How AXA Assistance separated the Good from Bad and Ugly Affiliate Publishers with Roivenue Attribution Solution:

"Separating affiliate partners who bring real value from those who are only cannibalizing your marketing budget can be a nightmare. Especially if you want to compare their real ROI, not only among themselves but also to other performance channels, and take the whole customer journey into account."

That’s why AXA Assistance decided to use Roivenue Attribution Solution to shed a light on the matter.


The Challenge

Everybody who uses the services of affiliate partners is aware of their threats to a marketing effectiveness evaluation. Often, less serious affiliate partners can leach on the performance of other channels by showing ads to customers who are very likely to make a conversion regardless the affiliate ad.


An ideal affiliate partner will bring new customers to your website without any help from the other channels. Such a publisher deserves the commission fee that he claims for your conversions. If the publisher steps into already existing customer conversion paths, he should deserve much less credit. But how much exactly?


To make the matter worse, reported data from the affiliate provider often serves results in bulk and with no regard for attribution. The client has built a previous analysis that has revealed that a deeper look into the issue would be needed in order to produce proper publisher evaluation. Also at this point, it became clear, that a data-driven attribution based solution could be helpful in this case, as it would account for the entire customer conversion path.


This is a typical scenario which could be solved by Roivenue Attribution Analysis, a feature which is built in Roivenue by default. But in this case, this feature could not have been used, as there have not been enough conversion paths related to affiliates, which means that the built-in algorithm would not have enough data to make its calculation reliable.


Nevertheless, this did not stop us, and to solve this problem for our client, we have designed a custom solution, which is based on Roivenue Attribution measurement, but calculates and presents the data in a different fashion.




Roivenue Attribution solution was used to capture customer paths across the whole marketing mix. All touchpoints and their costs were automatically loaded to the system to be transformed and later exported for further analysis.


As a next step, we analyzed the mapped journeys and divided all paths that included an affiliate touchpoint into three categories:



All touchpoints in the path are related to an affiliate – optimal solution.



Path 1 



An affiliate has attracted the customer, but the customer later came into contact with other paid campaigns. This situation is sub-optimal.

It is valuable that the affiliate has caught the customer for the first time and that the customer later proceeded to visit the conversion website, but the fact that there has been contact with other paid channels signals that the message of the affiliate has not been strong enough to close the customer by itself. This increases the cost of the path as other non-affil channels need to take action.


14-984Path 2 


An affiliate has entered a conversion path after paid channels. In this case, the affiliate’s value to the whole marketing mix is highly questionable and it is correct to assume that this affiliate publisher cannibalizes on other paid channels in the mix.


Path 3


In order to be able to divide customer paths into the three described categories, we first had to map these paths.


This is where Roivenue Attribution measurement comes into play, as it was designed for reconstruction of paths and data driven attribution modelling.


All traffic has been divided into channels of two types:

  • Traffic from affiliate partners
  • Traffic from other channels – paid channels (such as AdWords or Facebook) and unpaid channels (such as organic and direct).

All of these other channels – both paid and unpaid – fall into the paid channels group in this article.


Our next step was to gather data for one month, to collect enough customer paths for the algorithms to produce reliable results – the more paths we collect here, the higher is the precision of the evaluation that follows.


After the data gathering period, collected material was exported to Power BI for the final analysis.


Its outcome can be seen below.


 Publisher overview OK    Publisher evaluation TEST         



In the bottom left corner you can see an overview of active affiliate publishers and division of paths related to them based on the description above.


On this division is then based the calculation of Publisher Quality Index, which quantifies publisher’s tendency towards paths of type 3. This index represents reversed value of the division of count of type 3 paths to the sum of counts of paths type 1 & 2. The Publisher Quality Index (PQI) for publisher x is defined as follows:16

The higher the value of this index, the lower is publisher’s tendency towards paths of type 3 and thus higher is his perceived quality.

Publishers are evaluated and ordered by total paths related to them in the table on the bottom right. In the default setting, this dashboard considers both conversion and non-conversion paths. Filtration on only conversion or only non-conversion paths is possible via slicer on the top right – “true” selects only conversion paths and “false” selects only non-conversion paths.


It is finally possible to compare publishers between themselves through optics of their quality and decide if cooperation with lower quality publishers should be retained or stopped.


Not cooperating with these publishers would lead to decrease in marketing costs while performance of the marketing mix as a whole is retained, as majority of conversions related to type 3 publishers will take place even if they are not present in the mix.


The user of this analysis should consider the value of PQI below which the publisher is low quality – our suggestion is 0.6, as this value already signals strong tendency towards type 3 path. It is also important to view value of the index for each publisher with accounting for the number of paths related to the publisher – the more paths, the higher reliability of the index.


After the initial setup of data collection account in Roivenue, this analysis can take place periodically, so the behavior of the retained and new publishers can be continuously monitored and cleanliness of affiliate portfolio can be maintained.




Results and Takeaways 

AXA had been struggling with evaluating the quality of different affiliate publishers for years. 

The client uses Roivenue Data-driven Attribution Optimisation suite for managing and optimizing its budgets in several countries.

Due to the lack of enough data the standard algorithms couldn’t be used to optimize affiliate partners as there is too many of them and not enough conversions is happening.


That’s why Roivenue offered a customized analysis stemming from the knowledge of attribution paths.

Which happened as follows:

  •  Inputs: Roivenue account, Google Analytics free, measurement setup in Google Tag Manager
  •  Collection of data in Roivenue, mapping of customer paths
  •  Division of affiliate paths based on their structure into three groups
  •  Calculation of Publisher Quality Index, its low value signals risky behavior of the publisher
  •  Retention of only high quality publishers, based on their PQI
  •  Ongoing monitoring and evaluation, clean publisher portfolio ensured


As a result, AXA saw an increase of ROI of its affiliate program as well as of the entire marketing mix.


If you want to know more, check out our case study about how AXA increased ROI on affiliates by 180%.


Topics: Case Studies

Subscribe to Newsletter

Recent Posts