We believe that in order to make the attribution modelling as effective and applicable as possible, it is necessary to customize its settings. This is usually done after the initial verification of conversion paths and it includes several configuration parameters. The most important one is the choice of user path lengths.
When we configure attribution models for the first time, we always run the audit of user paths. This is done in order to see what are the usual journeys that customers undergo, prior to any conversion and to select the most effective approach to user paths will be most effective.
Let’s imagine the following situation. There are three users that went through the following paths:
Markov algorithms may be based on an assumption that the analysed paths end either with a conversion or with last visit, without any conversion (at least so far). As a result, we can use one of the following approaches. Each has its pros and cons, and each should be selected according to customer journey types.
Short paths
This is the simpliest approach in which we focus only on those elements of paths that finish with the first conversion. In our case this will lead to the following paths analysed by the model below:
This approach benefits from the precise modelling of how channels influence the first user conversion, and diminishes customer retention and all marketing activities focused on bringing customers back to the website. It works best when the majority of marketing efforts are focused at acquiring new customers, the brand is not that strong and customer retention is either not existent (because of business characteristics) or not that important.
Long paths
The second method is all about focusing on each conversion separately and including all touchpoints that happened prior to this conversion, disregarding the fact that some sessions may be calculated twice or more times (for each conversion).
In this case our user #1 will be split into three paths, each used for the calculations:
Please note, that in this approach instead of 3 paths (user 1, 2 and 3) we will have 5: user 1a, 1b, 1c, 2 and 3.
A very strong advantage of the abovementioned solution is that it is based on using all steps on every customer journey. In other words, traffic sources that happened at the beginning of user path, benefit from all conversions that happen afterwards. E.g. if a user is first acquired by Google Ads search campaign, Google Ads will get credit for all conversions made by this customer in the future.
In fact, this feature may be perceived as both – advantage and disadvantage, depending on the business needs. It is an advantage when the real influence of retention channels is small because of strong brand values and customer loyalty. So if we know that the brand is so strong, that the customer will return to the shop without the need of any targeted campaigns.
On the other hand, this model should not be used if customer journeys are long, even between conversions. It should not be used when customers make the decision about second or the following purchase in many days or weeks, and require several touchpoints.
Segmented paths
The last approach to paths is the one that may seem the most straightforward, that is we split paths so that each contains only one conversion. So here our User #1 will be split into the following paths:
This approach is usually chosen when all customer journeys are always long no matter whether they are the first or returning ones. So in other words it should be used when it requires some touchpoints to bring a customer back on the website.
As you can see, working with paths is not that straightforward as it might seem. That is why we set up every project using individual settings that depend on our initial analysis.
