Excellent Analytics Tip #15: Measure Latent Conversions & Visitor Behavior

Off CenterHere is an astonishingly brilliant, yet simple idea (if I may say so myself!):

Why just measure conversions as one purchase or conversions just as a submission of a lead or opening of an account on facebook / twitter / what ever?

Why not measure Visitor behavior after that first purchase / lead / membership sign up (or the first super poke)?

Why obsess about the "quickie" that is opening an account? Why not the Visitor behavior in the 30 days after sign up?

For your Search campaigns.

Or your Display ads. Or Affiliate programs?

If you do that, you are not just measuring the "one night stand quickie" what you are measuring is something more of value: Was there a second date? Perhaps a proposal in five months? Maybe a marriage in nine months (because that's when the baby showed up). : )

We do this so rarely.

Success for campaigns (Search, Social Media, Display, TV and what not) is always measured based on that one visit or what happens when the campaign runs.

Yet it is likely that you are measuring incomplete success.

Let's do an example.

Facebook, in its quest for world domination, runs a world wide campaign (of whatever sort, really, make one up).

Success is to increase the membership on Facebook and thus making it a more "valuable" property (so that even in a weaker economy they can raise another billion dollars of cash).

So they get another 500k memberships.


How about this….

Rather than measuring membership sign ups why not wait 30 days and look at the Recency report for Visitors that came to the site as a result of that campaign?

visitor recency

Did that just blow your mind? : )


Ok, what you are measuring in the case of Recency is not simply people who signed up BUT rather the behavior of the visitors who signed up.

You are measuring if people you acquired from this campaign are visiting the site frequently lot in this 30 day time period after the campaign launched.

My hypothesis being that simply signing up is not enough. That's not enough success (not in this economy baby!). Real monetizable success is:

1) Do those people visit facebook again? And this is key…
2) They visit every day, or every other day, or every few days.

The more frequently people visit facebook.com, the more valuable they are to facebook. They see more ads, they super poke more, they friend more, they add more apps, and so on and so forth.

All activities (sexy translation: "visitor behavior") that is multiple multiple times more valuable then simply signing up.

When you look at campaigns not just from a "quickie" view (did they sign up?) you are not sure if you really added value to your business.

This is the kind of analysis you'll do. . . .

google analytics top box recency scores

And here's a specific example of what you are looking for:

Campaign 1: 100k sign ups. "Top Bracket" Recency: 10% (i.e A month after the campaign 10k of these Visitors visited in the last 23 hrs on the 31st day.)

Campaign 2: 50k sign ups. "Top Bracket" Recency: 50% (i.e A month after the campaign 25k of these Visitors visited in the last 23 hrs on the 31st day!.)

Which one is more valuable?

With the "quickie" method it's campaign #1.

With the "post quickie" method (visitor behavior) it's campaign #2.

Super Sweetness: Combine looking at the Recency report with the Loyalty (Frequency) report (below) and you have a gold mine of insights. You'll now how exactly how many times Campaign 1 and Campaign 2 Visitors visited. That's helps your truly find your BFF campaigns.

Now the reality is most of us will follow the traditional "lets make sure we measure conversions" path and rarely do this.

Usually because we don't know that free or paid tools now allow you to do this reporting standard off the bad. AND that they are so easy to use that even my 4 year old son can do these segmented views (while, and I am proud of Chirag for this, remembering to filter "New Visits" out!).

This is one other hidden reason: Marketers are impatient. They want to know now what happened so they can collect their bonus (not really their fault, its how the compensation works).

But if you as HIPPO's and Leaders put the right structures in place to incentivise the right behavior and then encourage the right measurement than Queen Success will be yours!

If you buy into measuring Visitor Behavior then another fantastic metric to measure your campaign "conversions" is Visitor Loyalty.

Your display campaigns are purely for a "branding" impact and the measure of success is nothing.

Ok just kidding (am I? :)).

Your display campaigns are geared towards driving people to your brand and not bounce. Great. Measure that.

But also measure this, do these campaigns results in a increased likelihood that people will come visit your site again? And which display campaigns (on Yahoo or New York Times) result in Visitors that after a 90 day period come again and again and again?

Campaign 1:

visitor loyalty analysis

Or "glorious" Campaign # 2:

google analytics visitory loyalty report

Clearly New York Times display campaigns (#2 above, as an example :) are terrible and should be abandoned because they are not resulting in a longer term relationship with the acquired Visitors.

This is what I mean by measuring "latent conversions" (by that I mean using visitor behavior as a "conversion").

More marching orders:

  • If you are a Social Media website (with no discernable outcomes) don't use "membership acquisition" as a measure of success.

    Show your investors and potential advertisers that your property is driving the kind of "sticky" / "engaging" / "viral" / "insert buzzword here" actual visitor behavior. They will love you more.

  • If you are a non-profit like NTEN (the "craigslist of non-profits") then use Visitor Loyalty metrics to measure success of your AdWords campaigns.

  • Even if you are a ecommerce website take the data from your web analytics tool (not easy with Google Analtyics at the moment, maybe with the API, but you have Omniture so you should be doing this already) and measure not just the first purchase based on campaign traffic.

