Loading ...
Sorry, an error occurred while loading the content.

Re: Delayed conversions

Expand Messages
  • matt.jacobs
    Lars, this is something I ve spent a great deal of time analyzing and testing for nearly a decade now...and it s no less controversial today than it was back
    Message 1 of 13 , Feb 28, 2007
    • 0 Attachment
      Lars, this is something I've spent a great deal of time analyzing and testing for nearly a
      decade now...and it's no less controversial today than it was back then! You have two
      groups to look at
      1. Click-based visitors (those who click and land/convert)
      2. View-based visitors (those who see or are exposed, don't click and land/convert)

      These two groups are important to distinguish because visitor quality and volume often
      differs significantly between them, and when you start conducting tests to fully optimize
      both audience groups from "ad side" to "site side," there are things you'll want to consider
      in your test design for each.

      The basics.... First off, you'll want to make sure you can get pretty granular data - user-
      level if possible with segmented post-click and post-view data. You'll want to review your
      attributions in hourly time increments (you'll be looking at time to conversion from the
      click and the view). Second, be mindful of the consideration/purchase hurdle ($10 vs.
      $1,000). Third, think about the brand - does it have a lot of equity (Nike) or not (Joe's
      Junk)? With all of that in hand and mind, you'll want to analyze the decay curves of your
      post-click and post-view segments by those hourly increments I mentioned above. You'll
      likely find basic similarities in the decay curves, with post-click being much more
      aggressive and immediate than post-view. You'll find your success metrics (say,
      conversion rate) also vary substantially between these groups. And you'll also likely find
      that there are spikes that occur in your post-view data every 24 hours, which is associated
      with user browsing patterns (diluted among those spikes are arguably conversion
      attributions). Once you have your decay curves, you can make some basic determinations
      on the conversion window you want to accept. I've worked with numerous Fortune 50-500
      clients on this over the years and most wind up
      (a) giving post-click attribution credit for conversions within a 1-5 hour timeframe
      (b) giving post-view attribution credit for conversions within 24, 48 or 72 hours. Some
      have a more extended 5 or 7 day windows but not too many.

      For the sake of this string, I didn't go into the granular details. Drop me an email if you
      really want to get into the guts of this - from tracking to the various types of testing you
      can do to prove this stuff out, and more (it's important to note that web analytic tools are
      actually NOT the most appropriate or recommended tools to be using for tracking and
      testing this - they help but they're very limited. People who really understand digital media
      optimization know exactly what I'm talking about - it's very different from "web analytics."
      It's also important to note that assigning 100% attribution credit to the last click or
      exposure is fundamentally flawed and absolutely not advised...don't optimize against that
      unless it's SEM). There's a massive amount of research behind this stuff, including how to
      correctly attribute across digital touch points/marketing channels. Unfortunately, much of
      the research is conducted by agencies with a wealth of experience in this area like Digitas
      and AA-RF, but that research is done on behalf of their clients and the granular details
      aren't out there for public consumption.

      Best,

      Matt (matt@...)
      http://digitalmediaanalytics.com/
    • Lars J
      Thanks for all the insightful responses. Say you are told: Sell as much as you can as long as you do not exceed a cost per order of x dollars per campaign!
      Message 2 of 13 , Mar 1, 2007
      • 0 Attachment
        Thanks for all the insightful responses.

        Say you are told:
        "Sell as much as you can as long as you do not exceed a cost per order
        of x dollars per campaign!"

        Say this happens prior to conversion:

        1) You look at a banner (CPM) - Campaign #1
        2) You click on a different banner (CPC) - Campaign #2
        4) You click on a sponsored link (CPC) - Campaign #3
        3) You convert through an affiliate network (CPA) - Campaign #4

        What is the CPO per campaign?

        I would like to hear how more of you approach this question.

        Lars


        -----------------------------------------------------------------
        In response to:
        http://tech.groups.yahoo.com/group/webanalytics/message/9593
      • Oliver Schiffers
        Hi Lars, In this case It is as it is, you pay for times the campaign, the Conversion just happens one time. This is why you might come back with an additonal
        Message 3 of 13 , Mar 1, 2007
        • 0 Attachment
          Hi Lars,

          In this case It is as it is, you pay for times the campaign, the Conversion
          just happens one time.
          This is why you might come back with an additonal KPI (like looking at an
          eCPM) where you "Sell as much as you can as long as you do not exceed an
          _average_ cost per order of x dollar" over _all_ campaigns.
          Because conversion might not have happened if you would only have one of
          these campaigns.

