Friday, April 10, 2015

Linear Allocation (visit based) and Exaggerated Numbers while using Calculated Metrics - Exact Reason

Adobe Analytics// Issue with calculated metrics!!!
Wrong calculation when one of the value used in the calculation is zero. For example: Whenever the order value is zero and revenue has some value, the AOV calculated metrics divides revenue/orders. Ideally the output should be zero; it throws up wrong value here.
I saw this across Client RSIDs; whenever the order value is zero and revenue has some numbers, the value is calculated wrongly.
Need help!

Answers

02:19 PM
Hi Kamal,
Standard metric attribution for the Site Content> Pages report is linear therefore (see reference info linked below) even though it appears the Orders metric is 0 in the report it actually is a small fraction. Therefore when the calculation is made you have a very high # (Revenue) divided by a fraction (.10 fraction values of Orders for example) which produces a very high 'non-sensical' value. You can verify yourself by creating a calculated metric for orders that allows for decimal places. If it were the case that the Orders metric was truly zero than your calculated metric value would be an error DIV/O since you cannot divide a number by zero.
What analysis question are you trying to evaluate Site Content> Pages? My main question would be what is the intent behind using AOV against linear allocation metrics within the pages report? Since linear allocation is variable depending on the count of pages within a given visitors conversion visit it is very likely not what you are looking to review.
Reference Link:
When linear allocation is selected, success events are evenly divided across all variable values seen in the visit. For numeric and currency events such as Revenue, the monetary amount is divided. For counter events such as Orders, a fraction of the event is awarded to each variable value in the visit. These fractions in reporting are summed, then rounded to the nearest integer in reporting.
For example, in a visit where four pages are visited prior to a success event, each page would receive credit for 25% of the event. If, in the same visit, campaign had two values, each campaign value would receive 50% of the credit for the event.
Best,
Brian

REPLY
    01:15 PM
      Dear Brain
      Thank you very much. Great answer!!!
      We are trying to understand the effectiveness of each page that contributes to the conversion. I had recommended participation revenue metrics instead of measuring this way. Believe even participation metrics are based on per visit, and perhaps a better one.
      Do you want to suggest on any other way to understand the role of each page in producing conversions.
      Best Regards
      Kamal Chembath
      REPLY
        01:18 PM
          Dear Brian,
          Cross Visit Participation should work across visits right. Will this be an ideal choice for the purpose.
          Best Regards
          Kamal Chembath
          REPLY
            02:10 PM
            Hi Kamal,
            Participation metrics are great for analyzing site content influence on conversion events since it attributes whole end conversion event values to a given page over a visit. I would recommend leveraging the Orders and Revenue metric with participation.
            Best,
            Brian
            REPLY
              02:49 PM
                Dear Brain
                Thanks a ton . One more final clarification.
                Does regular participation metrics traverse more than a visit or only cross visit participation keep track of multiple visits?
                Best Regards
                Kamal Chembath
                REPLY
                  02:53 PM
                  In R&A participation metrics are tied for a given visit only. If you have access to Ad Hoc Analysis (formerly Discover) there is a Visitor Participation metric option:
                  Best.
                  Brian
                  REPLY
                    11:09 PM
                      Thank a ton Brian, very helpful answers.
                      -Kamal Chembath
                      REPLY

                        No comments:

                        Post a Comment