    Look back 90 days and see which campaigns drove more repeat purchases. Value those campaigns higher, even if initial conversion was lower.

And more stuff like this (add your examples in comments).

Any Reporting Squirrel can report instant conversions. Only a true Analysis Ninja will do this kind of "post quickie" analysis and help the business make thoughtful and cost saving intelligent decisions.

Good things come to those who are patient, and those who focus on Visitor Behavior.

Ok your turn now.

Do you measure "latent conversions"? How about Visitor Behavior as a success? Especially pan session visitor behavior? Have I missed any other examples of analyses that can be done (it is late in the night)?

Please share your thoughts and feedback via comments.

[NOTE: I am using a new new wordpress plugin that will allow you to edit your comment for a couple hours after it has been posted. In case you make a mistake.]

Couple other related posts you might find interesting:


  1. 1

    Great tip. However, surely you should also benchmark both campaigns against the normal traffic to the site?? i.e. if the top bracket recency for campaign 1 is 10% and for campaign 2 its 50%, then campaign 2 is better. However, if the top bracket recency for all new site visitors is 75%, then both campaigns were rubbish!

  2. 2

    I will start my comment with something that you said: Marketers are impatient. Oh my god! YES, they are!
    And.. yes, it is so common to see people loving campaigns with a high "fast conversion rate" and forgeting the even higher value that loyalty can bring. Even me.

    Well, I would like to write a bit more, but I need to check somethings about some campaigns here (hehe).

    Thanks, again!

  3. 3

    Some interesting food for thought as usual.

    Won't a lot of this swing all sorts of previous findings tho?

    I mean, if for example now I am achieving 20% conversions on direct visitors, then I find that this client originally visited 2 months ago from a msn search for example, then this direct conversion rate is now actually defunct right?

  4. 4

    I guess we're all guilty of "one-night stands" (well I know I am…OH get your mind out of the gutter, you know what I mean, lol :))

    I guess the reason we do this is once the business man or woman receives that sale or that lead, everything else becomes somewhat irrelevant. This is probably why so many online businesses fail to offer some return customer incentive program or some return customer coupons or something like that – we care so much about that first acquisition, that everything else becomes less important.

    Should it be important? Yes I do think so, and this latest Excellent Analytics tip by Avinash is a great place for us to start making it important. How do we make it important? Well, for me, whenever I wanna make something important, I'll throw a slide or two in on a presentation for the client to get them at least exposed to that information.

  5. 5

    This is much more critical in long sales cycle processes like in B2B. The campaigns that generate qualified leads for sales, even if months later, are the successful ones. The campaigns that generate short term traffic with no added interest are a waste of money.

    However, when you add in the "proxy endpoint" concept (a qualified lead) and analyze to that, you get a much more complex situation to analyze. A true analysis ninja is needed, as you would say.

  6. 6

    Good stuff!

    At Nedstat we segment by 'visitors' not 'visits' thus enabling us to understand the exact behaviour Avinash is describing.

    This is key for 1.idenifying success of campaigns but 2. identifying upsell/cross sell opportunities in your existing customer base.

    It is all about looking at your traffic, not as aggregated 'visits', but as a big room of 'visitors' who each have many 'visit' stories.

  7. 7


    Could you share step-by-step sequence that your 4-years old executes in order to get Visitor Recency report?

    Other step-by-steps might be helpful too.

  8. 8

    Avinash, here is what is stunning to me: the fact that Marketing people cannot see how incredibly important this idea is, and are not demanding this kind of data from analysts.

    Recency is the most important metric nobody uses yet, in part because (prior to inclusion by Google Analytics) it was only available in the very high end tools. Now anybody can use it.

    What doesn't make economic sense about getting 5 visits / sales for the price of 1, as opposed to 1 visit or sale for the same price? And given that knowledge, allocating campaign budget accordingly?

    Further, the Recency metric is *Predictive*, it tells you what is likely to happen in the future. Forecasting is the most powerful tool a Marketer has / an analyst can provide. Don't Marketers want to know about the future viability of Campaigns / Customers / Businesses?

    Is it possible that Marketers are not even aware of or interested in this most powerful metric, they don't even ask analysts for it? Please let me (us) know, because if it's true, I'm wasting my time speaking at eMetrics…

    Analysts: Nobody asks you to provide this metric?

  9. 9

    With regard to the Visitor Recency report, the interpretation given is "10k of these Visitors visit in the subsequent 30 days visited once a day or more."

    I always thought that report told you how long ago their most recent visit was, but not how many times per day they had visited. Wouldn't the correct interpretation be that 10k of these visitors visited again yesterday? Have I been wrong all this time?

  10. 10
    Alice Cooper's Stalker says


    Pure Awesomeness! You are the best, of the best, of the best, of the best, of the best. (Okay, my kids have been watching Kung-Fu Panda way too much and the lines from the movie are burning into my brain)

    It's common to dangle that carrot or coupon or white paper out there in front of people and try to get them to give us their deomgraphic and contact data. They grab the piece of cheese we offered and they leave, never to return.

