Here I would like to draw attention to the following. At the end of the article, Peter (and not without reason) admits that his method of calculating associated conversions is incorrect:
The weak point of my segment is the exclusion of direct traffic. In fact, by excluding it, I exclude the paths konkurenty_search — direct (none). Conversion on this path has already been attributed to the konkurenty_search list of japan cell phone numbers campaign in other reports. Therefore, it is incorrect to consider it auxiliary, but together with this segment I exclude such paths as: konkurenty_search — yandex / organic — direct (none). In this case, the campaign was an auxiliary stage and it should be taken into account.
Due to the relevance of the issue, in this article I would like to describe the way in which we at 5 o'click calculate the number of associated conversions.
So, let's say we need to calculate how many associated conversions the y_celevie_rus_poisk campaign brought in April 2015.
We know that the number of associated conversions in the corresponding report is not entirely correct, since it takes into account chains where this campaign was the last indirect source (the same chains that are taken into account in standard Google Analytics reports), i.e. chains of the type y_celevie_rus_poisk — direct (none), and chains in which this campaign was both an auxiliary source and the last one, i.e. chains of the type y_celevie_rus_poisk — google / organic — y_celevie_rus_poisk.
Therefore, to find the number of correct chains, you need to subtract the number of chains of the first and second types from the number of associated conversions in the report.
Step 1 - Go to the standard "Source/Channel" report and see the number of conversions for this campaign:
We know that this report attributes the conversion to the last indirect source. Therefore, this report takes into account, for example, the following sequence chains:
As you can see on the slide, in April 2015 there were 10 such chains for the “Order” goal.
Step 2 - Go to the "Assisted Interaction Analysis" report, set the maximum number of days in the retrospective review window, select the desired campaign and conversion type and view the report:
We see that in April the y_celevie_rus_poisk campaign became the last source of conversion 9 times. That is, these are chains of the following type:
We also know that all such chains are taken into account in the standard Source/Medium report, therefore, the difference between the number of conversions from the standard report and the number of conversions by the last or direct interaction is the number of chains that ended with a direct visit, and where this campaign was the last indirect source, i.e. chains of the type:
These are exactly the chains that we need to subtract from the number of associated conversions. In the example under consideration, in April 2015 there was only one such chain (10 − 9 = 1). Therefore, in April 2015, there were 6 associated conversions for the "Order" goal without such chains (7 − 1 = 6).
Step 3 - Now we need to subtract the chains in which this campaign was both an auxiliary source and the last one. To do this, in the "Analysis of auxiliary interactions" report, we overlay the segment: last interaction y_celevie_rus_poisk
And we look at how many associated conversions for this campaign were in this segment:
This number will be equal to the number of chains in which this campaign was both a secondary source and the last one.
Step 4 - Subtract it from the difference obtained in step 2, and we get the required number of associated conversions in April 2015 for the y_celevie_rus_poisk campaign (in this example it is 5).
Unfortunately, this calculation method is not suitable for daily reporting, since the data in the multi-channel sequences report is received with a delay. However, it can be useful for weekly planning meetings.
This method can also calculate not only the number of associated conversions, but also their value. Usually, such reports are more interesting for clients — reports related to money. The algorithm is absolutely the same. Such reports are especially useful when deciding to disable a particular source/channel. So, in our example, in a regular report, we see that posting on the site tema.livejournal.com brought 2 transactions at a price of 15,715 rubles:
