Web Analytics Sales Attribution model

Lunametircs's Robbin mentions that Web Analytics sales attribution has three models, namely (first interaction, average distribution, first interaction and last) model, but it is actually four models, ignoring the last interactive model by default, Most Web analytics tools and web analytics models present the credit for sales conversions to the final interactive channel and source (last interaction model), of course, the end of this is not necessarily true in the end, when the final interactive channel is a direct source, the credit will be attributed to SEO,PPC, Or the source of the referral, it can not find other people's situation, the good thing is attributed to direct traffic.

For example, take a look at the picture below, the natural search will get all the credit for this conversion, but in fact, the first time visitors use paid ads to find the site:

For a more understandable example, the European Cup is now in the hot stage, assuming that one of England's goals is done together by the following players, as follows:

There is no doubt that the goal was recorded on Wayne Rooney's head, Gerrard received an assists, but in order to better measure the player's performance, need to analyze who planned the attack, who participated in, so there are four of the previous analysis models mentioned.

These website analysis models, Google Analytics Advanced version can be achieved, said Google Analytics (Google Analytics) Advanced Edition can be the marketing channel to interact in the order of the sequence clearly (Google Analytics Advanced version is 150,000 U.S. dollars a year, the main feature is the sales of conversion channel attribution , Web analytics Services and training, data not sampled). But in the world, few money-poor owners can use the Google Analytics Advanced edition, How should deal with this problem?

In Google Analytics, the solution's data source comes from the following report: Transformation > Multichannel paths > Popular conversion paths.

Verify that the most important transformations are selected. Then, export all conversion paths greater than 0.

Remember that after you export the data, the recommended format for exporting is. csv, opened in Excel, as shown in the following illustration:

In Excel, use the data > columns, and select the separator "> symbol", and then you can get the formatting you want.

Take a look at the picture below:

Note: This produces a lot of space, can confuse the data, so you need to use the search and replace function to remove the space.

Model One: Last interaction

Google analytics E-commerce Tracking By default is the last interaction model, here is not explained in detail.

Model two: the first interaction

If you want to credit all the credit to the first interactive marketing node, you only need to sort and summarize the first column and convert the value rollup, and the following figure is the result of the natural search rollup.

You can also get the first interactive data through the Google Analytics (Analytics) Multi-channel path report, select Secondary transformations, and then select the first interactive analysis. Now this is done through Excel because it is easier to compare with other models.

Model three: Average distribution

If you want to divide weights into all the marketing nodes that are involved in the transformation, you need to do some challenging work.

1. It needs to be clear how many times each channel is in each transformation. (You can use such a formula: = COUNTIF) Range, "direct" – this tells you how many times "direct" has occurred in the transformation. Similarly, the second parameter can be replaced by natural search, email, paid search, etc.

2. Then distribute the values in the conversion proportionally, taking into account that some channels may get more weight because they appear more frequently in the conversion path. In the first picture, we see that paid search appears once, so it gets the weight of ¼. Natural search appears three times, so it gets the weight of ¾;

3. You can use such a formula to calculate the value of the Channel: = (the number of channels to appear/total channels) * Total conversion value;

4. Here is a screenshot, the blue box represents the count, is this channel is 1, not on the 0; the red box is the conversion value.

Model four: First and last interaction

This is the Google Analytics advanced version of "location-based" customization, can be customized according to their needs. Here are the weights assigned to each of the first and last interactions 50%:

1.5% Credit to the first interactive marketing channel;

2. To find the last interaction, you can use such a formula: = LOOKUP (REPT ("z", 255), range), note that this Excel function can only deal with English, can not handle Chinese, so if you deal with Chinese, the channel to replace the Chinese to English;

Another formula is also OK, INDEX (Range,match (REPT ("z", 255), range)--only need to replace two places where the range appears;

3. Give another 50% value to the last interactive click.

Summary of Sales Attribution model

Personally believe that the sales attributable to the Web site analysis of a more difficult problem, the reason for sales to attribution, because the channel from the performance of the distribution of resources, and the allocation of resources is very valuable, is a very important part of the marketing strategy.

For large E-commerce sites, if a site has a variety of traffic channels, then you will encounter the complex problem of sales attribution. But if a variety of channels to implement a separate site, is another way to play, such as the establishment of ad-hoc sites, specifically to create e-mail sent pages and so on.

In short, the above four kinds of Web site analysis model is to solve a site with a variety of traffic channels and put forward, I hope to help.