The Synerise AI Developer Hub

Welcome to the Synerise AI developer hub. You'll find comprehensive guides and documentation to help you start working with Synerise AI API as quickly as possible, as well as support if you get stuck.
Let's jump right in!

Attribution analysis

Introduction

Attribution Analysis gives you more insight into your campaigns. It allows you to see how customers react to your marketing strategy and identify the channels/methods/sources that give the highest conversion, whether it is revenue generated or subscription to your newsletter or any other goal of your marketing campaign. Gaining this insight allows you to adjust/optimize the campaign accordingly, creating more conversion while reducing the costs.

When you open the Synerise Attribution Analysis, you will see a chart and a table.
You may switch between two charts and a transition matrix. The line chart, comparing chosen models in chosen dimensions, as shown below:

The stacked line chart, comparing secondary dimension options within one specific primary dimension:

Or a transition matrix. It is built on the Markov chain concept for attribution. The Markov chain model is described here.
The Markov model for attribution is a probabilistic model, as opposed to heuristic ones (first touch, last touch, position-based, time-decay and linear). Therefor the heatmap of the transition matrix shows the calculated probability values of a transition from one channel to another one.
Below is a simple overview of the steps used to build the transition matrix:
• Splitting paths depending on purchase counts.
• Replacing some of the channels. The channels that are unknown are deleted and direct channels replaced with the previous channel in the path. Two things are important here. The first thing is that unknown channels are not informative, even if they bring conversion. The second thing is that we cannot influence the direct channel, and replacing it with the previous channel doesn’t affect the model because duplicated touchpoints don’t affect the outcomes.
• Calculating the transition probabilities.

The example heatmap of the transition matrix is shown below:

For example we can see that the probability of a transition from Paid Search to LinkedIn is 0.5, or that the probability of a conversion for a customer coming from Facebook is 0.2.

To switch between the charts, you have to choose the desired chart button, beneath the settings button, where the left is the line chart and the right is the stacked line chart.

You can adjust multiple features of the chart and table by opening the drop-down settings menu above the chart or table.
The first thing you can change is the Chart Model, which allows you to switch between multiple models shown by the chart.

The next thing you can change is the Table Models. Here you can add models that will be included in the Model Comparison table.

Another useful option is the possibility to choose the Primary Dimension. The customer conversions are divided into Medium, Source or Medium/Source options. This happens thanks to the UTM parameters in the link through which a customer reaches your site.

The Secondary Dimension option allows you to divide the Primary Dimension into smaller groups, resulting in additional insight.

The last option, Metrics, provides you with a possibility to choose whether you want to analyze conversion, revenue or both. This will have effect on both the chart and table.

By clicking the Channels button, in the top right corner, you can adjust the channel settings.

You can choose which channels will be analyzed.

You can also rename channels or add multiple channels into one group.

The Model Comparison table shows you data according to the options chosen above it. You can sort and filter the results for better readability.
On the screenshot below, in the Metric drop-down menu the Conversions option was chosen,

and here both Revenue and Conversions were chosen.

In order to add a comparison in the Model Comparison table, simply press the Add comparison button and choose the two models that you want to compare. They will be added to the table.

To save the results you can export them either as a PDF or in a CSV format. To do this click the Export button next to the Channels and choose the preferred option.

Synerise AI provides both basic attribution models:

  • first touch model
  • last touch model
  • position-based model
  • time-decay model
  • linear model

as well as advanced data-driven models:

  • Sales funnel model
  • Markov-based model
  • Shapley model

Implementation

To make full use of funnel model you need to provide not only page.visit and transaction.charge events but also product.addToCart events. You can fetch attribution results using Attribution API