (and it is available out of the box). Please read this post on the GA blog by Gauri Deshmukh: "Absolute Unique Visitors" versus "New and Returning".

      In ClickTracks, I am using Version 6, it also looks back for any time period you are reporting on and it will report Real Unique Visitors out of the box. I think in future versions it will be configurable value (you pick the time you want it to look back).

      I am aware from Eric that Visual Sciences also computes Real Unique Visitors (I encourage you to check with your Visual Sciences company representative).

    All three of the above are tag based solutions and will give you what you want. I am sure other vendors might have such options as well. I guess the "secret" is to 1) ask your vendor what they measure and then 2) push on them hard to give you want you want (because for the most part it "costs" them something but they will give it to you :)).

    Hope this helps.

    -Avinash.

  28. 28
    Hugh gage says:

    Thanks Avinash,

    I’ve seen those metrics in GA but didn’t know that Omniture had the capability to provide them too, albeit as additional functionality. I bet they do charge but it’s useful to know about nonetheless.

    Very best, Hugh

  29. 29
    Matt D. says:

    Hi there

    I find it really weak that Omniture's SiteCatalyst is not allowing Weekly or Monthly visitors. Once you buy their tool and get started, retag all your pages and get to the point of using it, then they disable these metrics for most reports and say it is too costly for them to provide.

    Matt D.
    Texas

  30. 30
    Minjae says:

    Hello,
    Let me ask regarding to cookie deletion issue.
    You said "The numbers for first party cookies are a 2 – 5%".
    Do you mean 2-5% of people delete first party cookies?
    Tha't only one question about this post. :D

    I appreciate this post solve my long vague problem on visitors and UV.

  31. 31

    Minjae : In Web Analytics there is no such thing as "people". It is all about "browsers", as in IE & FireFox etc. :)

    Cookies acceptance can be set at No at a individual browser level. Cookie deletion can be set at a browser or a system level (firewall / spyware / other "protection" software).

    But Web Analytics tools can help you measure some element of this. Here is my own IndexTools report for cookies. . . .

    IndexTools Cookie Disabled Report

    Hope this answers your question.

    -Avinash.

  32. 32
    Steve says:

    Ok Mister Avinash. I'll bite. :-)

    1. It *is* about people – we just proxy for people via # of browsers. If it's not fundamentally about people, then whom are we optimising our web sites for? Number Five? ;-) (http://www.imdb.com/title/tt0091949/ [1])
    I know what you mean/meant but … well… red rag to bull. :-)

    2. I *assume* that that Cookie Support measure is purely done via page tagging measures? ie The gif was requested, but the attached cookie in the html headers keeps changing, or is null'ed/stripped?
    That leaves oh so many places where cookies can and frequently are zapped. :-(
    Also subject to where you set the cookie – client side with JS, or server side.

    And none of which answers how often they get zapped either….

    I've always worked… backwards to compute deletion/rejection rates. Work out what IS accepted. Estimate how many… :-D *browsers* are left and call that a NUMBER.
    Rough but well… open to better ideas. :-)

    Cheers!
    – Steve
    PS Liked the context post! Nothing to say tho. Surprisingly…
    [1] Possibly showing my age. :-D

  33. 33
    Jason S says:

    Something is bothering me about the way Google analytics handles absolute unique visitors for the site above. I realize this 'metric' is now outdated as a modern KPI device but I need to understand something:

    It basically comes down to analyzing how many unique visitors or 'people' visited our site in 2009. I have read the 'hotel' example on Wikipedia using 'rooms' as an example and I also cross reference with independent audit tags via Quantast to assess site as a whole.

    BUT…. I am still confused (see table below). Review 2009 unique visitors:

    Jan No data
    Feb No data
    Mar 29,269
    Apr 29,121
    May 25,843
    Jun 31,705
    Jul 33,385
    Aug 35,181
    Sep 31,813
    Oct 33,183
    Nov 29,980
    Dec 31908
    ———-
    TOTAL UNIQUE VISITORS: 311,388

    Notice that Google ADDS the montly visitors together for 2009 in the analytics program by default which surely can not be correct. In other words, there is no way that 311,388 'people' visited our site? What value would I get from this 'fake' number?

    I could average the results externally in excel to produce this metric:
    25,949 average visitors per month

    When presenting stats to management should I present the average?

    Is there a better way to get some idea of the amount of unique people that visit a site in on a yearly basis?

    Thanks
    Jason

  34. 34

    It's interesting that it's been 5 years since this article was written and it can still be difficult to identify an individual user.

    This topic should be revisited!