    I'm still coming up to speed with google analytics. Can you take this a step further and segment what activites the people that visited more than 1 time performed on the site? Or even more targeted segments (the traffic that visited 9-14 times). I guess, which lacency segment is giving you the best return. If your business model involves seeing ads, like you proposed, then the people coming back the most are probably going to offer the best value. However, not all businesses are alike and I'm sure you might find that in some cases, one lacency segement is providing better roi than another.

  11. 11
    Ned Kumar says

    I agree with you that the 'quickie' (account openings) should not be used as an end-all to measuring conversions or the impact/ROI from a campaign. However, I do believe it can be the 'starter' for the relay of metrics that measure this conversion — with RETENTION RATE being the anchor in this relay. Here is how I visualize the relay line-up to be — I am keeping it simple without going into detailed breakdowns:

    * Account Opening Rate / Membership-signin rate or whatever makes sense as the first step

    * Activation Rate (Opening an account does not add anything to the bottom line unless they make a transaction, whatever is meaningful for that site)

    * Attrition Rate (Track the number of folks that signed in and how many of them revisit over a period of time). One would be amazed at the exponential nature of this drop.

    * Retention Rate — the ultimate measure of success. How successful were you in keeping these new vistors/customers/members over a period of time.

    Like I said, this is a simple framework. Dispersed between these, one can do a RFM (Recency, Frequency, Monetary) on these customers to quantify their value and agree with Jim that vew few realize how simple yet powerful this can be. Or one can do a penetration model – is the returns from these customers increasing over time etc. etc.

    Great post on a topic worth listening.

  12. 12

    I love reading stuff like this. Direct Marketing Analytics are always the best. Like you, @JimNovo and @NedKumar outlined, there is so much more needed for this analysis. Being able to calculate all these measures and calculating the Life Time Value (LTV) is critical to successful market anything in any directly measurable media.

    Two quick resources on LTV:





  13. 13

    Avinash, another great post.

    As I understand it, GA and most other web analytics tools use cookies to identify visitors and track repeat visits. For a website with a long sales cycle like Stephen Woods mentioned above, the shortcomings of data collection through cookies (some people browse with cookies disabled, delete cookies periodically, change browsers, or use a different computer) might lead to significant data loss.

    Any tips on how to mitigate those issues?

  14. 14

    I'm going to open up AdWords and try that out right now.

  15. 15

    Clive: I am afraid I don't understand your question. So sorry.

    In this context I think of conversions as a secondary (and perhaps a lamer) metric. For the success of your business you want customers that are valuable (those you can retain, those who exhibit behavior that adds value to you, those who might buy more). So when you judge campaigns use visitor behavior as the primary measure.

    Sure measure conversions to get a pulse. But don't reallocate the budget or declare winners until you see that the people you acquired are behaving in a way that adds long term value to you.

    Joe: That would be a great way to start! To change minds I have presented what they wanted (conversion) but then later added / presented / "slapped them with" :) the latent conversions data. Slowly but surely they start to change.

    Jim: Marketers don't ask because they don't know (don't think?). Analysts are not really business savvy (really hurts). Sans you there is little practical consistent advice on how to do this (so no education). Potent trifecta. Sad.

    Alice Cooper Stalker: You can certainly segment more than x number of times. Just as an example, click drag drop and boom!

    Segmenting Valuable Website Visitors

    I want people who visited more than 9 times from a certain paid search campaign and there I have it, sad that it was only 23! :)

    Ben: You are right and in my late night blog writing I did not phrase it correctly.

    The first line will indeed show the number of visits that were in the last 23 hours. Certainly a great indicator of success. But not as I phrased it.

    The sum of the table in the recency report will show you how many you how many visited again and over what distribution of time (since the conversion).

    You can apply the same segment "Cool Lovely People" on the Loyalty report and that would tell you the number of times you saw them. Between the two you have a winner. :)

    I have corrected that part of the post. Thanks so much for pointing it out.

    Ned: Bravo! Love the specific drill down framework. Gracias.

    Jake: With cookies my biggest advice to you / anyone is to not buy into the hype. Understand. Internalize. Measure. Form an opinion. :)

    I had deliberately included a link to my "all you ever wanted to know about web cookies" primer at the end of this post. You'll find my tips there.


  16. 16

    Hi Avinash,

    Very nice way to explain the usefulness of visitor analysis. However, I am not quite sure you can do this with Google Analytics. As far as I understand, advanced segments let you segment at the visit level, but not at the visitor level.

    Hence, when applying a campaign advanced segment to the visitor recency report, you're only considering the visits that came from that campaign, which is not really what you want. What you want is all users who first visited you from that campaign.

    Do I make any sense here?


  17. 17
    awhinston says

    Great insights, Avinash. We've been thinking a lot about whether/how free analytics can help with longer-term retention analysis. Your post is very timely.

    I see how GA could do this for a single campaign with a definitive start and end. But what about an ongoing campaign like paid or organic search?

    To measure retention here, it seems you'd need to assign a date parameter to the first visit (e.g. ‘User Type=New’ and ‘Source=Google PPC’ and ‘Month of First Visit=January’).

    Otherwise you’ve got activity from every new user ever acquired through the ongoing program (w/ stable cookies) corrupting your ability to isolate retention activity in (for example) month 1.