          Cheers,
          Oliver

          On 01/03/07, Lars J <freshflavour@...> wrote:
          >
          > Thanks for all the insightful responses.
          >
          > Say you are told:
          > "Sell as much as you can as long as you do not exceed a cost per order
          > of x dollars per campaign!"
          >
          > Say this happens prior to conversion:
          >
          > 1) You look at a banner (CPM) - Campaign #1
          > 2) You click on a different banner (CPC) - Campaign #2
          > 4) You click on a sponsored link (CPC) - Campaign #3
          > 3) You convert through an affiliate network (CPA) - Campaign #4
          >
          > What is the CPO per campaign?
          >
          > I would like to hear how more of you approach this question.
          >
          > Lars
          >
          > ----------------------------------------------------------
          > In response to:
          > http://tech.groups.yahoo.com/group/webanalytics/message/9593
          >
          >


          [Non-text portions of this message have been removed]
        • martincwesley
          Lars & Group, This is my first post to the group, and have been watching the threads for some time, and this one appears to be right in my sweet spot. The
          Message 4 of 13 , Mar 1, 2007
          • 0 Attachment
            Lars & Group,

            This is my first post to the group, and have been watching the
            threads for some time, and this one appears to be right in my sweet
            spot.

            The digital technologies that you all have been mentioning on the
            post for some time all rely on the "last click or Last view logic" to
            attribute credit for a conversion. We all know that is a flawed
            methodology as it minimizes the entire effect frequency has in the
            marketing and media mix.

            The other challenge is that typically you will all have more than one
            delivery technology in play with different cookie spaces. For
            example, there has been a lot of reference on this post to Omniture
            as the site analytics and search tool. The challenge is that your
            affiliate and your ad server impressions files are in different
            cookie spaces, and this is what leads to duplicate conversions or
            incorrect attribution of the "credit".

            From a top level you need to:

            1. Sync all the cookie spaces across all your digital channel
            technologies (Email, PPC, Natural, PI, Ad Server, Rich Media Server,
            Affiliate Server, etcÂ…) Then you need to tie that to your Account ID
            and Order/Trans ID system from your CRM or finance.
            2. Once that is done you then need to run logic that looks at an
            individual activity on your site or a subsequent transaction in your
            call center, and analyze that user to everything that they have been
            exposed to in modern time. You will need to store the data at the
            level of time stamps across the entire sync'd cookie/User ID database.
            3. Now you will never really analyze on an individual cookie,
            but what you will do is cluster say your 3 or 4 main customer
            segments, and then you will apply data models that will enable you to
            analyze a group of like customers and what the creative, media and
            frequency of consumption was prior to the conversion event. You will
            need to take into account the type and impact of each interaction in
            this model. For example, a branded search term vs. unbranded. A
            468x60 vs. a 120x120, rich or not rich, interactions or no
            interactions, etcÂ… This is media mix modeling.
            4. Data size for any average customer will range between 20 and
            30Gig of data a day that will need to be normalized and stored
            indefinitely.
            5. Here is an article I wrote on the topic not long ago that
            entails traditional media overlays as well:
            http://directmag.com/disciplines/marketing_right_combination_2/index.h
            tml

            Hope that helps.

            Martin Wesley, BlackFoot Inc.
            martin@...



            --- In webanalytics@yahoogroups.com, Lars J <freshflavour@...> wrote:
            >
            > Thanks for all the insightful responses.
            >
            > Say you are told:
            > "Sell as much as you can as long as you do not exceed a cost per
            order
            > of x dollars per campaign!"
            >
            > Say this happens prior to conversion:
            >
            > 1) You look at a banner (CPM) - Campaign #1
            > 2) You click on a different banner (CPC) - Campaign #2
            > 4) You click on a sponsored link (CPC) - Campaign #3
            > 3) You convert through an affiliate network (CPA) - Campaign #4
            >
            > What is the CPO per campaign?
            >
            > I would like to hear how more of you approach this question.
            >
            > Lars
            >
            >
            > -----------------------------------------------------------------
            > In response to:
            > http://tech.groups.yahoo.com/group/webanalytics/message/9593
            >
          • matt.jacobs
            Lars, not sure what you re trying to get at here. If you have a tracking solution that s mapped/incorporated across those digital touch points, the CPO should
            Message 5 of 13 , Mar 1, 2007
            • 0 Attachment
              Lars, not sure what you're trying to get at here. If you have a
              tracking solution that's mapped/incorporated across those digital
              touch points, the CPO should reflect costs associated with each
              touch point (whether you choose to include production, agency and/or
              operational costs beyond your media expenditures is up to you).

              If you're asking about audience quality, attribution and the
              influence of one touch point as incremental (positive or negative)
              to another, then what you're asking about can be answered in a
              multitude of ways, including but not limited to the following:
              1. Via media mix modeling (I'll spare others the length of that
              reply in this forum)
              2. Via audience value segmentation by channel (OLA vs. sponsored
              link vs. affiliate, etc.)
              3. Via allowance for "portfolio management" of CPO
              4. By combinations of 1, 2, and 3 above.
              5. By user-level analysis (which can be done within #1 above if you
              have the right data - time consuming but it answers your question
              and many others more completely than other methods)

              Each approach is different, but all will yield data that highlight
              the - or give a proxy for - different "quality" of users from each
              channel. With enough data you can model what the LTV might be for
              each, your optimal mix and what your CPO ranges can be for a given
              channel and user type.

              ~Matt

              http://DigitalMediaAnalytics.com
            Your message has been successfully submitted and would be delivered to recipients shortly.