    Current trends?
    FB and FB sign in is so ubiquitous now that that may be a better way to learn about your visitors. The nice thing about FB sign in is if the user is encouraged to do it then it is cross platform… Speaking of which…

    New challenges?
    I personally own 6 devices that can browse the web and I use 3 different browsers. I'm definitely a special case, but more and more people are owners of multiple devices. This makes identifying unique visitors a more difficult problem and it artificially increases your visitor count.

  35. 35

    Bonnie: It is very very important to mention that Unique Visitors are not Individuals. If you do many things right then UV's get close to Individuals, but they are not the same thing (as clearly highlighted by your multi – device fragmented existence! :)).

    Over the last five years somethings have gotten a lot better. Almost nobody, except the most suicidal, use third party cookies. First party cookies are much much better and all browser defaults favor it and hence the UV counts have gotten better.

    Technologies like the Async javascript tag from Google mean that you are able to collect even more data cleanly without having any detrimental impact on the consumer experience.

    We are still not people centric, yet. Tools like GA don't allow you to collect PII data. Let us add to this the government scrutiny and regulations.

    Over time this will evolve and if the Users are comfortable with it then maybe tracking Individuals will become the norm. A la Facebook.

    Device fragmentation is easier to solve if the company (you) cares enough. For example New York Times makes it worth my while to login and so across my desktop, my Times Reader Adobe Air app and my Android cellphone I am always logged in, making it easier for NYT to track me.

    I must point out that understanding each person is vastly overrated, unless you have a system to react to each individual with automatic customization (and massive massive content for pico-customization). Tracking Unique Visitors and segmenting them into micro group usually suffices for 99% of the cases and drives valuable action. So not having 100% clean data is not an excuse not to use the gold we already have.

    -Avinash.

  36. 36

    Hi Avinash – ah yes, unique visitors defined by cookies do not track individual users. I agree – for most publishers/product owners this problem is "solved."

    Honestly the "how to identify an individual user" is just a fascinating intellectual problem for me :)

    I initially started thinking about it from the perspective of

    – designing personalization and recommendation systems (a la if you're netflix or amazon etc)

    Right now, most people won't customize their website to that level, but I wonder 5 years from now if there will be new tools that will make it easy enough for your everyday publisher to do that type of implementation. Wouldn't that be crazy? I should put a note on my calendar to check on that in 5 years…

  37. 37
    Rafael Galante says:

    Why when we export unique visitors data from GA, it brings the last row as a sum?

    Isn't that just wrong?

  38. 40
    Mahalakshmi says:

    Hey Avinash !

    At the cost of sounding like a total dud … Can You Please explain sessions , visitors , page views in GA context ?

    I am new to web analytics and GA and I am soooper confused

Trackbacks

  1. Standard Metrics Revisited: #1: Visitors – Avinash Kaushik…

    Avinash Kaushik is going back and getting to the bottom of basic metrics like visitors and uniques in Standard Metrics Revisited: #1: Visitors.  I have suspected the main reason no two metrics packages give you the same numbers (even the……

    http://www.webmetricsguru.com/2006/09/standard_metrics_revisited_1_v.html

  2. Goad.net { » Getting Started With Web Analytics: Step One - Glean Macro Insights. says:

    […] These metrics look deceptively simple, they are not. For more on what you should look for and things to be careful about please read this post: Standard Metrics Revisited: #1: Visitors. […]

  3. […] These metrics look deceptively simple, they are not. For more on what you should look for and things to be careful about please read this post: Standard Metrics Revisited: #1: Visitors. […]

  4. […] Standard Metrics Revisited: #1: Visitors » Occam’s Razor by Avinash Kaushik How to count visitors? Nice discussion… (tags: analytics) […]

  5. […] This is an incredibly broad topic and I'll certainly take the input for the updated Beginner's Guide. In the meantime, I might suggest you read Avinash's series on Standard Metrics Revisited – #1, #2 and #3. That should at least be a good start. […]

  6. […] This is an incredibly broad topic and I'll certainly take the input for the updated Beginner's Guide. In the meantime, I might suggest you read Avinash's series on Standard Metrics Revisited – #1, #2 and #3. That should at least be a good start. […]

  7. […] You can learn a lot more about Visits and Unique Visitors in this post: Standard Metrics Revisited: #1: Visitors. […]

  8. […]
    To make things even a little more interesting (meaning even more confusing) the help results includes a paper from Avinash Kaushik, from over 4 years ago – Standard Metrics Revisited: #1 : Visitors. Avinash is an important contributor to our understanding of web analytics and in this paper argues:
    […]

Add your Perspective

*