    You did mention filtering out new users. However, if I define my segment as (e.g. ‘User type=Repeat’ and ‘Source=Google PPC’) then I have all the repeat users that click on my PPC ads (of which there are many) showing up in my retention report and corrupting the data.

    Can you please provide some additional thoughts on this? Thanks very much!

  18. 18

    While I understand the importance of watching customer behaviour over time and also beyond sign up, I do agree with Ned and think it can be a starter. I'm just thinking of people also subscribing to the newsletter when signing up for an account.

  19. 19

    If we have 3 goals, how can we know each other visitor behaivours? because in default segment part we can only choose visits with conversions not each conversions. I mean we cannot choose each goals' conversions.

  20. 20

    Hi Avinash,

    That is really a fantastic way of tracking retention success. Actually, I would add the metric "quantity" or better say, revenues per visit (when available).
    RFQ is a famous way to find gold in your group of clients when doing database marketing. And it is also great to use it while measuring success of campaigns results in terms of leads qualification.

    I will post an example of this soon based on your post/idea.

    However, I agree with Pere, (or at list, I share the same doubt). Is GA measuring visitors or visits when applying advanced segments? Could you please clarify this?




  21. 21

    Hi Avinash,
    Whilst I agree wholeheartedly with the philosophy here, I agree with Pere's comment (17 as I write this).

    If you do a visitor recency analysis, then the breakdown of recency is for the visits that occurred from that segment during the timeframe specified in the date range. If you segment by campaign, then we are segmenting at the *visit* level and we cannot ascertain an aggregate view of all their subsequent visits after a campaign event.

    Another way of putting it, is that if a visit to your website from a segment happened 30 days ago, but then does not return to the website during the timeframe you have set in your date range, when you look at the recency chart the data corresponding to the visit 30 days ago will not appear. I'm afraid this fact, unless I'm very mistaken, flaws your analysis.

  22. 22

    Great subject! This is actually what we are doing at the company I work for (an (online) media company). One of our sites can be considered a social network. When you haven't loged on for, let's say, 30 days. You'll recieve an e-mail. Within a couple of months from now we'll also start an ip-based campaign; you didn't log on for a while, you'll see all kinds of reminders around our entire network (including news and e-commerce sites). Collecting accounts is, like you said, one thing. Generating active users is pretty much the next level…

  23. 23

    This is all very applicable to the way in which one values online advertising as well. If you are valuing ads on the LTV of the customer it acquires while your competition is valuing their ads solely on the first purchase you have a competitive advantage. How? Well, you are valuing your ads more accurately and therefore will determine that you can afford to pay more to acquire a customer because you are looking at the lifetime contribution vs. the one-time contribution. There will be some ads that do not make economic sense when judged on the basis of the first order only. Therefore, your competition will shy away from these types of ads. You however, will not shy away from these ads because you have a lifetime view.

  24. 24

    Pere: It is accurate to say that you can't segment at a visitor level in GA. But like all things in life there are nuances! :)

    Let me explain.

    There are four reports in GA that implicitly use the anonymous unique visitor id (yes, without letting you or I choose it / use it). Recency, Loyalty (Frequency), Days to Purchase and Visits to Purchase.

    Those are all pan session reports that hold the unique visitor id "constant" to report the visit level data across time out to you and I.

    With that in mind here is what I did to accomplish what I wanted….

    Here is a campaign I wanted to analyze: Paid search. Ran through Oct and much of Nov. Then it was stopped for six weeks.

    Paid Search Keyword Campaign

    Lots of lovely people came to the site during that time period, which is nice….

    Visitor Recency Campaign Analysis

    But I don't care about the conversions during the campaign, I want to know if any of those people came back to the site. Any latent behavior?

    So I segment out those people and apply it to the next month period (Nov 24 to Dec 24). And here we go….

    Post Campaign Recency Analysis

    Essentially a pathetic amount (looking at the total). Sad. But worthy analysis.

    The distribution is nice to know, atleast got some people who came recently. I also go back and check their Frequency (using the Loyalty report).

    Here is the segment I had created to accomplish what I wanted to for the post 30 days analysis:

    Visitor Segmentation for Recency Analysis

    Notice both the segment created as well as the emphasis on just Returning, to ensure that I filter out laggards.

    What do you think?

    In GA you cannot segment by individual visitors, even anonymous (how super lame is that?) but you can do the above analysis to atleast get going on measuring visitor behavior for campaigns.

    If you are lucky enough to have dropped couple hundred thousand to a couple million dollars to get the DW version of other Web Analytics tool then you can do it even easier! And you should do it.

    What's important is that this technique / analysis is very critical for every business to do so that they can focus on long term sustained value. If Google Analytics can do it, fine. If it can't then you should get a different tool.

    I hope people take that away from this post.


    • 25

      I know this is very old, but still so very relevant so please allow me to ask a question here.

      I wonder if what you did actually makes sense.

      What you are essentially doing is create a filtered view for the period 30 after the period you had your campaign data come in with the following settings:

      Keyword: exactly 'X'

      Visitor Type: exactly 'Returning' (this is changed now and not available in metrics in segments)

      Medium: exactly 'CPC'

      To me what this means is you are looking at everyone who came in through the same campaign but is a returning visitor. Doesnt that mean that someone who visited your site BEFORE your campaign is also counted?

      Wouldnt you want a filter stating: I want to see all returning visitors that were first time visitors 30 days ago?

      Im struggling with a similar issue trying to figure out how much our online database is used by our registered users. I have data on when visitors completed the goal: 'registered for database', but cant seem to figure out how to see what their recency & loyalty is.

      For example:

      over 6 months I have 100 registered users for the DTB. 'Goal x completed' sessions: 100
      now I want to see how many of those 100 visited the site in the next 6 months. Would you create a segment: 'Goal X completed in the first 6 months' then run the Frequency & recency report for the next 6 months? You are not measuring a vistor tagged with 'goal X completed', you are measuring sessions with 'Goal x completed'

      I get confused even writing this down :P hopefully someone still reads this!

      • 26

        Vincent: You don't have to spend time understanding this post. Primarily because technology marches on and we have so many more ways of doing what this post was recommending (in Jan 2009!).

        Tools like Google Analytics have moved to being Visitor-centric (even in the free version) so you have several options.

        In order to understand the starting points of a visitor journey, and which campaign/channel starts it and all the steps along the way, you can simply go to the Top Conversion Paths report in the Multi-Channel Funnels folder. It is there in all its glory.

        Beyond that, you can also click on the Conversion Segments link on top of that report and you can choose the standard built in segments, like First Interaction is Paid Advertising, or create your own to get what you are looking for.

        Another option is to use Advanced Segments inside standard reports. Depending on what you want to do you can use Sequences (in the Advanced section), or Traffic Sources where you have an ability to create segments by Sessions or Users.

        Another option, specifically with regards to your question is to create a combination segment of people who are registered (you are likely using custom variables with scope equals one to track people who completed a goal, it is much more accurate), and go to the Behavior Section.

        Another option, is to use create your registered users segment, and apply it to the Active Users report in the Audience section of Google Analytics. It shows you 7-day, 14-day, 30-day actives.

        One final tip to alleviate your confusion. For your purposes, focus on User/Visitor analysis, use that level of segmentation to apply to the optimal report, make sure you are using custom variables to tag the registered people. That should open up a lot of possibilities for you.

        If you still need more help please reach out to a GACP, you'll find a list here: http://www.bit.ly/gaac


        • 27

          Avinash, awesome! Thanks for so much information. I will have to read this a couple of times, but it gives so much more entry points. great!


          (i ticked notify of followup comments, but never received…)

          • 28

            Still wondering, after going through and trying some things. (Active users is not part of my analytics unfortunately)

            I would like to track the people who have registered during the past year. If i create a segment (goal x > 0) that would only give me data for visitors who registered at that very moment.

            When I look at returning vs new visitors, this metric will not say: after registering, 30% returned to the site. It would say: out of all the people who registered at a certain date, 30% of those were returning visitors.

            How would I be able to measure how often the (goal x > 0) users come back to the site?

  25. 29

    Loved this post. My boss has been juggling a few different adword's adds around for a while.

    He's trying to target need niches for our larger business in order to try to create return visits and establish long term customers through very basic initial purchases.

    I absolutely have to dive further into this with your mentioned technique. Some of them aren't even working short term, so any money I can save him by focusin down on those that are giving us more return visitors is so KEY!.

    Thank you, (your article was my caffeine for the day!)


  26. 30

    Awhinston: For ongoing campaigns you would have to apply different kinds of analysis. For example maybe someone like Justin or Robbin could tell us if we can use a combination of "filters" and "profiles" to accomplish this with GA.

    But you can't do this using the segmentation feature in GA. Or atleast I don't know of a easy way to do it.

    Most web analytics tools would struggle with doing this in their tools (except the expensive Data Warehouse options you can buy from them on top of the web analytics tool). It's because you want to have a "trail" beyond the initial conversions for a "ongoing stream" of visitors.

    The only default report I had seen in a tool that does this was in Microsoft AdCenter Analytics. See this image:


    from this post:

    Goodbye Gatineau, hello…

    It was on Ian Thomas's "we are going to do this" post from a while back. But I don't use campaigns in my implementation of AdCenter Analytics so I don't know if it is out there and you can use it.

    But Ian's report is exactly what you want. Ask your vendor, Google Analytics or Omniture or WebTrends or CoreMetrics of IndexTools or whatever, if they can do it for you out of the box (not after you pay them lots of money for data warehouse solutions).

    Hope this helps.

    Bertwin: That is both awesome and scary (depending on whose point of view you are looking from!). : ) But it is a great example of how to productively apply this type of analysis.

    Ed, Juan Cruz: Please see my reply to Pere, on the blog now (http://cli.gs/MJvXDH), to see what I did.

    Speaking of the tool, GA places limits as to how far I can go, which is sad, but I think I can atleast do something to get going. And it is productive.

    I wanted to stress again that I want you/everyone to focus on the value of this analysis and the technique. The tool part, while important, is secondary.

    Serbay: I am afraid I might not be understanding your question. Perhaps you could post it to the Analytics help center?


    If I look at segmentation in GA I can choose overall goals, I can pick individual goal values etc. But I might be missing something here. My apologies for not being able to help.


  27. 31

    Comment #25 is remarkable. Conceptual suggestions are great, but nothing is better than seeing an expert roll up their sleeves and address an example, showing their work as they go.

  28. 32

    On a corporate site I measure we are looking at a group of articles that were created to increase customer awareness and are featured on home page. I've set up the tagging structure to call them out as articles of interest, and using the WebTrends custom measures I'm able to add up the dimensions for time frames of a day, week, month, quarter, and year.

    I hadn't thought to look at them other than for the monthly reports, and didn't think to add to the custom report the other measures. Now I'll put this post into my tool bag and pull it out to see if those who read the 'interesting' articles become return visitors and continue to connect with the corporate site.


  29. 33

    Re: Comment 22 and your reply comment 25…

    Hi Avinash,
    If I understand this correctly, what you are saying is that even though your segment specifies the ppc campaign, for the period you are measuring a month later (when the ppc campaign is not running) you will see any visitors that previously came to the site via that campaign. That would lead me to conclude that segments operate at the *lifetime* level??? This confuses things a bit for the analyst using segments doesn't it – they have to understand that when they look at these reports, they are not only looking at people that arrived *during that period* from that particular source/ppc/whatever, but anytime previously as well. Do I make any sense?

  30. 34

    Hi Avinash!

    Visitor behavior is definitely a part of our success measurement. It’s important for most of my clients to know what users are doing on the site and if it’s plausible to continue putting money towards visitor engagement. Before this, it was a guessing game. There would be a lot of sad people if this ever went away. ;)

  31. 35

    I am trying to replicate these reports and segments in Google Analytics and am coming up short because GA isn't telling me what it means by a "campaign visitor" (in terms of whether the initial, most recent, or current-visit-only campaign applies). I join my question with Ed Ralph's — what are the underlying dynamics here, within the GA tool?

    The concept of campaign latency is of course colossal – Jim Novo has been writing prolifically about it for years in the web analytics arena, and it's bread-and-butter for the offline marketing world. No question. But a step-by-step and some concrete definitions are needed, maybe not from Avinash, but definitely in the Google Analytics documentation. Neither "Campaign Visit" or "Campaign Visitor" is even in the GA glossary!

  32. 36

    Ed Ralph: I would not call it "lifetime concept" but if you think of the two cookies, session and persistent, that every tool sets. It is fair to say that the persistent cookie is present in very session of a person/browser and it can be used to tie sessions.

    Now, as I explained in comment #25 above, it does seem odd on the surface that you can't segment by that persistent cookie and still get the "lifetime" concept. Its just in those four reports that it is possible because I am "exploiting" the only four pan session reports (I can see the returning because GA is still categorizing them (visits) using the original unique id).

    I hope that the GA team will allow more than what it does today. Much cleaner. One where no exploitation would be required.

    Chris: Agree with you both on Campaign Latency and the Supreme Champion of Awesome Concepts Mr. Jim Novo!

    Please see my reply to Ed above, hope this helps a bit.


  33. 37

    From email campaigns, can you segment out individual buying habits in Google analytics?

  34. 38

    Re: Leaving where we left off; comment 31

    This is a good thread, with very important repercussions so I hope I'm not banging on too much. I don't know many other ways to say this, so let me try with this simple example which has a yes/no answer.

    A visitor arrives at your site via a *ppc* campaign 30 days ago.
    The same visitor arrives at your site *direct* 5 days ago.

    If you run the *ppc* segmented recency report for the period 10 days ago to now, will you see the above visitor in the report?

    I think no…

    Thanks Avinash for your continued support in clearing my foggy mind on this one.

  35. 39

    Measuring latent conversions especially makes sense for social media sites, where user activity is valuable. Lots of registrations are just spammer accounts with no visits or only one visit after the registration.

  36. 40

    Hi Avinash,

    Another good insight that makes you remember not to think in the box and always be critical to your processes, even though you know you´ve been following them for a while.

    Yes! we should measure loyalty and returning visitors and not only a sale.



  37. 41

    Fantastic post – It's a bit maddening when everyone becomes so focused on the quick conversion that they miss the forest for the trees.

  38. 42
    awhinston says

    I'd like to hear people’s thoughts on segmenting by visit (as GA does) and by visitor. It seems to me that – as long as the attribute you want to segment by is recorded in a persistent cookie that carries across visits – it shouldn't matter.

    For example, returning visitors: Is there any difference between 1) Segmenting a *visitor* as type='returning' and placing them and all subsequent activity in a segment and 2) Lumping together all *visits* in which the user type was 'returning'? Am I missing something?

    The GA custom variable also allows you to write to the GA cookie. So if I want to segment users who clicked a link on my site (not just visits in which that link was clicked), could I not write ‘type=clicked my link’ into the custom field, which should then carry across all subsequent visits by those users?

    So if I segment all *visits* to my site in which user type='clicked my link’ am I not effectively segmenting those *visitors*?

    Thoughts? Can anyone name common analytics functions that can *only* be accomplished by segmenting visitors?

  39. 43

    AWhinston: Typically you want both options, segment by Visits or segment by Unique Visitors. A very small % of people will productively use the latter, but you want it. :)

    You can make a lot of progress, like wonderful examples you have shared in your comment, with just Visits (or by holding UV as a "constant" in the background and bucketing Visits as in my example in the post).

    You have cleverly hinted at using the custom variable (or "user defined value" as it is technically called) as a way to get what you want, or define uniqueness – and custom variables are available for segmentation.

    Think of it this way. There are reports in Google Analytics (or Omniture or IndexTools or WebTrends) that only reports Visits. But in the rare cases when I want to segment deeper customer behavior I would want to see UV's in those reports (after I group the behavior into a segment).

    A good example is something you don't have as a feature in pretty much any standard web analytics tool today (excluding the DW additional expensive option), the ability to do advanced multi touch attribution analysis. If I could segment by unique visitors that would be a bit easier to accomplish that.

    I suppose your point is well taken. There is a lot you can do with segmenting by Visits and by other means, including user defined. But I would want GA maybe some day far away when wishes come true to also have the ability to segmentation by Unique Visitors, that would be good.


  40. 44

    You've given a lot of food for thought.

    I just opened my first analytics account; I'm still getting used to reading the metrics. So to find out that analytics can go so deep, is a wonderful thing!

    I enjoyed maths in high school, and psychology in university was fun, so analytics is gonna rock :)

  41. 45
    Adam Berlinger says

    Awesome and relevant posting!

  42. 46

    Hi Avinash

    I think the link: NTEN (the “craigslist of non-profits”) goes to NTEN.com . I beleive you meant to say http://www.nten.org not sure, http://www.nten.com takes me to a site with full of Sponsored Ads

    Thank you


  43. 47

    Avinash – yet another great post!

    We're starting to develop a Member Centre area for our Intranet at the British Parliament. Of course we'll get a whole surge of traffic to it the week we launch because we'll make a big splash about it on the corporate newsletter etc (and I'm sure we'll bathe in the warm rays of those early numbers)… but you're bang right about watching what happens to all those visitors down the line. It won't be much of a Member Centre if no one bothers to use it after a month. I know that is a little different to the idea of latent conversions for individual campaigns but it's the same principle.

    The good news for us is we actually know the number of Members we have in our organisation so we can actually see what percentage are actually using the Member Centre on a regular and ongoing basis. I feel a KPI coming on!

    Ok, better stop commenting and actually get on with building the damn thing or there will be nothing to measure! :-)

    Thanks again and keep them coming,


  44. 48

    Great post!

    We ran into this recently with an ecommerce client who was looking for instant purchases.

    Question: I thought that the GA cookies were re-written based on latest visit? So if their first exposure was PPC, they're 2nd & 3rd visits may have been organic. If this is the case, then would that visitor still show up in my PPC segment for the second and third visits?

    Thanks for your help!

  45. 49

    What about cookie deletion. ComScore com up last year (or was it two year ago?) about cookie deletion being up to 40%. I personally calculate a 22% monthly cookie deletion on our web site (base on traffic source (direct + organic than include our url) and new visitor segment.)

    This behaviour may affect result no?

  46. 50

    Sebastien: There is a lot of FUD floating around about cookies. Please check out one of my recent posts about Cookies:

    A Primer On Web Analytics Visitor Tracking Cookies

    In the article I attempt to provide clarity about this issue (including cookie deletion).


  47. 51
    Random_Steve says

    I implemented a potential solution to the unique visitor problem with GA today and then came across this post when trying to determine if I imagined the term 'latent'. The main reason to implement was that my company sells higher priced items that take at least a few visits and some invested time to purchase. With GA measuring visits only, I could not follow the path from somebody googling 'Blue Widgets' to reach our site and gaining site name recognition and then coming back 3 days later by googling 'site name' to make a purchase.

    So, a unique string for each visitor is passed to the user defined value. This is checked against the cookie to make sure the user defined value is not overwritten. Now, when the database reports a new web customer with an order for say $5515.49, I segment by the unique user defined value revenue that matches that order total. This creates a segment for the customer and I can see every historical move made up to the purchase instead of only the current visit.

  48. 52
    Matt Wilcko says

    Can Advanced Segments be used to track latent conversions? For example, if ABC was used as a landing page for a campaign, could I setup an Advanced Segement that counts visitors that hit landing page ABC at some point in the past? I would then apply the segment to our goal pages on our site. Would this work, or would GA only count visitors who went to landing page ABC and my goal page during the same visit?

  49. 53

    Matt: You can't do this with Google Analytics in the web interface. But checkout the free conversion funnel tracking tool PadiTrack.

    PadiTrack uses the Google Analytics API And creates pan-session (across visit) funnels (for same Visitors). And it also allows you to create segmented funnels (which you can't do with GA experience).

    Check it out, it is free. : )


  50. 54

    I love analyzing visitor behavior.

    It is very exciting to try out different internal link text to see click through rates change and become optimized.

  51. 55

    Great post, Avinash. Just wanted to add that if you run an e-commerce site you can also measure purchase latency in Google Analytics if you feed the data into GA via event tracking.

    Basically, you just run a quick database query when the customer reaches the checkout confirmation page to determine how many days have passed between the current order and their previous order. You can then pass the data to GA in a trackEvent call. GA will then record purchase latency and show you the average for your site, or if you combine it with advanced segments, for specific parts of your product range.

    You can use the knowledge of the purchase latency to spot potential tripwire events if you do a similar thing in your CRM. You can then create a marketing campaign to target customers before they defect. Jim Novo explains how to do this in his book 'Drilling Down', which is superb.

    I wrote a little blog explaining how to do this for Econsultancy, which is here. Hope you don't mind me sharing the link: http://econsultancy.com/uk/blog/7860-monitoring-purchase-latency-in-google-analytics


  52. 56
    Lindsay says

    Hi Avinash,

    Is there a way in GA to create a segment based on only those users who visited X amount of times within the given time frame (not over lifetime). I want to create a "Loyal Visit" segment where Visit Number is greater than 2 times within the given month (not that the days since last visit was less than 30).
    I guess this would be a combination of the Frequency and Recency report but I can't figure out how to do it.

    Thank you,

    • 57

      Lindsay: First create a segment using the dimension Count of Visits. In the value column you'll type in 2 (per your comment above).

      Then apply that segment to the report you are interested in, say Top Content.

      Then restrict the time period to what you are interested in, say Jan 1 to Jan 30.

      The data you are looking at should be for people who visited two or more time during that month (and not lifetime).


      • 58
        Lindsay says

        Thank you Avinash. I'm not sure why I thought that would pull in the number of visits over the lifetime of the visitor.

      • 59
        Lindsay says

        I guess a follow-up question would be whether there is a way to determine how many of the "New Visits" became "Return Visits" within a given month.

        User A comes to Site for the first time on April 10 (New Visit)
        User A returns to Site on April 12 and April 18 (2 Return Visits)

        Essentially, how many of the Return Visits were originally New Visits in this same month. Or How many of the New Visits returned to the site within the same month.
        Does that make sense?


  1. […]
    Prompted by Avinash’s post on Recency (if this topic interests you, there is much more here), I have to return to an idea that keeps running through my head:

    Why do so many Marketing people fail to understand the basic underlying dynamics of Interactive / Online Marketing? Relative to the Comments on Avinash’s post, why would Marketers not be interested in the Recency metric? If the Marketers are not aware of it, why would Analysts not push it to them, show them the power of it?

  2. […]
    This list wouldn’t be complete with out a post from Avinash Kaushik. This time he is talking about Measuring Latent Conversions & Visitor Behaviour using Google Analytics and advanced segments. As always, top stuff.

  3. Bloggers Digest 1/23/09 » Ecommerce Blog says:

    In the Web Analytics category: Measure Latent Conversions & Visitor Behavior by Avinash Kaushik.

  4. […] For detailed examples of Visitor & Customer Retention Mapping, see the Measuring Engagement Series.  For examples of how to measure Visitor Recency in Google Analytics, see this post by Avinash. […]

  5. […] Avinash's recommendations to track the value of a customer of time includes using a mix of the visitor recency and visitor loyalty report – but it would be so much more powerful if you could combine this thinking with an accurate traffic source report, not just the last source that drove a user to (re)visit your website. […]

  6. […]
    I know, this can be frustrating for those of you without brick and mortar retail stores, but don’t let it become discouraging. In the process of setting goals, remember to include micro conversions as well. If macro conversions are things like downloading your programs and purchasing your products, micro conversions can be used to measure how many times “contact us” was clicked, or “about us“, or jobs were applied for, or “print this page” was clicked, and the other signs that your site is being used for research, even if nothing was purchased. Someone interested in your site enough to research it is likely to come back, especially if you were convincing enough with your web copy, which will help you calculate your conversion ROI, which is every bit as helpful as the conversion data itself. After all, if you aren’t measuring the metrics that matter, you’re wasting time that could be used on more valuable metrics.

  7. […]
    in the pilot epi, ted sees robin (“the girl of his dreams”) across a crowded bar. the first night they meet, he tells her, “i think i’m in love with you.” that’s just crazy, right!? likewise, how can a brand expect you to “fall in love” them after a 30-second tv spot or a single banner ad? in order to create legen…wait for it…dary relationships with customers, it takes more than a “one night stand” kind of strategy. how about building a community of engaged, life-long customers? obviously, this takes more time and effort as there is no clear-cut, blanket answer, but more likely than not it will prove effective.

  8. […]
    Primarily developed for e-commerce businesses who want to drill down and see who is using different aspects of their site. They’ve built wizard-like interfaces for setting up retention messages based on the people data they collect, which can be sent out using email, or within your app. Mixpanel is a popular option for tracking micro conversions and user behavior, a best practice.

  9. […]
    Let’s talk about some metrics. I won’t pretend to know the right key performance indicators (KPIs) for your business. Every business is unique. Every funnel is unique. But here are several metrics that we here at Stripes like to look at: Cost per acquired visit. Page depth. Visit latency.

Add your Perspective